Every instinct you have says that’s backwards.

Is it real? Is it a modeling artifact? Or is it a glitch…in the Matrix?

Is what I’m seeing real, and if so, why? That one question is a recurring theme that runs through everything we’ve been doing on a client’s strategically important CCS project. So, grab a coffee (or take the red pill), and let’s see how deep the rabbit hole goes.

Why CO₂, and why now?

CCS (carbon capture and storage) has quietly stopped being a corporate sustainability talking point and started being an engineering backlog. The Global CCS Institute’s Global Status of CCS 2025 report counts 77 CCS facilities now in operation globally, capturing and storing around 64 Mtpa of CO₂. This is up from 50 facilities in the previous year. With a further 44 Mtpa of capture capacity currently under construction, and operating capacity projected to grow roughly fivefold by 2030, the growth trend in the CCS industry is clear. The reason CCS matters is simple: a whole tier of the economy (i.e. cement, steel, chemicals, gas processing) can’t be decarbonised through electrification alone. For those hard-to-abate sectors, capturing the CO₂ and putting it back underground isn’t one option among many. It’s more or less the only feasible one on the table if we are serious about cutting our emissions.

A consequence of this advancement is that plenty of engineers who’ve spent their careers keeping hydrocarbons flowing are about to be handed a very different fluid. We were invited to the SLB OLGA CCS Roundtable earlier this year as specialists with first-hand experience in modelling CCS, to present our findings and share what we’ve learned working with OLGA and its new NEO (Olga for New Energy Operations) solver, developed specifically for CCS systems. This piece is a distillation of what we brought to that table.

Same discipline, a very different fluid

Most of our readers are familiar with the traditional oil and gas playbook: multiphase flow where gas, liquid and water coexist and much of the job is managing that coexistence. Multiphase is a given; you design for it.

CCS flips the emphasis. For long-distance transport of CO2, the goal is to operate in the dense-phase and the primary reason is commercial: it packs liquid-like density into the pipe, so a single line can carry the tonnages of CO₂ a project needs that it never could as a gas. Therefore, for a CCS export line, the whole point is to stay single phase. The moment it splits into two phases in a line that wasn’t designed for it, you inherit the entire hydrocarbon problem list (plus a few extra quirks unique to CO₂) with none (or very little) of the tolerances to absorb it.

Here’s the part that makes CO₂ genuinely tricky, and it’s worth one good analogy. Picture a knife-edge mountain ridge. Out on the broad lower slopes where hydrocarbons live, you can walk ten metres in any direction and your altitude barely changes. The ground is forgiving. Now walk up to the razor ridge at the summit: one step left and you drop hundreds of metres into the liquid valley; one step right, the same plummet into the gas valley. A single pace is the difference between two different worlds.

CO₂ operates right on that ridge, near its critical point (about 31 °C and 74 bar), where small changes in pressure or temperature can lead to dramatic changes in its properties. A model that’s “close enough” on the slopes can be catastrophically wrong up on the ridge, and the ridge is exactly where our fluid lives.

Get the fluid right first

So before we run a single simulation, we get the fluid model right - the EoS (equation of state) that tells the simulator how CO₂ behaves at every pressure and temperature it might see. Choose well and everything downstream stands on solid ground. Choose poorly and you’ve built a very expensive, very confident incorrect answer.

There’s a whole family of EoS to pick from, from fast, familiar workhorses (eg. RKSA and CPA) to high-accuracy models built for CO₂-rich mixtures with very specific sets of components (eg. GERG-2008, EoS-CG). On the gentle slopes (i.e. far from the critical point) they’re practically identical. However, up on the ridge they can potentially part ways. It’s the least exciting decision in the study and the one most likely to save a project. A quieter choice sits alongside it: feed the simulator a pre-computed property table, or make it solve the thermodynamics on the fly. We tested both - the fast option held up here, the gap only opening at very low flow rates.

Enter NEO

Which brings us to the guest of honour, and the reason our roundtable deck was called Into the Matrix.

NEO is the next-generation solver for OLGA, developed specifically to handle CO₂, and in our experience, it’s been a genuine step change. Modelling a fluid this non-linear used to feel like trying to dodge bullets in slow motion. Numerical instability, regular crashes, and questionable results operating up on that ridge. NEO in conjunction with the new CO₂ flow model (activated in OLGA by SOLVER=NEO, FLOWMODEL=OLGA, CO2=ON), handles the unique characteristics of CO₂ with a stability and accuracy the older approach struggled to match. The difference between a model you trust and one you babysit.

The NEO solver is still in its early iterations and being such a new feature, it isn’t flawless, and pretending otherwise helps nobody. It is for these reasons that collaborating at industry events like the Roundtable that we presented at are important. At this event, we highlighted three specific examples from our own experience with NEO to follow up on: slugging in the dense phase, occasional non-convergence with the steady-state pre-processor, and flow instability at very low flow rates. None are a showstopper, and all are the kind of things you only find by pushing a numerical model hard against a real, challenging CCS problem.

NEO may be The One for CCS modelling, it’s just not done training yet (are any of us?).

Reading the Matrix: seeing the code behind the plot

Now assuming your fluid model and simulator have got the physics of living on that ridge correct, the final challenge is to interpret the results that are output from your simulations. If you’re coming from a traditional flow assurance background and familiar with multiphase hydrocarbon systems, the results for a CCS system can at times appear quite counter-intuitive and just plain wrong. This is where that red pill comes in.

For the analysis of hydrocarbon flow assurance, we mostly live on two kinds of charts: trend plots (values over time) and profile plots (values over distance along the pipe). They’re the right tools for oil and gas. For CCS, take the blue pill; keep using those traditional methods and the story looks calm, right up until a quirk you never saw coming bites you. This is because those plots don’t show where the fluid sits relative to the boundaries that matter - that ridge.

The red pill is to plot your operating conditions straight onto the phase envelope, so you can see at a glance how close you are to the bubble line, dew line, critical point, Widom line and hydrate curve. These results are then presented on a pressure-temperature (PT) plane that the system operates in, or for a more rigorous but perhaps less intuitive approach, on a pressure-enthalpy (PH) plane. However, we go one better, with what we call a PTH plot: overlay enthalpy contours across that pressure–temperature diagram, and now pressure, temperature and enthalpy all live in a single pane. The Nebuchadnezzar’s dashboard, if you like.

Remember the operators aboard that ship, reading the Matrix straight off the screen? Cascading green code to the uninitiated, “blonde, brunette, redhead” to them. A PTH plot is the same. A tangle of lines to those not familiar with this presentation method. However, once you’ve learned to read it, you see exactly where your fluid is and how much room it has before it does something you didn’t ask for. Free your mind, and the plot stops being a plot. There is no spoon.

Three glitches in the Matrix

Read the data this way and the “glitches” (results that make you doubt your model) resolve into explanations you can point at. Three specific examples we have encountered:

1. Hotter fluid when it’s colder outside. The mystery we opened with. Drop the ambient temperature and the CO₂ arrives at the bottom of the riser hotter. As the dense, heavy CO₂ descends the riser it compresses under its own weight, and with the rapid compression, heats up. That part is as expected. The twist is how much: a colder environment lowers the operating pressure in our system, and at lower pressure CO₂’s Joule–Thomson coefficient is higher. In other words, the temperature is more sensitive to a change in pressure. Same descent, bigger temperature rise. Cooler outside, warmer inside. Our PTH plots clearly show this phenomenon.

2. The −13 °C plunge. Late in field life our reservoir pressure reduces and tubing pressures are low enough that the fluid runs two-phase in our well tubing. CO₂’s Joule–Thomson coefficient is far larger in two-phase than in dense phase, so the temperature drop across the choke turns severe (rapidly plunging into sub-zero temperatures), once the operating point crosses the boundary. Again, the PTH plot shows the crossing; the moment we cross into the two-phase region and temperatures plummet. Real physics, not an artefact.

3. Flow-rate spikes. A flow-rate profile that jitters during steady-state operation for no good reason at all looks alarming. However, this one is numerics, not physics. As the fluid crosses the Widom line, OLGA flips how it labels the phase; all gas on one side, all liquid on the other, and that switch can introduce a discontinuity. Plot the operating point against the Widom line on a PTH plot and the spikes line up with the crossings.

The common thread with the three examples above? Knowing where you sit relative to the thermodynamic boundaries turns weird into explained. Sometimes that’s elegant physics; sometimes a numerical wrinkle to chase down. Either way, you can’t easily tell them apart until you can read the Matrix.

What we took away — and where we’re taking it next

Three things we’d write on the whiteboard. Get the fluid model right before you run anything; the cheapest couple of hours and the most expensive to skip. Visualise where you’re operating, not just what the numbers are. And treat NEO as the real step forward it is, while keeping a running list of its current limitations and known issues.

Thanks to all participants at the roundtable. Conversations like the ones held are how the industry’s tools get more features, faster and more stable performance, and most importantly, greater accuracy.

If this is your kind of thing, come and chat with us in person. We’ll be presenting on CCS flow assurance at the SPE Workshop “Beyond Boundaries — Advancing Cost-Effective and Sustainable Solutions in Flow Assurance,” 8–9 September 2026 in Kuala Lumpur, Malaysia. Fittingly there will be a session at the workshop dedicated to flow assurance for energy-transition fluids and CCS. If you’re there, we’ll be the ones talking about knife-edge ridges, cascading code and the absence of spoons.

Lots more to come. Stay tuned, and free your mind.

If you made it this far, thank you. Subscribe to Data and Discipline for more from the Pontem Analytics team on CCS, flow assurance, production chemistry, data analytics, and the occasional glitch in the Matrix.

Subscribe now

Adoption is climbing, people are genuinely faster, and the feedback from the team is good.

That is exactly the problem.

What you are seeing is real, but it is individual productivity, and individual productivity is the part you get for free. The thing that actually changes how a business runs sits one level above that, and reaching that is much more complicated.

What follows are the five things we consistently see companies struggle with when they try to make that jump.

These are not the generic warnings about hallucinations and data privacy that you have already read a dozen times. They are the ones that show up in asset-heavy operating businesses, where the answer AI hands back ends up inside an invoice, a nomination, or a decision that somebody has to defend later.

To keep it concrete, we will use a midstream gas plant as our running example, and one question that gets asked every single morning:

The part you get for free

Everyone reading this has used AI plenty of times by now.

You may not remember the details of the first time, but you probably remember the reaction, which was likely close to “wow”.

AI is an undeniably impressive technology. As Steve Jobs would have said, it just works. You throw some random question or task at it and it typically hands back a well crafted response, a code snippet, an image, or nowadays even runs a full workflow for you.

The chat interface is a big part of why. It encourages you to experiment, to test it, and ultimately to develop an intuition for it. It often encourages you with positive reinforcement (“great question, Steve”). In a work setting that often starts as simply as asking AI to explain something to you, we find an early use case we wrote in 2023 when ChatGPT was just taking off and called “Google Plus”. It expands quickly from there.

And you get all of this “for free”. Not literally free, of course, because AI compute is expensive (and getting more expensive). But once an organization has made the investment to get AI-enabled, that individual productivity gain shows up on its own, without any additional training or infrastructure, and it shows up very quickly.

It is a true step change, and it is absolutely worth having.

It is also where most rollouts quietly stop.

The second tier, where AI becomes something the enterprise depends on rather than something individuals happen to be good at, is a completely different problem. Below is where it tends to come apart.

1. You will mistake individual productivity for enterprise capability

Six months in, the numbers look great. Licenses are being used, people report real time savings, and a few teams have built genuinely clever workflows for themselves.

None of that - on its own - means the business has changed how it decides anything.

Individual gains are additive. Enterprise capability is structural. The first does not turn into the second on its own, and the early success can hide the fact that the structural work has not started yet.

What this looks like: five people can now get an answer to “how is our plant operating today” in about thirty seconds, and no two of those answers match.

2. Your most important information is not in a form AI can use

The operating record of the business lives in daily production reports, pig reports, trucking tickets, allocation statements, incident reports, and email threads.

AI can read any one of those beautifully. That is not the same as being able to answer a question that spans a year of them.

Until that information is extracted and stored in a structured way, every question is answered from whatever happens to be within reach at that moment, which is usually the most recent document rather than the most relevant one.

What this looks like: answering the plant question properly means pulling four separate reports, and all four of them arrive as PDFs and email attachments.

3. The same question will get different answers depending on who asks

This one deserves a closer look, because it is the failure that does the most damage.

Two people ask how the plant is operating today.

The first, who is less technical with their prompts, gets back: “The plant is running normally with no major issues reported.”

The second, who is more technical, asks the same question and gets: “Inlet is 82 MMscf/d, recovery is 91%, and residue is on spec.”

The second answer feels much better. It has numbers in it.

But neither answer is verifiable.

Where did 82 come from, and is that gross or net of fuel and flare? Is 91% ethane recovery or total liquid recovery? Is “on spec” coming from the online GC, and was that GC frozen this morning like it was last Tuesday?

Ask again tomorrow and the numbers may move for reasons nobody can reconstruct.

This is not the AI being broken. It is the AI being asked a question that the business itself has never defined a single answer to.

What this looks like: two confident answers, both defensible, neither reproducible.

4. Nobody owns the business logic

Most companies assume the AI platform or the implementation partner brings the business logic with them.

They do not. They bring access.

The actual calculation, the one your contracts, your counterparties, and your board depend on, is yours to define. Somebody has to decide what recovery means at this plant, which meter is authoritative, and how to handle the days when a value is missing or frozen. Codifying this (i.e. putting a number/formula in-place to define what “good” means) is harder than it sounds.

That work is unglamorous, it is nobody’s favorite project, and it is almost never assigned.

What this looks like: ask who owns the definition of recovery for this plant and every answer points at somebody else.

5. The last mile is IT, not AI

The pilot works. Everyone is impressed. Then it needs to run in production.

Production means identity and access control, network boundaries, a live connection to the historian and the accounting system, an audit trail, and somebody on the hook to support it at 2am.

That is where a surprising number of promising pilots stall out, or quietly revert to the spreadsheet the pilot was supposed to replace.

What this looks like: the demo answered the plant question perfectly using data someone exported in March (potentially in a sanitized, structured, and static way for a pilot test), and three months later it still cannot reach the historian on its own.

What all five have in common

Every one of these comes back to the same root cause.

There is no single authoritative answer sitting underneath the chat window.

When someone asks how the plant is operating, the AI is not retrieving the company’s answer, because the company does not have one written down anywhere. It is assembling a plausible answer out of whatever it can reach at that moment.

That is fine for an individual. It fails at the enterprise level for two reasons:

• The process is slow and never fully automated, because a person has to keep prompting until the output looks right

• More importantly, the process is not repeatable or reproducible, so the same question can produce a different answer tomorrow with no way to explain the difference

Mark Cuban has written about these same issues himself.

And the stakes are not academic.

Nobody actually makes a decision on “how is the plant operating today”. They make decisions on whether to reject ethane, whether a pig is going to make it through, and what this month’s margin looks like. Those answers become invoices, nominations, and commitments to counterparties.

A plausible answer is not the same as a right one.

A right answer you cannot reproduce tomorrow is not a business process.

What has to be true instead

Two things, and only one of them is new.

First, the critical business logic has to live outside the chat window.

One definition of recovery. One authoritative meter. One agreed way to handle a frozen GC. Written down, deterministic, tested, and owned by someone.

This is the familiar part. It is business as usual, or business as it used to be. Those processes get built the way they always were, and they get exposed through interfaces so other systems can call them consistently, most often through what are called APIs, or Application Programming Interfaces.

API’s enable systems to communicate with each other

Second, AI has to be able to reach that logic.

AI capability gets extended in roughly four ways, from the simplest and most limiting to the most capable:

• Prompting and reusable prompt templates

• Custom instructions and shared knowledge bases

• Retrieval across company documents and data

• Direct access to the company’s own algorithms and systems

The first three are individual level moves. They are useful, they can be shared across a team, and they will never make an enterprise critical answer reproducible.

Only the fourth does that, and today it is typically implemented through something called an MCP server, which is just a standard way of telling AI what your systems can do and how to call them. That standard will change, probably more than once. The requirement behind it will not.

What you get once that is in place is not really an AI benefit at all. It is the benefit of having a governed source of truth:

• Consistency and reproducibility. The answer to “how is our plant operating today” is the same regardless of who asks or how they phrase it.

• Reusability. The logic is built once and shared across teams and workflows, instead of being reinvented by every individual in their own chat window.

• Control and trust. Access to sensitive processes and data runs through a defined interface rather than ad hoc prompting.

• Scalability. New systems plug into the same standard, so capability compounds instead of restarting each time.

This is a domain problem, not a software problem

Here is the part that gets missed.

• The hard piece is not the plumbing. Anyone can stand up an interface.

• The hard piece is knowing what the right answer actually is, and being able to prove it.

Defining what recovery, plant margin, or liquid holdup means for a specific asset, under specific contracts, with specific meters, and then defending that number to a counterparty, is domain work. It requires people who have run these calculations, been wrong, and figured out why.

That is also what makes validation possible. If you have built the physics based and first principles models that describe what the system should be doing, you can tell when a number coming back is wrong. Without that, all you can say is that the answer sounded reasonable.

This is where the shape of the market matters. Broadly, companies looking for help today are choosing between AI platform providers, who give you access and tooling but have no basis to judge whether your recovery number is correct, and systems integrators, who will build you excellent pipes and leave the definitions to you. Both are doing exactly what they are good at. Neither is accountable for the answer.

The firms that can do both are rare, because it requires living in the domain and in the software at the same time.

At Pontem, we have been building that deterministic layer for years, for humans. Every model, calculator, and monitoring tool we have delivered to an engineering team is exactly the asset AI now needs in order to be useful inside an enterprise.

We are not learning a new business, the consumer just changed.

Same bridge, new user

Pontem was founded on December 30, 2022, one month to the day after a then unknown product called ChatGPT launched. We had not heard of it when we started.

What we did have was a conviction that the hard, technical problems that actually run an industry are best solved by domain experts working from real data.

That has not changed. Since day one, we have been in the business of bridging data and discipline, pairing the raw information that runs an industry with the expertise needed to turn it into better decisions.

For years we built that bridge for people.

Today we are building the same bridge for AI, giving these new systems controlled, repeatable access to the processes and data they need to be genuinely useful.

Same bridge. New user.

When we talk about Mercury, we are talking about elemental Mercury, chemical symbol Hg, not the late, great lead singer of Queen, Freddie Mercury.

The main claim to fame and probably what gives it one of its other names “i.e. quicksilver” is that Mercury is the only metal that is a liquid at room temperatures. This is also mirrored in its chemical symbol Hg, which stands for Hydrargyrum which is “Water-silver” in Latin.

Where to find Mercury

Mercury (Hg) occurs naturally in geologic hydrocarbons, it is thought to originate from volcanic rocks, often underlying the reservoir. There is a possible inherent relationship involving CO₂ and Hg (CO₂ can facilitate the transport of Hg in the gaseous phase).

In some instances, it may be possible for Geologists to predict with some accuracy whether a reservoir will contain appreciable concentrations of Hg. The correlations involve source rock proximity to underlying volcanic formations, reservoir CO₂ content, regional trends and various parametric associations. Geologists in locations that have a history of Hg in reservoirs can estimate the range of Hg concentrations with some confidence. Early indication of Hg in produced fluids is useful for planning well tests. Such studies have occurred in the gas-condensate fields of Oposhnya (Ukraine) and Astrakhan (Russia).

High Mercury Levels are Associated with the Ring of Fire: Deep Crustal Fault Zones, Plate Boundaries and Volcanoes

You may ask yourself “Why is it important to know how much mercury is in your oil and gas streams?”. The main reason is that it is highly toxic and understanding how it affects your health is so important.

Mercury and its impact on Health

Elemental mercury was first used thousands of years ago by the Ancient Romans, Greeks, Chinese and Egyptians and even then they knew it was a dangerous material to work with because people mining and working with it kept getting sick.

The reason for that is Hg and Hg compounds are neurotoxins and bio-accumulate in your body. Hg enters the body principally through the inhalation of its vapour; dermal absorption (Hg⁰ and organic Hg compounds) and ingestion (ionic Hg) are also possible.

Neurological Symptoms of mercury poisoning include:

• Tremors

• Personality changes

• Memory loss

• Coordination problems

In fact, the common phrase “as mad as a hatter” came about because of the use of mercury in felt cloth production, which was used to make hats in the 18^th and 19^th centuries.

Elemental mercury

An interesting nuance is that elemental mercury is much more dangerous when inhaled as vapor than when touched or swallowed. Modern toxicology established this distinction only relatively recently. Ancient observers knew mercury could make people ill but did not understand that inhaling invisible mercury vapour was often the primary route of poisoning

Mercury Toxicity

One of the most notorious cases of mercury poisoning in the 20th century occurred in Minamata, Japan in the 1950’s

What happened?

A chemical factory discharged methylmercury into Minamata Bay over many years. The mercury accumulated in fish and shellfish, which were then eaten by local residents. This caused widespread mercury poisoning.

Victims suffered severe neurological symptoms, including:

• Numbness of the hands and feet

• Loss of coordination (ataxia)

• Vision and hearing impairment

• Paralysis

• Coma and death in severe cases

The poisoning also affected unborn children, leading to severe developmental and neurological disorders known as congenital Minamata disease.

Why is it historically important?

Minamata disease became one of the world’s most famous environmental health disasters and played a major role in the development of modern environmental regulation and pollution control. It is often cited as the classic example of mercury bioaccumulation in the food chain.

A more recent 21st century example of Health and Safety issues associated with mercury production in the oil and gas industry is on the Ichthys platform operated by Inpex offshore Western Australia.

In this case workers and the environment were exposed to mercury on the Ichthys platform.

Read more about it here

Inpex slammed by Australia’s regulator for mercury leak a...

Ichthys Platform

Not only is mercury toxic to people it can also have an adverse influence on topsides processes especially if there is aluminium present. We go into more detail regarding those issues later in the post.

Types of Mercury

Unlike coal where Hg presents in pyrite or other sulphide minerals, several forms of Hg are found in crude oil and gas condensates; they include:

• Elemental mercury, Hg⁰

• Suspended mercury compounds, notably mercuric sulphide, HgS

• Organic mercury compounds, R₂Hg and RHgCl where R is CH₃, CH₃CH₂ etc

• Mercuric halides, notably mercuric chloride, HgCl₂

There is some evidence that other Hg species may be present, but those listed above are considered the main constituents, with Hg⁰ dominating the others.

The total mercury concentration (THg) reported in hydrocarbons varies considerably (THg concentrations in crude oil above 5 parts per million are known). Gas condensates in SE Asia have dissolved THg concentrations in the range of 10 to 800 parts per billion.

Elemental Mercury

The volatile nature of elemental mercury (Hg⁰) makes it the dominant form in natural gas, allowing it to concentrate up in the lighter fractions of refined petroleum. At ambient temperature Hg⁰ is soluble in aliphatic liquids from 1 to 3 parts per million, this is several times greater than its solubility in water (<0.05 parts per million) and TEG (<0.1 parts per million) and greater than its solubility in the amines used for gas sweetening. The solubility of Hg⁰ in produced hydrocarbons is highly dependent on temperature, thus when saturated liquids cool, Hg⁰ can precipitate and then readily absorb onto metallic surfaces and silicates suspended in liquids and reacts with corrosion products (iron oxide) in pipes and equipment thus causing it to accumulate in oil and gas production systems.

Suspended mercury compounds

Unprocessed hydrocarbon liquids and many process flows contain suspended mercury compounds, notably mercuric sulphide (HgS). (Meta-) HgS particles are usually small (1-10µm ) and are insoluble in liquid hydrocarbons and water.

Organic mercury compounds

Organic mercury compounds such as Dialkyl mercury compounds (R₂Hg) are highly soluble in crude oil and gas condensates and partition to the hydrocarbon liquid phase according to their boiling point. The prevalence of R₂Hg is a matter of some debate, but they are considered to be a very low percentage of the THg content of crude oil and gas condensates.

Methyl mercury halides such as methyl mercury chloride (CH₃HgCl) partition preferentially to the water phase when the water is separated from the oil.

Mercuric halides

Mercuric halides, notably mercuric chloride (HgCl₂), are approximately 10 times more soluble in crude oil and gas condensates than Hg⁰. However, HgCl₂ will still partition preferentially to the water phase in primary separations. The extent to which HgCl₂ is soluble in the aqueous phase depends on the salinity and pH of the produced water. The natural abundance of ionic compounds is unclear, as Hg⁰ oxidizes in liquid hydrocarbon.

Hg⁰ + HgCl₂ ——→ Hg₂Cl₂

Understanding better the types of mercury that can occur and where it may be found is very important due to the toxicity of mercury to humans. Too much exposure to mercury can have a serious effect on people’s health.

Impact of Mercury on oil and gas processes

Corrosion of Aluminium by Mercury

We should also mention that as well as its toxicity, elemental mercury has another unwanted property, which is that it is very corrosive to aluminium.

This nasty property of mercury led to an infamous failure of the LNG process at the Skikda LNG plant in Algeria in 1975.

The LNG systems failed due to a Brazed Aluminium Heat eXchanger (BAHX) in the liquefaction plant suffering a catastrophic failure due to the presence of mercury in the natural gas feed. The mercury attacked the aluminium exchanger through corrosion mechanisms such as amalgamation and liquid metal embrittlement (LME).

Failure mechanism

Natural gas can contain trace quantities of elemental mercury. In cryogenic LNG facilities, aluminium heat exchangers are widely used because of their excellent thermal conductivity.

Mercury can:

1. Penetrate the protective aluminum oxide layer.

2. Form an aluminium-mercury amalgam.

3. Cause rapid loss of metal integrity.

4. Promote liquid metal embrittlement, resulting in sudden brittle cracking under relatively low stress.

The resulting exchanger failure was severe enough that it became a landmark case in LNG process safety.

Industry impact

The incident fundamentally changed LNG plant design. Following the failure, LNG operators began installing Mercury Removal Units (MRUs) upstream of liquefaction trains to reduce mercury concentrations before gas entered any aluminium cryogenic equipment.

Typical LNG specifications today limit mercury to extremely low concentrations, often around 0.01 μg/Nm³

When it comes to oil and gas processing, the most severe effect of elemental Hg (Hg⁰) on gas processing is corrosion; both amalgam corrosion and liquid metal embrittlement (LME) can occur. In the presence of moisture, Al based alloys are particularly prone to amalgam corrosion.

Hg⁰ + Al ——→ Hg(Al)

2Hg(Al) + 6H₂O ———>2Al(OH)₃ + 3H₂ + 2Hg⁰

The reaction regenerates Hg⁰, enabling it to form more amalgam in a continuous corrosion process.

LME is caused by Hg⁰ diffusing through the grain boundaries of metal to form a liquid amalgam. In contrast to amalgam corrosion, LME does not require moisture; it can result in a rapid brittle fracture. This type of corrosion can affect a range of metals - Al alloys, Cu based alloys and some steels.

There are many recorded cases of cryogenic equipment failure in older gas plants such as Skikda, however the introduction of Hg⁰ removal units and ‘cold box’ designs that have been specifically designed so that they are resistant to Hg⁰, have served to reduce the rate of failure.

Other detrimental impacts of Hg⁰ on gas processing include:

• The deposition of Hg⁰ in equipment poses a health and safety risk to personnel involved in maintenance or inspection.

• Spent Hg sorbent materials are a hazardous waste that operators must store or process for disposal.

• Contamination of treatment processes such as glycol dehydration systems, molecular sieve and amine acid gas removal units

• Sludge containing Hg from water treatment systems, separators, de-salters and heat exchangers represent a toxic waste that is difficult to store or process for disposal.

• Hg⁰ in gas plant products can affect downstream manufacturing. Chemical plants, notably those that produce ethylene, olefins, aromatics and MTBE (methyl-tert-butyl ether) are at risk of Hg⁰ in feeds due to its corrosion and catalyst deactivation potential.

The information provided in our post to this point quite clearly details why knowing how much mercury is in your oil and gas stream is important. Consequently knowing what to look for is really important when trying to measure how much mercury is in the oil, gas and water samples that are taken from wells during exploration, appraisal and development drilling.

Sampling and Analysis

Establishing the concentration of Hg⁰ in the reservoir is key when designing gas plants. However, the ‘actual’ concentration of Hg⁰ in recently discovered reservoirs is often elusive. Hg⁰ measurements determined at the surface during flow tests or using bottomhole sampling devices are mostly lower than those achieved during production as the metallic surfaces in contact with the reservoir fluid scavenge the Hg⁰.

The potential for Hg⁰ absorption from samples collected at the surface is particularly great. Fluids must traverse thousands of meters of tubing plus topside piping and equipment prior to sampling. Furthermore, the volume of gas needed to saturate the string with Hg⁰ is usually much greater than that delivered during the period of a drill stem test (DST). Consequently, samples retrieved downhole in suitably coated sample bottles to prevent the Hg⁰ from absorbing on the inner surface of the sample bottle are the recommended method to sample for Hg.

To remove possible mercury contamination of liquid samples from mercury containing particulates (including reservoir sand) and drilling mud it is recommended that the sample be filtered prior to analysis. However, any filtration equipment must be made from suitable materials that don’t absorb Hg⁰ or be suitably coated to prevent solubilized Hg⁰ from absorbing onto the metal surface of the filter

If sampling procedures are carefully considered and estimates of Hg⁰ losses to metal surfaces are applied, determining the lower limit of Hg⁰ in produced hydrocarbons is achievable. Establishing the upper Hg⁰ limits requires regional correlations or rigorous downhole sampling procedures, including the use of special alloys and/or suitably lined sample bottles.

Ultimately there is a skill and a lot of experience that is needed when taking samples for mercury analysis. It is not a straightforward process and needs a separate detailed sampling and analysis procedure to maximize the accuracy of the process. Not all companies provide the same mercury sampling service, so it is important to factor this into whom you choose to perform the sampling and analysis. This is where Pontem Analytics can provide valuable assistance as we are able to assist in the design of the sampling and analysis program and vet any prospective service provider chosen to do the mercury sampling and analysis.

Mercury analysis in Natural Gas

You have to approach the analysis of natural gas and liquid hydrocarbons separately. Natural gas contains principally Hg⁰ and possibly Hg(CH₃)₂, though in minuscule amounts. Prior to analysis, Hg must be separated from the matrix as the presence of hydrocarbons, particularly aromatics, interfere in the determination of Hg by atomic absorption spectrometry (AAS) and atomic fluorescence spectrometry (AFS).

A summary of the ISO methods for sampling and analysis is presented in the next Table.

Mercury analysis in Liquid Hydrocarbon

In contrast to natural gas which contains mainly Hg⁰, a variety of Hg species can be encountered in crude oil and gas condensates.

According to ASTM 7482-08 “Standard Practice for Sampling, Storage and Handling of Hydrocarbons for Mercury Analysis” normal hydrocarbon liquid sampling procedures are employed to obtain discrete, homogeneous samples. A grab sample should suffice if particulates such as HgS are not expected; otherwise iso-kinetic sampling is favoured.

The determination of THg in liquids is highly sensitive to seemingly minor aspects of sampling such as, sample container type, capture of volatile Hg and the amount of headspace present in the sample. ASTM 7482-08 recommends the use of pre-cleaned (with nitric acid) 40mL borosilicate glass vials with a PTFE lined septum cap, commonly known as VOA (Volatile Organic Analysis) vials. If larger volumes of samples are required (>1L) epoxy-lined steel containers are preferred (uncoated metal containers are unsuitable and will significantly deplete the Hg content in the sample). Furthermore, sampling equipment should not contain the following – Cu, Zn, Sn, Al or alloys composed of these metals.

To avoid the loss of volatile Hg species, sampling should occur with minimal delay (dissolved Hg evaporates with an apparent vapour pressure similar to butane or pentane). To limit the partitioning of volatile Hg to the vapour phase, as little a headspace as possible should remain in the vial (1 to 2mL in a 40mL vial)

Mercury analysis in Produced Water

Produced water from petroleum production facilities contain mostly Hg in ionic or suspended forms (Hg⁰ has a low solubility in water).

The choice of sample container is similarly important for the determination of THg in water samples. ISO 5667-3:2003(E) “Water quality – sampling – Part 3: Guidance on the preservation and handling of water samples” recommends the use of pre-cleaned (with HNO₃) 500mL borosilicate glass bottles. Preservation of the sample is required, and the preferred technique is the addition of HNO₃ to pH 1 to 2 and K₂Cr₂O₇ at 0.05 m/m% (in the final solution).

Predicting Mercury Distribution

Once you have an idea of how much Mercury may be present in the reservoir fluids it is important to understand how the mercury distributes itself across the various phases at different temperatures and pressures. To achieve this Pontem use software packages that can predict the concentration of Hg in gas and liquid hydrocarbon process streams. Generally, only Hg⁰ is considered in the models; for natural gas this has proven entirely suitable (less so for crude oil). The models use thermodynamic properties and empirical data to forecast the distribution of Hg⁰ between the natural gas, liquid hydrocarbon and water phases. When there are 3 phases present, the partitioning of Hg⁰ to water is usually ignored due its low solubility in water and the accumulation of particulate Hg in separated water streams. Wherever possible it is good to compare the simulated results with measured data from similar processing plants.

Mercury Removal

Finally, once you know there is mercury in your production stream it is important to be able to remove it. Over the years a number of different mercury removal systems have been developed.

The removal of Hg from natural gas and liquid hydrocarbons is achieved using sorbents. Sorbents consist of a substrate support and a reactive component that is bound to the support. The sorbent reacts with Hg⁰ (or a compound of Hg) to form an insoluble and inert species such as HgS, HgI or amalgam. Several commercial processes are offered for the removal of Hg, some are suited to gas, others are applicable to liquid streams.

There are a number of different sorbent types that can be used to remove mercury. Some of these have been listed below.

Using sorbents to remove Hg from liquid hydrocarbons is more difficult; a variety of Hg species is usually present in the matrix, and most Hg removal systems are chemically specific to a particular form of Hg, usually elemental. If the feed contains significant amounts dialkyl or ionic Hg the removal system may not function efficiently.

An emergent and important development in natural gas production is the relocation of Hg removal units (MRU) to upstream locations. The development of high-capacity sorbents resistant to entrained water and liquid hydrocarbons (such as metal sulphides) have resulted in the positioning of MRU upstream of dehydration and acid gas removal. Furthermore, light liquid condensates that have undergone primary water separation can also be successfully treated for Hg removal.

To get the most out of any mercury removal technology, Pontem Analytics using our expertise in system optimization can create a state-of-the-art physics based machine learning model to monitor closely how the mercury removal unit is performing and how mercury levels are changing over time.

To conclude, the management of mercury in a producing oil and gas system is a complex process starting at the point you take a sample for analysis all the way through to its removal but in many instances because it is so difficult to take a representative sample and easy for the mercury to be lost from any sample taken. Many systems only find out once they start producing that they have a mercury problem and by that time it is usually too late to develop a cost-effective solution. Brownfield Engineering modifications are expensive $$$.

Taking that into account, it is important to fully understand all the issues associated with developing a mercury management strategy and that knowledge is not always present in every oil and gas operator. So with that in mind, come and get that knowledge by speaking with Pontem Analytics who have years of experience dealing with this exact problem.

Subscribe now

Background

When most people hear “AI” these days, they picture a chatbot spitting out text or a model conjuring an image from a single sentence. But the data that actually runs the world (i.e. the stuff sitting in spreadsheets, databases, and historian tags) isn’t articles or pictures. It’s tables. Rows and columns. Of all the data types out there, tabular data is far and away the most common, and it’s the bread and butter of most companies’ day-to-day work. And that’s no accident: to make data usable, we don’t just collect it, we organize it into a table, with each row a single record (a car, a well, a day of production) and each column a variable we measured (horsepower, pressure, rate). That simple grid is the standard way this kind of data gets stored and accessed, which is exactly why we call it tabular.

On June 30th, Google released a new foundational model called TabFM, built specifically for tabular data. Foundation models have completely reshaped how we work with text and images, but tables have mostly been left to the “traditional” machine learning toolkit. Naturally, the data scientist in me wanted to test drive it, first on a clean textbook dataset, and then on a real oil well’s data.

Foundation Models: What Are They, Really?

Before getting to results, we should talk about what foundation models are.

A foundation model is a large model pre-trained on an enormous, diverse pile of data so that it develops a general “understanding” of a domain, one that can then be pointed at brand-new problems with little or no additional training. The large language models (LLMs) behind tools like ChatGPT and Claude are the most famous examples: they’re trained on massive text corpora and can then answer questions, write code, or summarize a document without being retrained for each new task. We’ve all used these tools by now.

TabFM is that same idea aimed at a different target. Instead of learning the patterns of language, it’s pre-trained on a huge variety of tables, so it can walk up to a dataset it has never seen and start making predictions, often skipping the usual cycle of picking a model and tuning it.

The “traditional” way to model tabular data is to build a brand-new model for each dataset from scratch: pick an algorithm, feed it your rows, and tune it until it performs. A foundation model flips that around. It has already studied a huge, diverse library of tables, so it shows up already knowing what tabular data tends to look like, you just hand it your dataset and ask for predictions. It’s the same shift we saw with text: you don’t train a new language model every time you want to summarize a document, you just ask ChatGPT or Claude.

And while Google’s release made the headlines, it’s worth saying they weren’t first to this party. Researchers had already shown the concept worked; the crowd behind TabPFN demonstrated a couple of years back that a single pre-trained model could handle tabular prediction shockingly well. What makes Google’s entry notable is the scale it’s operating at and, frankly, the weight of the name behind it. When Google plants a flag on “the most common data type in the world,” it’s worth paying attention.

Test Drive #1: A Simple Cars Dataset

To get a feel for how TabFM behaves, I started where just about every data scientist starts: a clean, well-behaved dataset. In this case, the classic “cars” dataset, predicting a car’s fuel efficiency (MPG) from familiar characteristics like horsepower, weight, and displacement. A quick look at the data shows the relationship you’d expect: more horsepower and more cylinders, lower MPG, but there are also a few other variables in the dataset not shown below, including model year, origin (which is categorical not numerical), displacement, weight, and acceleration.

Getting set up was much easier than I expected. You load TabFM, pass in the dataset, and then ask for predictions. No training loop, no grid search, no tuning. That’s a big part of the appeal.

But before grading anything, it’s worth a quick word on how you grade a model fairly. If you let a model both learn from and get scored on the exact same data, it can essentially memorize the answers and look far better than it truly is. To guard against that, I used an 80/20 train/test split on every analysis in this post: the model learns from 80% of the rows and is then scored on the remaining 20% it has never seen. So every metric I report and every parity plot you’ll see comes from that held-out 20% test set, which is what lets us trust that the model is genuinely performing well rather than just repeating what it already saw.

To grade the results, I used RMSE (root mean squared error). If you’re not a stats person, RMSE is just a tidy way of summarizing how far off the predictions are from the actual values. Lower is better, and it’s reported in the same units as whatever you’re predicting (MPG in this case) so interpretation is easy.

I also pulled up a parity plot, which is one of my favorite quick gut-checks for a regression model. If you haven’t seen one before, the idea is simple: put the actual values on one axis and the model’s predicted values on the other, and plot one point per prediction. If the model were perfect, every point would land exactly on the 45-degree diagonal. The tighter the cloud of points hugs that diagonal, the better the model is doing. Here the points hug the line closely, with only mild scatter, which is exactly what you want to see.

Test Drive #2: A Producing Oil Well

Textbook datasets are nice, but they’re not what we actually deal with. So for a tougher test, I pointed TabFM at production data from a producing well.

The variables in this dataset, which span January 2023 until roughly January 2025 (~2 years), are:

• BHP (bottomhole pressure): the pressure measured down at the bottom of the well.

• THP (tubing head pressure): the pressure up at the surface where the fluids come out.

• GOR (gas-oil ratio): how much gas is produced alongside each barrel of oil.

• Water cut (WC): the fraction of the produced liquid stream that’s water (vs oil)..

Using measurements like these as inputs, the goal was to predict the well’s liquid production rate (Liquid_Rate).

However, this well is a little messy, which you can see in the data above. Its operating conditions have changed frequently over its life, so the model isn’t looking at one steady regime, it’s looking at several. And one of the messiest variables is the water cut, which sat flat near zero for two full years and then jumped abruptly into the 20 to 40 range. Sudden regime shifts like that are exactly the kind of thing that makes models struggle. In other words, this is a proper test of TabFM, not an easy example.

To make it a fair fight, I ran TabFM head-to-head against an out-of-the-box gradient boosted model (think XGBoost or LightGBM), the workhorse most of us would reach for by default on a problem like this because they have historically been the best models for tabular data, even out performing neural networks.

What surprised me most is that TabFM took the win without any tuning at all. Both models track the well reasonably well across most of the range, but TabFM stays tighter to the diagonal while the gradient booster scatters more, especially at the higher rates. The messy, abruptly changing behavior that I expected to trip everything up is exactly where TabFM held its edge.

For the detail-oriented, here’s each model on its own:

Wrapping Up

A few things stood out to me from this work:

• Foundation models for tabular data have clearly crossed the line from “academic curiosity” to “something I’d genuinely consider reaching for.”

• The out-of-the-box performance, with no model selection and no hyperparameter tuning, is the part that should make everyone in this space sit up.

• Messy, real-world data (like this well) is still the true test, and it’s where these models will earn or lose their keep.

This is absolutely something to watch going forward. If a general-purpose model can match or beat a carefully chosen one on the first try, the way we approach a huge chunk of everyday modeling work starts to look pretty different.

And on a more personal note, I’ll admit there’s a bittersweet element to all this. For a lot of us, hunting for the best model and painstakingly dialing in the right hyperparameters is one of the genuinely fun parts of being a data scientist. It’s the “craft”. Watching a foundation model come in and do it in one shot is both exciting, but also a little deflating at the same time.

A Note on Licensing

One final thing before I wrap up. This was a review and evaluation exercise only, not commercial use. TabFM is released under Google’s TabFM Non-Commercial License v1.0, which is expressly not intended for commercial use. Everything I did here falls squarely under what the license calls a “Non-Commercial Purpose”: testing, evaluation, and research that isn’t tied to commercial gain, production deployment, or revenue generation. None of these results were used in commercial decision-making, client deliverables, or paid products or services.

Flow assurance was my entry point, but what kept me in this field was the deeper story behind every transient event, every surge, every cooldown curve, and every restart sequence. Over the past 12 years, I’ve worked on projects across Europe, Asia, Africa, and the Middle East, supporting developments for ADNOC, Aramco, Shell, BP, Tullow, Dragon Oil, Cadent and many others.

Alongside this regional breadth, my career has taken me through three very different engineering hubs: London, Kuala Lumpur, and Abu Dhabi. Each shaped me in its own way: London sharpened my technical discipline, KL immersed me in fast‑paced delivery and diverse operating environments, and Abu Dhabi grounded me in large‑scale, high‑stakes energy systems.

Along the way, I’ve had the privilege of working with brilliant engineers, specialists and mentors who pushed me, challenged me, and helped shape the way I work. Some of them are now at Pontem, and I’m genuinely excited to be working alongside them again.

I’ve spent days modelling wax deposition in OLGA, nights tuning fluids in Multiflash, and months unravelling the operational stories hidden inside shutdowns, surges, cooldowns, and restarts.

I’ve worked on hydrogen networks, non‑Newtonian sulphur pipelines, MEG systems, deepwater tiebacks, heated pipelines, and long‑distance export routes with each project adding another layer to how I understand flow assurance, thermodynamics, and the delicate balance between safety, operability, and performance.

This understanding was deepened further by stepping into the world of process dynamic modelling, where facilities reveal how they behave under stress, during transitions, and in the moments that never show up in steady‑state studies. I’ve built and tuned dynamic models in HYSYS, UniSim, and Virtuoso, analysing separator behaviour, tank vapour systems, controller responses, dehydration trains, and transient facility interactions to understand how systems respond to real‑world operating scenarios.

These models taught me how facilities respond to changing conditions, how control systems shape behaviour, and how transient interactions can expose risks long before they appear in the field. They also reinforced a lesson I carry into every project: data isn’t just an input – it’s a teacher.

That mindset naturally led me toward AI and advanced analytics, culminating in completing Microsoft’s Autonomous AI Architect Training – a step that expanded how I think about modelling, prediction, and decision‑making.

Today, I’m excited to bring that blend of engineering depth, dynamic modelling experience, and data‑driven curiosity to Pontem Analytics.

Pontem is a place where domain expertise meets analytical creativity, where complex systems are understood through both physics and patterns, and where solving hard problems is part of the culture. Joining this team feels like a continuation of the work I love: exploring complexity, learning from data, and building models (physical or digital) that help clients operate more safely, efficiently, and confidently.

I’m looking forward to contributing my experience, learning from exceptional colleagues, and helping deliver solutions that make a meaningful impact around the world.

Here’s to new challenges, new ideas, and the next chapter.

A valve closes somewhere on a large onshore pipeline network. Pressure spikes, flow reroutes and equipment is at risk. As an engineer you need to know what is happening in real-time, not after running a manual simulation on a desktop tool.

That is the problem we set out to solve. We built a tool that could be embedded, automated and deployed at scale. One that simulates the pressure surge and flow rerouting caused by a valve closure, and delivers those answers as it happens.

But to understand why that matters, we need to start with the physics.

What is Liquid Surge?

You may know this phenomenon as water hammer however, in the pipeline industry, we refer to it as liquid surge.

You've probably heard it before, it's the thump! you hear when you shut off a tap too quickly. The moving water suddenly has nowhere to go, so it pushes back. That pushback is a pressure wave, and in a large pipeline network, it can be powerful enough to damage pipes, fittings, and equipment.

How Do We Simulate It?

The basic way to observe it is simple. A pipe, a valve, a stopwatch and a pressure sensor. Let the water flow, then shut the valve in half a second. The sensor spikes and that's liquid surge.

For complex pipeline networks however, you can't just run physical tests every time you want to understand a closure event. That's where simulation tools come in. They allow engineers to model these transient events without being on-site.

A Note on the Math

Most commercial tools use Method of Characteristics (MOC) , and it’s the industry-standard mathematical technique for simulating transient flow (sudden changes in pressure and flow) in liquid pipelines. This approach discretises the pipe into spatial segments and solves for pressure and flow at each segment, at every timestep.

Liquid flow in a pipe is governed by two equations describing how pressure and flow change over distance and time. Solving these directly across a whole network is computationally expensive.

MOC sidesteps this by transforming these complex equations into simpler ones that can be solved step by step along the pipe. The trick is that these simpler equations only hold along specific paths called characteristic lines, which correspond exactly to the speed at which pressure waves travel through the liquid. By marching along these paths at each timestep, MOC efficiently tracks how a pressure wave propagates and reflects through an entire network.

Commercial simulation tools for liquid surge are mature and reliable. They have been used by pipeline engineers for decades and they do the job well.

So why build our own?

The answer is not accuracy. It is deployability.

Tools like these are designed to be run manually by an engineer sitting at a desktop. You build the model, you run the simulation, you read the results. That works perfectly for one off analysis. It does not work when you need a system that can predict pressure surge events automatically, in real time, embedded inside a live pipeline management platform.

That is what we needed. And that is what we built.

Our Pontem Model implements the core hydraulic equations using MOC to simulate pressure and flow efficiently across large networks. Critically, it models the entire network simultaneously. When a valve closes, we can track not only the pressure surge at that point but also how flow redistributes across the rest of the system in real time.

This is the tool we put to the test.

Putting It to the Test

To validate the Pontem Model against a commercial simulation tool, we started with the simplest possible configuration a single flat pipeline with pressure boundaries at both the inlet and outlet set to 300 psig, achieving a steady-state flow of approximately 1,000 barrels per day.

A control valve was placed just upstream of the outlet. Once steady-state flow was established, the valve was closed and pressures were recorded at two points the pipe inlet and just upstream of the valve. The same network was built independently in both tools and run under identical conditions.

Seven experiments were run in total, each isolating a different variable:

• Baseline - flat pipe, 0.5s valve closure

• Elevation - pipe inclined 500ft upward and downward

• Closure time - 10s and 30s closures

• Flow rate - increased to 3,000 bbl/d

• Outlet pressure - reduced to 150 psig

  

The Results

We started simple. A single flat pipe, a fast valve closure, identical conditions in both tools. The results came back almost identical.

Steady-state flow matched to within 0.2%. Peak surge pressure was essentially the same. For a first test, that was encouraging. But a flat pipe with a fast closure is about as easy as it gets. So we pushed harder.

We then introduced elevation. First a 500 ft descent, then a 500 ft climb. We increased the flow rate to 3,000 bbl/day. We dropped the outlet pressure to 150 psig. In every case, the two models tracked each other closely.

Then we slowed the closure down.

A 0.5 second closure is violent. The water has no time to adjust and the pressure spike is immediate. A 30 second closure should be gentle by comparison.

That assumption did not fully hold up.

The surge rise difference jumped to 12.7% for the 30 second case. At first glance that looks like a problem. But the absolute difference was only 2.5 psia. As a percentage of actual peak pressure the difference was 0.7%, consistent with every other experiment. The number looks large because surge rise is a small quantity to begin with.

The real explanation is in how each tool handles valve behaviour. The Commercial Model closes linearly over time. The Pontem Model uses a butterfly style curve, staying nearly fully open through most of its travel before choking sharply at the end. For fast closures this distinction does not matter. For slow closures it does. Not a flaw, just a known difference between the two approaches.

One final observation. The Pontem Model’s pressure peak consistently arrives slightly later than the Commercial Model across all experiments. For fast closures the offset is around 1.1 seconds. For slower closures it is 2.74 seconds. The magnitude is right. The timing is off. That is worth investigating further.

Scaling Up

Phase 1A proved the model works on a single pipe. The real test was always going to be a network at scale, and that is exactly where the Pontem Model is being deployed today, on a large onshore pipeline network.

At this scale, the model provides real-time insight into system pressures and flow routing during normal operation. That alone is not a small task for a network of this size.

Where it becomes genuinely useful is during valve closure events. The model indicates the maximum pressures expected across the network, and critically, how flow reroutes once that valve closes, including whether the rest of the network can handle the extra flow being pushed its way.

The model runs directly through the client's existing analytics platform. Engineers do not need a separate tool or a separate login. The insight is exactly where they are already working.

What’s Next…Future Work

Phase 1A and the work on the large onshore pipeline, give us confidence that the Pontem Model holds up, both in controlled conditions and at scale in the field. But there is more to validate.

Phase 1B will test a more complex network geometry, with two outlets instead of one, to confirm the model handles flow routing correctly when there is more than one path for liquid to take. From there, the goal is to keep expanding into more complex topologies, including converging and diverging pipelines, while continuing to refine how the model performs against real field data.

But the bigger picture goes beyond liquid surge. This work is part of a wider Pontem capability to build and deploy physics based models that run online, embedded directly into the platforms operators already use. Not models you run manually and interpret offline, but tools that sit inside your workflow and give you answers as conditions change.

We did not set out to replace commercial software. We set out to solve a problem it was not built for. A validated model, running on a live pipeline network, giving operators the answers they need when it matters most. And that is the kind of work we deliver.

Share

That title captured the central challenge: what should a company do once it has built the foundational layer of contextualized industrial data?

It was a great few days of networking, learning, and spending time with folks in the Cognite ecosystem of customers, partners, and providers.

Throughout the day, there was a consistent message about the freedom companies have; to create the tools they need to run their business once they establish a solid data foundation.

Since the NBA Finals were going on last week, we could not help but make the analogy that the speakers before us were like Kobe throwing up the perfect alley-oop, and we came in like 2000 Shaq coming in to take it home with real, practical examples of how we used that same philosophy to run our business.

Bad analogies aside, the setup was real.

• Paul Miller from Forrester laid out a strong value statement around Cognite: get the basics right, involve the frontline to identify real pain points, define the conditions of satisfaction with stakeholders, and find partners with strong technology, deep domain knowledge, and a shared outlook. Never lose sight of the human element!

• Our colleague Adam Ballard joined a panel with ExxonMobil and International Paper. Adam talked about how he used Cognite to accelerate the same AFoF — Autonomous Fields of the Future — program in an onshore setting, optimizing daily routes and helping operators in North Dakota spend more time on meaningful work.

• Right before our session, Chirayu Shah (Cognite Chief Product Officer) and Lauritz Lont (Cognite VP of AI) laid out a vision that echoed the theory behind our presentation. We then got to show what that vision can look like in practice. They showed the perfect visualization of what the technology stack looks like for an industrial operation — one we had been trying to demonstrate to our leaders for the last few years, with much poorer graphics.

By the time we got up, the setup was almost too perfect. The presentations before us had all pointed to the same basic idea: this was not about buying a platform, building another dashboard, or chasing digital transformation as a buzzword. It was about creating the conditions for a different way of working.

Our presentation focused on what happens when that vision hits the real world of operating an asset: the people, the data, the workflows, the legacy systems, the good pilots, the stalled pilots, and the constant pressure to prove that digital work actually moves the business.

Jacob talked about how we initially ran disconnected pilots that were not always tied clearly enough to business goals. A technology pilot can be technically sound and still struggle if the people closest to the work do not understand the “why.” He shared the story of how we almost had the Department of Homeland Security called on us because the timing and communication were off for a fixed-wing emissions capture pilot. We also discussed how much work it took to recover credibility after that and after the robotics fiasco with Spot, the Boston Dynamics robot dog.

Depending on how it is communicated, offshore robotics can be seen as a useful tool, a distraction, or even a threat. Business transformation is not only a technical exercise. It is also a trust exercise across all stakeholders.

The important point about the data foundation is that it was never just about storing more data. We already had plenty of data. Today, oil and gas companies probably use only a fraction of the data they generate. Useful information was spread across historians, SAP, documents, inspection systems, spreadsheets, engineering tools, field notes, and in people’s heads.

Pilots could prove a concept, but they could not scale into daily operations if the data needed to support them was trapped in disconnected systems. Equipment needed context that could connect time series, P&IDs, documents, SAP events, maintenance history, integrity anomalies, inspections, operator observations, and workflows.

That is when industrial data starts to become useful — not because it sits in one place, but because it begins to reflect how the asset actually works.

An offshore asset has only so many ways to create value.

For an aging offshore asset, the business case was pretty simple: produce more, spend less, spend smarter, avoid failures, and find production opportunities that were already there but not obvious.

We were looking at sensors, robots, computer vision, vibration data, drones, and AI. Not because the technology was interesting on its own, but because any one of those could produce an initial signal to help us protect uptime, reduce OPEX, make better decisions, and find “zero-cost barrels” hidden behind constraints, suboptimal configurations, or missed signals.

These signals could be the initial step into an operating philosophy that we call Find-to-Fix.

Find-to-Fix is simple in concept: detect an issue or opportunity, validate it, diagnose it, prioritize it, execute the corrective work, and learn from the outcome. The hard part is making that happen across real people, real systems, real assets, and real workflows.

We tested technologies that could help the asset sense and perceive more of what was happening.

• Remote SME support with Microsoft HoloLens

• Computer vision for fabric maintenance and structural integrity detection with Abyss

• Automated wall thickness monitoring with Inductosense

• Predictive vibration monitoring with RangerPro

• Drones and robotics to reduce rope access hours, taking people out of harm’s way during dangerous inspections

Not every pilot was perfect, but perfection was not the point. The more important question became: can this data connect to the foundation and become part of a connected operating system?

Once the foundation was in place, Cognite native applications helped us test that theory.

• Infield digitized operator rounds and captured more than 100,000 field entries in the first year.

• Remote helped visualize integrity anomalies in context.

• InRobot supported robotics trials with 3D mapping to collect critical data.

• Industrial Canvas helped our organization standardize root cause analysis. In fact, we created a business rule that all RCAs must be completed with the tool.

• AtlasAI queried our workorders to help find similar patterns in detect notifications and workorders

• We also started exploring Maintain to better support work planning during shutdowns and turnarounds.

Those applications proved real value, but they also exposed the next bottleneck.

It was easier to use the native applications because they were “cyber-blessed,” but the custom applications that could take us to the next level of action in Find-to-Fix were still hard to progress. They were slow, expensive, and forced to move through IT, cyber, supply chain, logistics, communications, and all the normal friction inside a large company.

We routinely challenged the system.

We became the “innovation cowboys,” fighting existing corporate institutions that could not keep up with the pace of digital innovation.

We introduced multiple custom to semi-custom solutions:

· a physics-informed neural network surrogate model for gas lift optimization which delivered production uplift.

· a visualization / engineering workbench that put reliability and integrity in plain view with SLB Optisite

· an executive asset scoreboard which kept leaders abreast of the financial and safety metrics of the asset with IQumulus

The data was there and the value was real, but the path to absorb, scale, and tell that story was harder than it should have been.

We were not being reckless. We were bringing awareness to the tension that exists when trying to move at the speed of bettering the asset versus systems designed to move at the speed of policy. Every company says it wants innovation, but the corporate immune system around innovation often struggles when something does not fit the normal path.

We later recognized the importance of the human side of change management and began incorporating these principles after attending Prosci’s change management program, which helped us communicate ideas more effectively and implement initiatives more smoothly.

We even learned there was a better word for “haters” — it was something like “resistors.” Hahaha

Most resistant groups inside an organization

Back to Find-to-Fix…

Find-to-Fix might start with an operator capturing an abnormal condition in the field or a sensor detecting a change in vibration. In the old way of working, that signal could die in a note, sit in a spreadsheet, get buried in email, or wait for the right person to connect the dots, never making it to fix.

In a Find-to-Fix workflow, that signal is detected and validated against adjacent contextualized data.

Cognite Flows showed us a path to create an application that fit our operating model.

It showed a path to build the kind of custom industrial applications we had been trying to create for years, but at a very different speed — and with a lot fewer people calling us cowboys and more people recognizing us as innovators.

Richard’s demo brought the story full circle: a Find-to-Fix application built on top of the data foundation, connected to workflows, context, and decision logic.

It does not mean every problem is solved. Trust, process, governance, and engineering judgment are still required. But when the foundation exists and the application layer can move faster, the conversation changes.

Ideally, you can now go to Gemba (for us it was offshore), have a human-centered discussion about how data is used, what additional context is needed, and what would make decisions easier for the OIM, operator, or mechanic — then turn around a minimally viable product by the next business day.

Cognite customers can move quickly, test ideas, and turn operational insight into working tools without waiting for the traditional machinery of enterprise software development to catch up.

For field-minded problem solvers, it is hard not to see the appeal.

That is also why the combined strength of our companies Pontem, Iron Stag Advisory, and Genesis Lone Star is worth paying attention to. We all have teams behind us that understand how to work as one when the problem demands it. Together, we can help companies move beyond digital ambition and start solving real operational problems with practical tools, strong engineering context, and people who understand the field.

Flows is a cowboy’s dream.

Find-to-Fix is what becomes possible after the foundation is built.

Link to our full presentation:

At Pontem, we spend a lot of time thinking about the challenges our customers face across the full lifecycle of an offshore subsea asset.

We often see how early design decisions are made with limited fluid data, uncertain production profiles, and unknown operational philosophy.

Customers often have to debate whether a flowline should be large enough to capture upside, while not being so large that the system becomes more vulnerable to slugging and liquid management issues.

Early on, we are also typically determining what chemicals we may need and how often we may need to intervene without a full operational picture. Similar questions come up around flowline routing or single versus dual flowline when drilling down into the chemical inhibition effectiveness and future production behavior.

Our industry is moving toward minimum viable design concepts (more single flowlines) in order to reduce cost, move faster, and capture the NPV upside of earlier first oil. Spending too much time designing around every possible issue on the front-end can erode project value before the field ever comes online. Sometimes the right answer is not to eliminate every possible problem through design, but to make the critical layout decisions with future intervention in mind.

With a partner like Paradigm Flow Services, early conversations about intervention planning and production deferment avoidance can make future remediations more practical, adding a layer of realism into flow assurance decision-making.

Paradigm are the creators of Flexi-Coil®, the world’s first flexible coiled tubing technology designed specifically for platform-based riser and pipeline intervention. Their field-proven technology gives operators a practical mechanical solution for addressing deposits and improving flow.

Together, Pontem and Paradigm see a real opportunity to help customers connect the dots between predicting flow assurance problems and solving them – faster, safer, and more cost-effectively.

We recently sat down with the Paradigm team to learn more about Flexi-Coil® and discuss how the industry can benefit through our collaboration.

Q&A with Paradigm Flow Services

Q1: For those who may not be familiar, can you describe what Flexi-Coil® is, what problem it was designed to solve, and what makes it fundamentally different from the conventional approaches that operators have relied on for riser and pipeline intervention?

Flexi-Coil® was conceptualized to help resolve industry challenges with blockages such as wax, sand, asphaltenes or hydrates that occur at distances from the production facilities with limited options available to resolve without major subsea intervention.

Flexi-Coil® utilizes a neutrally buoyant, composite coil which can pass through >1,000 degrees of bends and to distances of up to 6 km (~3.7 miles) from the entry point. The Flexi-Coil® tubing is made from composite materials so there is no risk of abrasive wear to the inner-wall of pipeline or flexible systems during deployment or recovery.

These key differentiators mean that gaining access to the production system is less invasive and quicker so therefore production can be restored faster.

Q2: Before any mechanical intervention takes place, there’s a critical step in understanding where the restriction is and what it’s made of — whether that’s scale, asphaltenes, wax, or hydrates. From your experience, how important is the upfront modelling and diagnostic work in planning a successful Flexi-Coil® campaign, and how does having a detailed picture of the deposit profile and pipeline hydraulics change the way you execute a job?

Before any live operation we gather key system data from the client along with completing a site survey by mobilizing a Field Technician to the facility – this usually happens very quickly, within a couple of days from first notice of an issue.

If the line is fully blocked, technicians carry out a Find-Block® survey using our proprietary technology connected to a topsides instrumentation point and accurately locate the face of the blockage. If access is available from both sides of the system, we can also confirm the full length of the blockage to determine the best side to target it from.

From this point our engineers conduct a deployment simulation using our in-house software called ‘Flexi-Sim®’’ to confirm that the piping system is navigable and that the blockage is within reach.

Preparation is critical to the safety and success of any offshore remediation activity. Before work begins, we engage closely with the operator to understand the system layout, line history, suspected blockage mechanism, prior chemical injection strategy, and any operational constraints.

As a collaborative team, Paradigm, Pontem, and the customer can review the operational history of the flow assurance issue and use deposition modeling to better understand the likely cause of the blockage. This helps build a stronger root cause analysis, increases confidence in job execution, and supports more accurate planning – ultimately reducing unnecessary platform downtime and cost.

Q3: What advantages do you see to customers getting Paradigm involved early in the life cycle of an asset and how can Paradigm and Pontem work together in that aspect?

Getting involved early allows us to help operators think about intervention and future production flexibility before the asset design is locked in.

Paradigm and Pontem can create real value for customers. Pontem can help operators understand where flow assurance and operability risks may emerge across the life of the field, while Paradigm can bring practical intervention and remediation experience into the design conversation early. Together, we can help operators understand not only what risks may exist, but what options they would have if those risks materialize.

Together we can help connect flow assurance prediction, production surveillance, remediation planning, artificial lift optionality, and field-proven intervention capability into a more complete lifecycle strategy for the asset.

Early planning is also important when thinking about artificial lift optionality. Flexi-Coil® can provide a built-in option for in-riser gas lift. This is not only limited to conceptual phases as Flexi-Coil® can also be retrofitted for flow improvement (as a velocity string or gas lift retrofit) for mid to late-life subsea wells. These options help to manage subsurface and production risks by giving operators more flexibility. If the asset is designed with future access, installation, and intervention requirements in mind, operators can preserve optionality without necessarily overcomplicating the base development concept.

By bringing Pontem’s flow assurance and lifecycle modeling experience together with Paradigm’s practical intervention technology, we can help operators make better decisions earlier and respond more effectively when issues arise. This collaboration is about building more resilient subsea systems, preserving operational flexibility, and helping customers move from prediction to action with greater confidence.

For Pontem, the value of this collaboration is simple: flow assurance should not stop at prediction.

As an industry, we are good at identifying risks, modeling outcomes, and defining operating envelopes. But over the lifecycle of a field, conditions change. Production profiles evolve, fluid behavior becomes clearer, equipment ages, and risks that once lived in a model can show up in the field. When that happens, customers need practical options for how to respond.

Pontem brings the flow assurance, operational data, and lifecycle modeling perspective. Paradigm brings field-proven intervention capability and practical remediation experience. Together, we can help operators make better early design decisions, preserve future optionality, and respond more effectively when production issues arise.

This collaboration is about connecting data, discipline, and practical field execution to protect production, reduce downtime, and extend the life of offshore subsea assets.

Fun fact: I did my kindergarten, middle school, and high school all in the same village, and grew up in a joint family of thirteen. There’s a particular kind of patience you learn early in a household that big - and a particular kind of stubbornness too.

Sidhwan Khurd - where it all started.

Education

I moved to Canada for graduate school at the University of Waterloo, where I picked up a Master’s in Mechatronics and Mechanical Engineering with a focus on Data Analytics and Green Energy, alongside a Graduate Diploma in Business Intelligence and Entrepreneurship.

Fun fact: Mechatronics is the field that lets you call yourself a mechanical, electrical, and software engineer in the same sentence - depending on which part of the project is currently on fire. It also gave me an unhealthy enthusiasm for soldering irons and a genuine love for problems that sit at the intersection of physical assets and digital tools.

Career Journey So Far…

My career kicked off in 2020 at LG Electronics in India, where I spent six months as an R&D co-op debugging PCBs, analyzing field failures, and chasing lean manufacturing wins wherever I could find them.

I then took on an internship with RealBotics, a Pittsburgh-based startup, where I led design of an IoT project full mechanical design in Fusion 360, a Python and Raspberry Pi software stack, and through-hole PCB layout for production. The kind of project where one day you’re CAD-ing, the next you’re soldering, and the next you’re explaining to non-technical stakeholders why version 7 looked nothing like version 1. A useful lesson in iterating quickly.

After Waterloo I joined Wood in 2023 as a Digital Reliability Analyst, and that’s where I really found my footing. Over the next three years I worked across some genuinely fun problems - sensor feasibility programs for refinery condition monitoring, prescriptive maintenance libraries for a greenfield LNG facility, and AI-assisted PMO discovery on years of SAP maintenance data.

Fun fact: “Process Automation Engineer” sounds neatly defined until you realize it can mean anything from configuring a sensor on a refinery rooftop to writing pairwise classification logic on millions of material records. Both happened in the same year.

By 2025 I was promoted to Process Automation Engineer II, and the work scaled up - automated corrosion modelling pipelines for upstream operators, large-scale spares interchangeability and reliability programs, and selection as a presenter at OTC 2026 and API 2026 on automated corrosion modelling and RPA tooling.

Lived through the worst snowstorm of my life on assignment in St. John’s, NL - 100 cm in 48 hours. Atlantic Canada doesn’t do anything by halves.

Joining Pontem

Now I’m at Pontem Analytics as a Senior Reliability & Predictive Analytics Specialist, working on predictive analytics deployment, maintenance modelling, and reliability engineering for industrial clients.

What pulled me to Pontem is the chance to work alongside people who genuinely care about both the data side and the engineering side of these problems - and to be part of a team where the work spans research-grade analytics through production deployment and hyper care. That combination of technical depth and operational impact is exactly what I’ve been looking for.

Outside of Work

Outside the day job, I spend most of my time with my wife - whether that’s planning the next trip, working through a new recipe in the kitchen (we’ve gotten dangerously good at North Indian comfort food), or just decompressing after work with a cup of tea or a pint on the balcony.

‘Rat’- our hamster and chief decompression officer.

I’m also incurably a tinkerer. If I’m not building something at work, I’m probably building something at home. Recently that’s meant a full-stack fintech web app - built mostly because I wanted to understand the stack end-to-end - and an IoT GSM-based remote irrigation controller for a family farm back in India, designed to do the same job as commercial systems at a third the cost. More recently I’ve picked up shooting as a hobby, and somewhere on the longer-term wishlist is rebuilding cars - once I have the space and the time to do it justice. The common thread across everything I do, at work and outside it, is the same: can this be done better, simpler, or smarter than how it’s being done today?

Excited for what’s next.

Manraj

London — April, 2026 — Eigen, a leader in connectivity, workflow automation, and advanced visualisation, today announced a strategic partnership with Pontem Analytics, a specialist in production optimization, process chemistry, data-driven monitoring and advisory applications. This collaboration brings together complementary expertise to deliver cutting-edge, AI-driven solutions for production chemistry optimisation across complex industrial environments.

The partnership focuses on combining Pontem Analytics’ advanced hybrid modelling capabilities with Eigen’s process digitalisation expertise and real-time analytics platform. Together, the companies aim to provide a seamless, end-to-end solution that enhances operational efficiency, reduces chemical usage, and improves production outcomes.

As part of a new project initiative, Pontem Analytics is deploying AI-powered hybrid models to determine optimal chemical injection concentrations (ppm) based on dynamic process conditions.

“This partnership represents a powerful convergence of domain expertise and digital innovation,” said Joel Chacon, CCO at Eigen. “By integrating Pontem Analytics’ deep knowledge in process chemistry with Eigen’s capabilities in connectivity and automation, we are enabling smarter, data-driven decisions at the operational level.”

Pontem Analytics brings extensive experience in flow assurance and production chemistry, leveraging hybrid AI models that combine first-principles understanding with machine learning. Eigen complements this with a scalable platform that connects disparate data sources, automates workflows, and delivers intuitive visualisation for operators and engineers.

“This collaboration allows us to move from insight to action” said Ian Kopperman [at Pontem Analytics. “Together, we are bridging the gap between advanced analytics and real-world application. By embedding our hybrid models into Eigen’s platform, this collaborations delivers real-time optimisation that directly improves chemical efficiency and production performance.”

The partnership is expected to expand across multiple use cases, delivering measurable value to clients in energy and process industries.

For more information, visit:
Eigen - www.eigen.co
Pontem Analytics - https://www.pontemanalytics.com/

About Eigen
Eigen specialises in connectivity, workflow automation, and advanced visualisation, enabling organisations to unlock the full value of their operational data through scalable, intelligent solutions.

About Pontem Analytics
Pontem Analytics specialises in production optimization, process chemistry, data-driven monitoring and advisory applications, delivering advanced hybrid modelling solutions that optimise chemical usage and improve production performance in complex industrial systems.

I have had a good career. One built on expertise and solving hard problems. But at some point I realized that building a career around expertise is not the same thing as building one around impact, and neither is the same as fully owning it.

I did not build my career around some perfectly defined master plan. I built it by solving the problems in front of me. I said yes to difficult work. I adapted to what the business needed. I became useful. Over time, usefulness became a reputation, and that reputation became a career.

A good career, and one I am deeply grateful for.

But there is a difference between building a successful career and fully owning it.

Fresh out of Georgia Tech, I started at MSi, putting in the reps and building a strong technical foundation in flow assurance modeling. More importantly, I was exposed early to great people, tough problems, and the kind of thinking that would shape the rest of my career.

Over time, I became the flow assurance SME who could step into complex situations and quickly help make sense of them. I am proud of that identity. It taught me how to think in systems, not just in tasks, and how to stay calm when the answer is not obvious.

It also taught me something else: expertise can become a box.

Corporations value specialists. They depend on them. But they do not always know how to see them beyond the role they have already proven they can perform. The deeper your expertise becomes, the easier it is for others to mistake that expertise for the full extent of your potential. You become highly trusted, highly regarded, and sometimes highly confined.

That tension has shaped much of my career.

Even while flow assurance kept pulling me back, I had the chance to see and learn a lot. I saw two large deepwater fields move from concept through startup. I built procedures and relationships during operational readiness phases with operators who would go on to become Production Supervisors and OIMs across the Gulf. I managed subsea ROV vessel scopes to execute integrity inspections and reliability work on mature assets. I also saw firsthand how seemingly simple early project decisions could come back later as operational challenges (mostly in the form of hydrates and asphaltenes - ouch).

That variety mattered: it was my first taste of Independent Spirit, doing what was needed to help the Corporation make money. It broadened my view and deepened my instincts. But the times in between those roles often came with inconsistent responsibility and authority.

Eventually, I moved into a more strategic technology role, and it confirmed something I had suspected for a long time: I was capable of contributing well beyond the narrow boundaries of the label I had carried. That role pulled me closer to the intersection I had been circling for years — operations, data, problem solving, and innovation. I had the opportunity to directly influence which technologies we tested, deployed, and scaled across offshore facilities.

And then, as often happens in corporate life, circumstance intervened.

A merger was announced, but there was some uncertainty with timeline and outcome.

During two years of uncertainty, my team continued to push forward with innovating. In fact, some of our most meaningful progress happened in that stretch. We rolled out digital operator rounds, capturing 100,000+ operator entries in the first year alone. We installed vibration and electrical condition monitoring to detect anomalies on critical machines, deployed automated integrity sensors, and developed a subsea condition monitoring program. We tested computer vision to predict corrosion on fabric and structures, deployed a physics-informed neural network model that boosted GoA daily production and created an Asset Management Center on our production floor at Hess Tower.

A merger changes things. Business lines shift. Priorities get reshuffled. Advocates move on. Roles that look like the future one day can disappear the next. And…flow assurance came calling again. It was a reminder that much of what happens in a career is shaped not just by merit, but also by timing, structure, and forces outside your control.

That realization did not make me cynical. It made me clearer.

At some point, you have to decide whether your career is something that happens to you, or something you are willing to actively claim for yourself.

For me, this is that moment.

I have spent enough years proving that I can operate inside the box. What I want now is the opportunity to build on the full range of how I think and what I care about.

From the customer perspective, I see major gaps across industrial operations in basic problem solving, root cause analysis, and the beneficial use of lessons learned. Digital companies are trying to build the next great tool without enough domain knowledge. Engineering companies are trying to modernize operations without wanting to take on full-stack software solutions. And somewhere in the middle, the customer is still asking the most important question: how do create ROI from technology in my business?

That question matters to me.

I have a deep personal passion for production surveillance automation, and I believe the future will be built through end-user-informed analytics and high-quality visualizations grounded in the principles of reliable operations.

Identify anomalies. Diagnose the problem. Provide a prognosis. Plan and execute the corrective work.

At its core, the workflow is not complicated. But much of the surveillance work done today is still repetitive, manual, and difficult to scale. Valuable knowledge gets trapped in spreadsheets, habits, and people’s heads. Too much of it disappears when someone changes roles.

That is not sustainable. And it is not necessary.

Pontem gives me what I have realized I need most at this stage of my career: trust, flexibility, autonomy, and an open canvas to turn expertise into broader customer impact. It is a place where I can bring together operational realism and a strong belief that our industry still has plenty of room to get smarter.

Pontem, Latin for bridge, sits at an important intersection. It represents a bridge between data and discipline, between engineering depth and modern digital capability, and between identifying a problem and actually solving it.

Working with Pontem does not feels like a cold start. There is a legacy here from MSi and Assured Flow Solutions. When I left MSi in 2012, we have stayed connected through work, friendship, and the occasional chance to simply spend time together. Over the years, those ties never really went away. That shared history makes this move feel both familiar and full of possibility.

Tommy even brought me into AFS in 2015 to make a nationally-televised client testimonial video with the promise of meeting Terry Bradshaw. (That kinda happened - haha, see below).

I have long believed that flow assurance engineers are more naturally wired for digital leadership than many people realize. Long before digital transformation became the phrase of the day, many of us were already living in that world. We were building models from scratch, automating repetitive analyses, structuring data, interpreting patterns, and trying to understand what was happening between two sensors when the full picture was not directly visible. We were taking physics, data, and operations and turning them into better decisions.

That mindset matters now more than ever.

Pontem has the ingredients: deep technical roots, practical problem-solving capability, and strong digital instincts. In today’s environment, that combination matters.

That is the kind of work I want to be part of.

In the months ahead, I will be spending time reconnecting with friends old and new across the industry, learning from the global Pontem team, and helping shape how we support customers through Intelligent Operations.

This moment feels fitting.

For years, I built inside systems that benefited from my expertise, but did not always create room for my broader impact.

Now I have the chance to own that next chapter more directly.

That is why this feels like the right intersection.

A chance to turn expertise into impact.

Let’s reconnect.

Jason.Harry@pontemanalytics.com

Driven by a love for engineering (and solving complicated problems), I went on to pursue a B.Tech in Petroleum Engineering from the Indian Institute of Technology (ISM) Dhanbad, graduating in 2017. Those four years shaped not only my technical foundation but also my ability to survive on hostel mess food — a skill every engineer secretly prides themselves on.

Fun fact: In my first year of college, I initially joined the Petroleum Engineering branch — only to have it changed to Electrical Engineering due to internal sliding. But with some hard work (and a lot of chai-fueled study nights), I managed to earn the grades to slide back into Petroleum Engineering. So technically, I’ve switched branches twice before even starting my core courses!

Career Journey So far…

My career kicked off at PetroMinds Consulting LLP, where I joined as an engineering intern. This is where I built a strong base in reservoir and production engineering, fluid thermodynamics, and modelling.

After a year, I moved to Wood as a Flow Assurance Consultant, a role that played a huge part in shaping my technical expertise. Over five years, I got to work on steady‑state and transient flow assurance projects for major clients like Saudi Aramco, ADNOC, CNOOC, Woodside Energy, and Shell. I worked across concept select, FEED, and detailed design phases, dealing with all kinds of challenging (and fun) engineering puzzles around flow behaviour.
Fun fact: Flow assurance engineers often get excited about wax and hydrates — not your everyday dinner conversation topic.

After Wood, I joined KBR, Inc. as a Technical Professional in flow assurance, continuing my FEED and detailed design work for ADNOC and Shell. This deepened my exposure to large‑scale design environments and cross‑functional engineering workflows.

I then pivoted towards digital innovation by joining SLB as a Senior Production Engineer. Here, I worked on building an integrated digital surveillance tool for ONGC, focusing heavily on digital twin modelling and workflows such as network model validation, optimization, and well rate estimation.

Most recently, I had the chance to work with ExxonMobil as a Flow Assurance Engineer in operations. This experience gave me hands‑on exposure to real‑world challenges like hydraulic analysis, hydrate management, pigging operations, and WAG procedures. Working closely with principals and advisors helped me sharpen both technical and practical problem‑solving skills.

With diverse experience across consulting, digital innovation, and operator roles, I’m excited to have joined Pontem Analytics. Working with industry specialists and learning from their experience is truly energizing, and I’m looking forward to contributing meaningfully to the team.
Fun fact: “Pontem” means “bridge” in Latin — fitting, considering they connect deep expertise with real‑world impact.

Outside of work, I’m married and love spending quality time with my family. I also enjoy swimming, sports, and traveling, which help me stay active and balanced.

Fundamentally, nothing changes in the thermodynamics when dealing with hydrate formation or dissociation, and models are quite clear in determining stable or unstable hydrates for a given system. There are several models to evaluate the thermodynamics of a system that are accepted as standards in the oil and gas industry, such as Multiflash, PVTSim, and CSMGem.

Chemicals will shift the equilibrium curve. Depressurizing the system or providing latent heat will move the system’s thermodynamics relative to the hydrate equilibrium curve. One must, obviously, consider kinetics when asking how quickly a hydrate plug will form or melt. Colorado School of Mines Hydrate Plug Dissociation Tool (CSMPlug) predicts how quickly a hydrate plug will dissociate by solving transient heat-transfer and moving boundary (Stefan-type) problems with hydrate thermodynamic equilibrium constraints.

This post will provide an overview on the principle mathematic equations to solve the “how long” question and we’ll look at some examples to highlight how this may apply in the field.

Math Is Everywhere

As one may imagine, hydrate dissociation kinetics is quite dynamic and there is not a single equation to describe this process. Within CSMPlug, for example, there are multiple equations coupled together: transient cylindrical heat conduction, hydrate equilibrium thermodynamics, moving boundary, latent heat balance, and prescribed pressure history.

The model works by solving heat conduction at the surface of a hydrate cylinder while estimating dissociation due to temperature changes and pressure drops. Then it predicts how long these changes will take to fully dissociate the solid hydrate plug at the boundary. Once enough heat has been put into the system to fully dissociate hydrates at the surface, a new boundary is set with a smaller radius. This process continuously repeats until the plug is fully dissociated. Two limitations within the model are it does not simulate axial convection nor consider fluid flow dynamics (which makes sense to put this limitation in place as the hydrate plug is typically immobile!). In other words, the dissociation front moves radially inward as conditions change.

Math Fun (don’t worry, it’s Type 2 fun)

For a given hydrate plug, the series of continuously solved equations include:

A critical component for this model (and really the equation that determines how fast the hydrate dissociates) is at the phase boundary (shaded equation in the table), which is moving. The final solution is presented (hydrate is fully dissociated) when

s(t) = 0

What does this mean?!

For a hydrate plug, dissociation rate is controlled by Heat flux vs Latent heat requirement.

If the heat supply is small then dissociation is slow, which is the case if the flow line is insulated. A primary example of this that many will be (unfortunately) familiar with is melting a hydrate via depressurization alone. A flow line with a hydrate plug may begin with a dual-side depressurization, which initially takes the surface of the plug outside the hydrate equilibrium curve. There is an initial rapid dissociation (at the hydrate surface). However, this is an endothermic reaction and temperatures decrease locally (at the new surface) because heat is absorbed. This forces the conditions back across the equilibrium curve and hydrates are once again stable. The temperature can only warm up from heat provided from outside the flowline. If this is subsea, it will not warm up very much and will do so slowly. If the line is insulated, then that process is even slower.

This is why depressurization-only remediation can take several weeks in a subsea system with an insulated flow line. The seabed temperature is cold and does not provide much heat to the system and the U-value of the insulation severely limits the transfer of thermal energy.

Do I Really Need A Fancy Model For This?

There are sound engineering first-principle calculations out there, so won’t those work? That’s a great question. Let’s go through some examples and compare. Let’s look at a hydrate plug that has formed in a 20-mile subsea tieback. Here’s the system:

Let’s assume a 50 m full-bore plug forms in the line. The volume of this plug can be calculated as

V = (pi)r²L = p(0.127)²(50) = 2.54 m³

We’ll also assume hydrate density to be ~900 kg/m³, so the mass of the hydrate is 2,286 kg. This is a very conservative estimate, as it assumes 0% porosity and full conversion of hydrates (no free water/gas within this volume).

First Principle Calculations for Depressurization

As mentioned previously, depressurization is a viable method to dissociate hydrates because depressurization makes hydrates thermodynamically unstable. However, dissociation absorbs latent heat (it is an endothermic process and is self-cooling). After the initial flash dissociation, the plug temperature drops toward equilibrium. The heat to warm the system back above equilibrium temperature must then come from the surrounding seawater through a layer of insulation.

The heat flux from seawater is estimated by considering the surface area around the plug:

A = 2(pi)rL = 2pi(0.127)(50) = 39.9 m²

If we assume an insulted pipeline, h = 8 W/m²K (1.4 BTU/hr^.ft^2.°F), and a small driving force after depressurization (DT = 5°C) then,

Q = hADT = 8 x 39.9 x 5 = 1,596 W

Assuming the latent heat of a methane hydrate is 450 kJ/kg then the energy required to melt the hydrate is

Q_total = 2,286 x 450,000 = 1.03 x 10⁹ J

The dissociation time calculation becomes

t = (1.03 x 10⁹) / 1,596 = 645,000 sec = 179 hours = 7.5 days

If there is a more effective insulation and the heat transfer value is only 3 W/m²K (0.52 BTU/hr^.ft^2.°F), the dissociation time increases to 20 days. Technology giveth and technology taketh away! What was preserving thermal energy from the reservoir to carry along the flow line is now severely limiting thermal transfer from the seabed temperature into the system. This obviously helps to reduce the risk of hydrate formation, however once hydrates are present the remediation becomes more challenging. The chart showing various flow line insulations below highlights the differences and drives the point home.

Another important factor to consider is when pressures become very low (for a subsea tieback). Let’s consider a system that has managed to be blown down to 150 psi. The hydrate equilibrium temperature for Green Canyon Gas (a representative Gulf of America gas composition) is shown below. At equilibrium the system will slowly warm to 40°F by the surrounding ambient temperature, which will melt the hydrates. This will, in turn, cool the system down below freezing temperatures. The operation then risks ice formation!

This Catch-22 practical example is one reason why there are usually additional methods that are used in parallel for hydrate dissociation when depressurization is employed, which we’ll discuss below.

First Principle Calculations for Direct Electrical Heating

If we consider DEH for our dissociation example and assume 150 W/m is applied to the flowline, the total heating power applied to the hydrate plug is

P = 150 x 50 = 7,500 W

Let’s also assume there is only 85% effective heat applied, so P = 6,375 W. The same energy is required to dissociate the hydrate, which makes the time requirement

t = (1.03 x 10⁹) J / 6,375 W = 161,568 sec = 1.9 days

Obviously, reducing deferred production from 3 weeks to 2 days is a huge advantage. Unfortunately, installation CAPEX and OPEX costs to provide power in order to maintain the system for a 20-mile tieback must be considered against the risk of a hydrate plug occurring (once or multiple times).

How is a model like CSMPlug different from what we just did with first principle calculations?

The impact of these differences can be significant. CSMPlug will produce:

· Slower early-stage dissociation (based on internal gradients)

· Nonlinear time evolution

· Self-cooling effects

· Interface motion

The main deviations come down to three main sources:

Internal thermal gradient – In reality a hydrate plug, much like an ice cube in your Manhatten, will melt from the outside inward. While the surface warms relatively quickly, the core remains cold. In our simplified example, the equations assume a uniform plug temperature, constant heat flux, and no moving boundary resistance. There really is an internal thermal gradient where in the early stages the surface warms quickly while the core remains cold. This assumption produces an error ranging + 10 – 25% for the dissociation time.

Self-cooling during depressurization – During a pressure drop, the surface temperature drops, therefore DT shrinks and heat flux decreases. A lumped model assumes a constant DT. This produces a larger error – typically ranging + 20 – 50%.

Variable heat transfer coefficient – As the plug shrinks, the surface area decreases and the contact resistance changes. Our simplified calculations assumed a constant area. This difference accounts for +/- 10 – 20% error.

Long plug lengths – For very long plugs, pressure gradients matter more and dissociation may start at the ends. Lastly, melt rates are not uniform. A simplified model will under or over predict the dissociation time by up to 2x in some cases.

The first principle calculations tend to break down when considering hydrate plugs that form in subsea tiebacks because:

· Strong depressurization (large subcooling)

· Thick insulation

· Very low external DT

· Thick plugs (> 0.2 m radius with low conductivity)

· Partial dissociation with trapped gas pockets

In our DEH example, we estimated 45.6 hrs (1.9 days), whereas incorporating a full Stefan solutions (solid melting to a liquid) will give a solution ranging 52 - 62 hrs (+15 - 35%).

Bonus Topic: Chemical Options

Methanol and MEG (mono ethylene glycol) are often used to prevent hydrates from forming as a thermodynamic inhibitor (THI), and they both can be used to dissociate hydrate plugs. Without going into the thermodynamics too much, a safe approximation for the volume of methanol required for hydrate equilibrium temperature suppression is ~ 0.8°C / wt% MeOH and ~0.3°C / wt% MEG. The concentration component is an important factor to consider because free water is released from the hydrate as it dissociates, which dilutes the inhibitor. In practice this means THI must be continually fed into the flow line to fully dissociate the hydrate. This also means there must be communication across the hydrate plug or else the injection side of the line will keep increasing in pressure and the amount of THI will be limited (or the line will burst!).

In 2024, Koh et al. reported a model to describe a two-step mechanism to dissociate a hydrate plug. In the first step, heat is applied to begin the hydrate dissociation process and create annulus spacing between the pipeline wall and the hydrate plug (images 1 - 3 below; this is critical as one needs to establish communication across the plug). In the second step (images 4 - 6), THI is delivered to one side of the hydrate to accelerate the dissociation of the hydrate.

In addition to heat, the annulus spacing can be established through one-sided depressurization. The second step is dependent on the THI concentration gradient. The model calculates the accumulation of water and THI over time based on the dissociation. The model also takes into account the transient axial convection of the THI flow rate and associated thermodynamic changes caused by THI concentration changes. Other considerations include the porosity and permeability of the hydrate plug, heat transfer coefficient, hydrate plug length, and annulus spacing.

When applied to the field, the model predicted a hydrate dissociation time of 1 - 2 days, depending on THI flow rates, hydrate porosity assumptions, and time to create an annulus spacing (shown below).

In practice, the field was able to utilize multiple techniques to minimize safety risks and reduce the downtime to a minimum level. The super interesting point of this example is a model was developed to predict what is possible, yet unknown variables still existed: plug length, porosity, and location to name a few. The model validation still worked and by incorporating actual field data the remediation times provide valuable options to get the hydrate safely melted and production up and running.

In other words, the best solutions (in our humble opinions!) come about when combining scientific theory (what’s possible) with actual limitations / real world scenarios (what’s practical).

Subscribe now

My wife and I are deep in the “pickup and drop-off” days of parenting (IYKYK). The big one each day is the afternoon pickup. We’re lucky to have grandma help several days a week, but when it falls to me or my wife, there’s one big decision:

My wife’s answer:

“Get there 30 minutes early and be one of the first to get our daughter.”

My answer:

“Get there closer to actual pickup so I minimize the time commitment.”

Now, I really don’t want my daughter to be the last one picked up (too many days in a row of that and you start getting looks from the teachers). But I’m also generally very busy and want to avoid a big break in my afternoon if I can.

After enough firsthand pickup-line experiences at my daughter’s school, I’ve built a decent intuition for how to minimize my wait and that’s how I came to my answer above. But the data scientist in me is curious to try and prove it with statistical simulation and modeling.

Simulation Approach and Key Assumptions

To try and settle this, I wanted to build a model of the pickup process that enables me to simulate thousands of pickup days and see what happens on average. Below are some of the details of the model as well as its key assumptions.

The Model

What I’m looking to model would be classified as a discrete-event simulation. Within that, the pickup queue is modeled as an M/G/k queue (a common type of queue process):

• Cars arrive randomly following a non-homogeneous Poisson process — meaning they don’t all show up at once, but cluster around the bell in a bell-shaped pattern (no pun intended). The peak arrival rate actually hits about 5 minutes before the official pickup time.

• Service times are random too meaning how long each car takes to pull up, load a kid, and pull away follows a log-normal distribution (most cars are quick, but every now and then someone needs to track down a missing lunchbox, I’ve even seen a shoe fall off).

• Multiple loading spots operate in parallel meaning just like the real curb, several cars can load at the same time in a first-in, first-out order.

I then ran 3,000 Monte Carlo simulations — each one representing a full school day — to build up a statistical picture of wait times across every possible arrival time.

Key Assumptions

• ~240 cars arrive per pickup session (roughly 60 kids per grade after accounting for after school activities for some of them × 4 grades)

• 4 parallel loading spots at the curb

• Average service time of 30 seconds per car with a standard deviation of 15 seconds, including pull-up, load, and pull-away (our school seems pretty efficient)

• Arrival peak is 5 minutes before the bell, with ~68% of all cars arriving within ±13 minutes of that peak

• The simulation window spans from 40 minutes before to 30 minutes after the bell (picking your kid up 30 minutes after the bell would be pretty bad)

• Cars are served first-come, first-served — no cutting, no VIP lane for parents who “just need to grab one kid real quick”

• Each simulation day is independent — no memory of yesterday’s chaos

Results

We’ll start first with car arrival distribution which shows when most cars tend to arrive. A quick sense check confirms this is reasonable. Most parents tend to arrive before the bell gets out, which is why we see the peak before +0. Having said that, there are still some arrivals after the bell gets out too (like me).

Next is what the queue (carpool line) looks like over time, factoring in the loading of kids once we reach +0. There’s always someone who shows up as soon as the doors open (~40 minutes before pickup). The line builds until it peaks right around +0 when kids start getting loaded. From there, cars start leaving the queue, but a few stragglers (again, like me) are still arriving so it stays somewhat flat before dropping off.

Finally for the most important part, combining the two pieces of information above, we can get a sense of the total time you spend getting your kid, depending on when you arrive. We can also look at this in distribution form for an early arriver vs. later arriver further below.

What This Means for the “Arrive Late” Strategy

The data tells an interesting story:

Arriving early is the worst strategy for total time spent (sorry honey). If you show up 30 minutes before the bell you’re sitting in line for over 31 minutes on average. Yes, you picked up your kid almost first, but you waited at least 30 minutes up front to be able to do that.

The wait time begins to taper off right when the bell goes off to ~20 minutes. This is when the queue is at its maximum point, but you’re only waiting for the cars in front to pick up their kids because pick up has already started.

The magic spot: if you wait it out and arrive 10 minutes after the bell, your average wait drops to just 16 minutes which is shorter than someone who arrived 20 minutes early.

Wrapping Up

You’ll notice that arriving 20 or even 30 minutes after the bell means you can almost pull right up and grab your kid. That makes sense intuitively, but I choose to live by the “happy wife = happy life” motto and 10 minutes after the bell is about as far as I’ll push it.

P.S. Other Scenarios are Possible Here

What I showed above, was custom tailored to my daughter’s school and from my sense checks on it, it seems to match. But if we increase the number of kids to be picked up, or we change how long it takes to pick someone up, or any of the other parameters, we can end up with quite different results. And in particular, ones where you would want to get there extra early if you can.

Here’s an example of loading 400 cars (vs. 240) and a slightly longer pickup time per kid. In this case, there’s a sweet spot about 20 minutes before pickup starts.

While Pontem works on the premise “Prevention is always better than the Cure” we also have excellent experience in the Hydrate Remediation process. In our latest post we share some fundamental processes and tools required to safely locate and remove a hydrate plug.

The “safely” part is super important because there are numerous examples of where hydrate plugs have become dislodged and caused catastrophic damage to flowlines, pipelines and equipment causing serious injury sometimes and in very rare occasions loss of life.

The removal of a hydrate plug is so dangerous because if you dislodge a heavy, compact hydrate plug with enough pressure behind it these can travel at speeds up to 200- 500 miles/hr. In one example, we modeled depressurization procedures involving a 200’ hydrate and found one option to be particularly disastrous for the subsea infrastructure.

Hydrate plugs come in all shapes and sizes. Below is a picture of compacted hydrates taken from a lab test with 100ml of fluids. Just imagine what a 1000 lbs hydrate plug would look like.

Unsurprisingly it can be devastating to whatever and whoever it runs into. The cartoon graphics below give a hint of the damage it can do.

This shows why it is so important to consider all the safety implications when managing the hydrate remediation process.

The aim now is to share with you the hands-on skills needed to safely remove a hydrate plug. For this post we will focus on:

1. How to locate the hydrate plug

2. Dissociate and remove the plug safely

3. Then restart the system without forming another hydrate plug.

While not all encompassing, this post should prompt enough questions to begin discussing the most appropriate next steps to get your system back online!

Subsea Hydrate Blockage Location Detection Technologies

How the hydrate plug is detected is very much down to the length of the line and whether you can easily access the outside of the pipeline. Accordingly, a topside pipeline is the easiest, whereas buried onshore and subsea pipelines are far more difficult.

It is also worth saying that no single plug locator method is perfect, a combination of approaches is often recommended to achieve precise results.

Recent advancements have introduced non-invasive tools specifically for deepwater flowlines where access is limited. The table below summarizes some of the available detection techniques.

Acoustic Pulse Detection Method - Image courtesy of Find-Block® Paradigm Group

Pontem are in discussions with a supplier of one of the acoustic detection tools to combine our modelling expertise in the use of physics based models to predict hydrate plug locations.

Hydrate Blockage Remediation Techniques

There are several different techniques that can be used to remove hydrate blockages but the most commonly used are:

Thermodynamic Hydrate Inhibitor (THI) Injection

THI’s such as Methanol, MEG and TEG are used to melt the hydrate. Most facilities will have access to at least one of the chemicals, if not two.

THI’s work by altering the freeze point of hydrates forcing it to melt at the system pressure and temperature.

For a THI to work it must actually contact the hydrate plug, which is not always possible. There is almost always fluids between the plug and either end of the flow line, which makes THI placement a non-trivial exercise. Additionally, it is not recommended to use Methanol to melt a hydrate plug in a vertical pipe as Methanol will sit on top of the free water that is released as the hydrate starts melting due to it being lighter than water. This barrier stops the melting process.

Depressurization

If the injection of a THI has failed to remove the hydrate plug, the next option is to depressurize the pipeline to move the system outside the hydrate envelope. When it comes to depressurization there are two possible options:

• Dual sided depressurization - The most common and preferred method used is two-sided depressurization because it minimizes the chance for a large pressure differences across the hydrate plug occurring, thus preventing the hydrate plug becoming a high velocity projectile.

• Single sided depressurization - Usually not recommended as you run a greater risk of having a high pressure differential across the hydrate plug (i.e. >10 bar / 145 psi). However, there are instances where single sided depressurization has been used successfully but it needs to be done in a controlled manner and requires a thorough risk assessment before it is performed.

Impact of multiple plugs in a pipeline

In the case where multiple plugs have formed in the pipeline, extra caution needs to be taken even with two-sided depressurization. This is because in the case of multiple plugs, high pressure gas may still be trapped between two impermeable plugs so unbeknown to the operator there may be a significant dP across one of the plugs causing it to become a projectile if loosened due to the depressurization process.

Cooling during Depressurization

One of the draw backs of using depressurization to dissociate a hydrate plug is that sub-zero temperatures can be generated. This is manageable in uninsulated flowlines as the ambient temperatures warm up the system within hrs/days allowing the plugs to melt. However, many deepwater oilfields use highly insulated flowlines (e.g. U-values <10 W/m²K), which prevent the internally generated cold temperatures from being dissipated prolonging the hydrate dissociation process. This may mean that the ambient temperatures take weeks / months to move the system outside the hydrate region.

In flowlines that sit in sub-zero seawater temperatures this problem is exacerbated as it can lead to ice formation. This is a problem because no amount of depressurization will remove an ice plug. The only way to remove it is to somehow add some external heat to the pipeline, which we will talk about more later in this post.

All of this hopefully shows the complex thermodynamic process that occurs during hydrate plug dissociation and what needs to be evaluated upfront.

Use of heat to warm up the pipe

There are a number of heating options that have either been tried or proposed. These are:

1. The Petrobras SGN method

2. Subsea Active Electrical Heating

3. Electric Blankets and Heat tracing for topsides piping

Petrobras SGN Method

If you are a chemistry geek this one is for you. SGN stands for Self-Generated Nitrogen. SGN is a patented, thermo-chemical technique primarily used by the Brazilian National Oil company, Petrobras to remove hydrate and organic (paraffin) plugs in subsea flowlines, trees, and wells. This method relies on an exothermic redox reaction between two specialized, nitrogen-containing salt solutions (referred to as “C” and “N” solutions) in the presence of a catalyst to generate a high enough temperature to melt the hydrate plug. This is delivered via Coiled Tubing (CT)

• Reaction: Ammonium chloride (NH₄Cl) and sodium nitrite (NaNO₂) react to produce nitrogen gas, sodium chloride, and water, releasing a significant amount of heat (∆H = -75kcal/mol).

Active Electrical Heating

Direct Electrical Heating (DEH): This is the most commonly used active heating method, which uses the steel pipe as an electrical resistor to generate heat. An AC current is fed via a cable to the far end of the wet insulated pipeline, where it connects to the pipe, causing current to flow back through the steel, generating heat through both Joule and skin effects. This heat then passes from the metal pipe wall into the fluid in the pipe. A drawback of this method is that the heating process is not always uniform due to the resistive properties of each section of the pipe being slightly different.

Electrically Trace Heated Pipe-in-Pipe (ETH-PiP): This works by using up to 3-phase trace heating cables strapped to the outside of the inner pipe of the P-i-P set-up and within the insulation layer of a flowline. Powered by subsea transformers, these cables generate heat via the Joule effect, which involves passing an electric current through the heat trace cables, generating heat that warms the inner production pipe and fluids to above the hydrate dissociation temperature. The copper cable achieves much better temperature control compared to DEH because the heat generated is consist throughout the length of the cable.

Subsea Hydrate Remediation Skids

When it comes to hydrate blockage removal, time is money because as long as the hydrate plug is there no production can happen. This is why the use of Active Electrical Heating is so useful. You can virtually flick a switch and the hydrate plug will start melting.

However not every field can afford active electrical heating and in a lot of instances the use of THI and/or depressurization just doesn’t work. This has driven the subsea service provider community to develop Hydrate Remediation Skids or Composite Coiled Tubing units, which attach to deepwater flow lines to help remove hydrate blockages quickly.

Hydrate Remediation Skids for System Depressurization

These are designed to connect to subsea manifolds and/or Production XTrees via a ROV to allow for quick dual sided depressurization of the system to <10bar and to inject THI into the system. Below are Hydrate Remediation Skids supplied by Subsea7 and Oceaneering that have been developed for this purpose. There are other providers of this type of skid as well.

Composite Coiled Tubing Units

There are a number of specialized coiled tubing technologies that use composite coiled tubing and specially designed heads to transport the coiled tubing along subsea pipelines. The lightweight and more flexible nature of the composite tubing allows the technology to do things the traditional stainless steel coil tubing can’t. Like reaching up to 20kms along subsea pipelines and being able to navigate through multiple 5D bends.

The composite coiled tubing can be inserted directly into the outlet of the subsea riser on the topsides process or via multiple subsea intervention access points (i.e. 2” ‘hot-stab’s) on subsea equipment. After insertion, the coiled tubing can be used in multiple ways to remediate hydrates. These are:

• To dose THI directly onto the hydrate blockage. The coiled tubing is fed down the riser or via the intervention points in the subsea equipment to the immediate vicinity of the blockage and THI is injected directly onto the hydrate.

• To inject hot fluids (e.g. diesel) directly onto the hydrate formation for thermal dissociation using the same methods described above.

• To inject gas to the base of the riser to unload the liquids, and remove the subsequent static head, from the system, attaining hydrate free pressures subsea. This is only required if the system cannot be depressurized via the usual flare path due to high liquid hold-up in the riser.

Hydrate Remediation Successes

As with everything it is important to know that there are effective solutions out there with respect to getting rid of hydrates. Below is a small list of successful cases where hydrate blockages were remediated successfully.

The above data demonstrates that most recorded hydrate remediation cases are gas or gas condensate systems and that the majority of the blockages occurred due to mal-operation or failure of the chemical injection and/or gas dehydration systems. Furthermore for uninsulated lines, hydrate plug removal took hours / days but for the only insulated line case included, hydrate plug removal increased to days / weeks. This corroborates the description previously supplied regarding what happens when trying to dissociate a hydrate plug in an insulated line.

The Hydrate Plug is gone so what to do next

Finally, there is no point in successfully removing the hydrate blockage without preparing the system properly for restart. Otherwise, you will probably just re-block the line.

To get the system up and running again Pontem recommend the following:

1. Remove the uninhibited water that caused the hydrate plug in the first place. This can be done by running a low-pressure pig at this point to remove the uninhibited fluids.

2. If you can’t run a pig, flood the system with as much Methanol or MEG as possible, with one eye on the costs and the other on how the aqueous phase will be disposed off once it comes out of the pipeline.

3. Once you think you have eliminated/inhibited the water in the line, make sure the chemical injection system or gas dehydration system is back up and running.

4. For the systems that normally operate outside the hydrate region during steady state flow, it is recommended to restart the system at as low a pressure as possible after dosing with Methanol or MEG until the flowing temperatures are above the hydrate dissociation temperature.

If you follow those points you will give yourself every chance to get your system up and running again but unfortunately there are never any guarantees in this game.

We hope this post has improved your understanding of the Hydrate Remediation process and as ever there is always more to these things than you think. It is not until you experience having to do this that the gravity of the process sinks in. This is where Pontem can help because we have that experience and are more than happy to share it so don’t be shy, contact us.

Subscribe now

Background

Pigging offshore pipelines is expensive, disruptive, and not without risk. A single pig run can cost hundreds of thousands of dollars, require careful operational coordination, and — at low flow velocities — introduce a very real risk of stalled or stuck pigs.

So when production declines in a late-life asset, a familiar question comes back into focus: Can we safely do this less often?

For this study we were involved in, the motivation was clear. The pipeline had been operating at low rates for years, pigging returns showed consistently small wax volumes, and yet pigging was still being carried out on a relatively frequent basis. Reducing pigging frequency would deliver immediate cost and risk reduction — if it could be done with confidence.

But there was a catch.

The fluids in the line were no longer the same as in previous years. Production was increasingly dominated by different wells, with different (higher) wax risks. Even at low rates, the concern was simple and valid:

Are we still seeing low wax because the system is benign — or just because we haven’t been caught out yet?

Answering that question — properly — was the real objective of the study.

What the Historical Data Did (and Didn’t) Say

Looking back at pigging history, the headline story appeared reassuring.

Over the last several years, pig returns consistently showed around 1 kg of wax or less per run, even with long intervals between pigs. On the surface, that trend alone might prompt someone to say “what’s the problem here?”.

However, relying on history alone masked two important uncertainties:

• Earlier pigging data reflected different production mixes, not the fluids currently dominating the line

• Historical low wax recovery did not necessarily bound what could happen next, particularly if waxier streams became more influential

In other words, the question was not “Has wax been low?”
It was “Does the historical behaviour still apply to today’s fluids and tomorrow’s operations?”

That distinction is what pushed this study beyond a simple trend review.

Letting the Data Define the Problem

Rather than starting with a model and asking it to predict behaviour, the workflow deliberately started with data aggregation and reconciliation across numerous sources, something we do often at Pontem.

In this specific case, it included:

• Daily averaged pressures, temperatures, and flowrates

• Detailed pig launch, receipt, and wax recovery records

• Laboratory wax appearance temperatures and heavy-end content

• Well-by-well production contributions over time

Only once the data was pieced together and told a consistent story was modelling used — not to invent outcomes, but to stress-test what the data implied under current and future conditions.

Turning Data into Bounded Predictions

A transient multiphase flow and wax deposition model was then used as an interpretive tool.

Instead of focusing on internal modelling parameters, results were framed in terms of expected and conservative outcomes — effectively P10 and P90 estimates — bounded by what had already been observed in the field.

The calibration logic was straightforward:

• Higher historical wax recovery cases were used to define conservative behaviour

• Recent low-wax pigging cycles anchored the expected case

• Historical fluid blending based on waxy and non-waxy wells was taken into account

• Predictions were only accepted if they remained consistent with above

This approach ensured the model stayed tethered to reality, rather than drifting into theoretical extremes.

What the Results Showed

Across current and forecasted production conditions, both the data trends and the calibrated simulations told the same story:

• Wax accumulation rates remained very low

• Predicted wax layers were fractions of a millimetre thick, even over long durations

• Resulting internal diameter reduction was negligible, with no meaningful impact on hydraulics or operability

Low flow velocity was recognised as a potential risk, but analysis showed it to be acceptable for this system. Predicted wax layers remained extremely thin, with no meaningful impact on hydraulics or blockage risk, and pigging could still be executed safely when required.

Crucially, the results did not rely on optimistic assumptions. Conservative scenarios still showed wax behaviour that was slow, stable, and non-threatening.

The Moment of Truth: Field Validation

The strongest test came at the end of the year, when the line was pigged under the very conditions the study sought to assess.

Approximately 3 kg of wax slurry was recovered, which was 3 times more than what any of the last 5 years data showed. The pig showed no damage, and the wax itself was soft and mobile — entirely consistent with expectations.

This matched closely with the predicted wax accumulation P10/P90 range for that operating window.

That alignment between prediction and physical recovery was the most important outcome of the work. It demonstrated that:

• The predictions were realistic

• The system behaviour was understood

• Future decisions could be made with confidence rather than caution

Conclusions

With data trends, calibrated modelling, and field evidence all pointing in the same direction, the conclusion was clear: routine quarterly pigging was no longer technically justified.

The recommendation was to extend pigging frequency to once per year, with the option to reduce further if future runs continue to show minimal wax recovery. Pigging was reframed as something done for inspection readiness, not routine wax control.

As previously mentioned, the economic and operational implications are significant. Reducing pigging frequency from quarterly to annual delivers material cost reduction and, just as importantly, reduces operational exposure. Fewer runs mean:

• Less exposure to pigging at marginal velocities

• Lower risk of stuck pigs and recovery operations

• Reduced operational complexity offshore

This case matters because wax risk does not scale neatly with declining flowrate, and legacy practices can persist long after their original justification has faded. By grounding decisions in data, evidence, and validated predictions, it is possible to move from conservative habit to confident optimisation — and in some cases, the safest option is not another pig run, but a better understanding of what’s actually happening inside the pipe.

The most honest theme for us is something much simpler: show up

As Pontem has grown these last 1000+ days, there is a lot to reflect on. While there are definitely highlights, the trip is best summarized in the aggregate. Taking small - sometimes unnoticeable, always intentional - steps to build the company we want to work for (and the people we want to work with).

This is article #167 in the series. By consistently showing up, we “have a stack of undeniable proof that we are who we say we are” (thanks Hormozi). And while showing up applies to everything we do, one place it is visible is in these articles, where we make a point to get everyone to contribute with regular cadence. That’s once/week consistently for the last 3 years that someone at Pontem has “showed up”, put effort in, kept their word, and delivered on what they were asked to do (and we didn’t even make a “67” joke, because that’s so last year…)

So in honor of our 3rd birthday, we thought we would reflect back on a few of our Greatest Hits.

Top of the Pops: Most Popular Topics

We have covered a range of topics, some technical and some editorial, but we can reliably count on three main themes if we want to drive clicks. Any guesses? Our top 3 topics (measured by views) have been:

• Hydrates

• OLGA

• Sports Analytics

Especially if we can find a good picture, our audience always seems to show up for thought-provoking and thirsty pictures of hydrates. Lucky for us, we have some of the most well-respected hydrate experts in the world under our umbrella and have collaborated with others (academia / industry) frequently in the past. Keep your eyes open for more - gotta give the people what they want and play the hits sometimes.

Chart Toppers: Most Viewed Posts

Most major sports have specific numerical milestones (i.e. 500 Home Run Club, sub-4 minute mile, etc.) that guarantee induction into their Hall of Fame. The best of the best. We are no exception, as we have a short-list of top-performing posts inducted into ours. In order, here are our most popular posts since inception.

Congratulations to Stephen Hamilton for having the #1 read post in company history. He truly is Mr. Social Media.

Deep Tracks: Down the Rabbit Hole

Not surprising, the list of topics has grown considerably since our 1st birthday recap. In pulling this together, it was fun to go back and re-read some of our earlier work, as a lot has happened in the world these last 3 years. We were also reminded that Pontem’s founding date coincided closely with another startup you might have heard of: ChatGPT.

It’s fair to look at many companies operating today as BC (Before ChatGPT) and AC, as the way we process data, look at tech stacks, and build companies has been forever altered. Launching right on the cusp of this offers a unique perspective. In fact, we even opined on this early with our thoughts on ChatGPT, if you want to see if we hit the mark:

Aside from that, poking around our Substack reveals thoughts on Machine Learning, NBA gambling outcomes, controlling OPEX in low commodity price environment, and various other good reads - many of which were written before the AI craze went ballistic - so interesting to see some of these data-driven feral ideologies play out. We will share some of the deep tracks throughout 2026 that you might have missed when they were originally released. It’s fine, much like Taylor Swift, we own all our masters, so we can reshare.

So, if you keep reading, we will keep showing up and writing (hopefully interesting stories). And while you are down the rabbit hole, stop, click around, and read a few of the classics…

Subscribe now

At Pontem, we are getting excited for the upcoming Winter Olympic Games, hence the catchy subtitle to talk about creating a smooth sampling and analysis program.

At its core, a good analytical program will provide enough leading and lagging indicators to know if a system is operating efficiently and when it is going off track. The trick, of course, is for the data to be useful and accurate, and while it sounds easy enough, the truth is collecting and running samples must be done “by the books” to avoid contamination or erroneous results. Here, our post explores some best practices to ensure your sampling and analysis program hits the bullseye.

All manned oil and gas facilities have a requirement to routinely sample and analyze the oil, gas and water to ensure that specifications are met to allow the hydrocarbon products to be exported for sale and the water to be discharged overboard. Failure to meet these specifications can result in fines, devalued product, or possibly outright rejection of the hydrocarbon product from the receiving company. Additionally, internal quality targets are often used to maintain asset integrity (i.e. corrosion and fouling) and/or maximize production rates. Failure to routinely sample and analyze the production fluids can significantly impact oil and gas production, hitting the bottom line of the operation.

In addition to the above, there will be process upsets that require numerous samples to be taken for analysis and troubleshooting to return the system to baseline. This can be a very important part of the sampling and analysis process because getting it right may mean the difference between producing and not producing hydrocarbons. Taking on deferred production due to unplanned shutdowns is definitely not “Cool Runnings”. See what we did there?! 😊

The ideal scenario is one in which samples are collected, analyzed, and results are in hand in a matter of minutes to hours. This requires the ability to do the majority of the analyses required on-site. To do this effectively means you need a lab. Sometimes it can be a bit of a squeeze to fit a decent lab onto offshore facilities but not quite as bad as this:

Correspondingly, the size of the lab limits what you can do and the equipment you can fit into the lab, but you still have a wide range of analyses to perform. Large footprint analyzers, rare or difficult to source solvents or disposables, and highly specialized training are all things that will limit the applicability offshore. We will go into exactly what type of analyses are needed later in this post.

So, with this limitation, it is unlikely that there will be much state-of-the-art equipment and sometimes you do feel like you are performing “bucket chemistry.”

The good news is that with the development in portable chemical analytical equipment there is still a lot that can be done and to a reasonable accuracy.

Taking that into account we will now go into a bit more detail on what analysis you need to do for oil, gas and water streams.

Sampling and Analysis Requirements

This process should be started during the design phase where the sample locations are initially identified. This process needs to make sure of the following:

1. The sample point location and orientation is set correctly in the pipe or vessel. It is critical to understand what type of sample is being pulled at the sample point location so it can be spatially placed correctly. This includes understanding of the flow regime and/or how the vessel is configured. Nothing worse than getting water when you wanted oil! This includes placing the sample point at the 12 o’clock vs 6 o’clock position. Or placing the sample point near the oil outlet or water outlet for a vessel.

2. The sample point is accessible. In many instances, sample points have been located where they can’t be reached or that they are in such a small space you can’t get a sample bottle to fit underneath them.

3. Is the sample point in a horizontal or vertical pipe? This will make a difference on what type of sample you get as gravity will play a part in what type of flow regime you will have, which in turn will affect the phase that you will sample depending on where the sample point is located.

Acceptable and Not acceptable Oily Water sampling points

The pictures below also give examples of where to locate sample points in vertical and horizontal pipes

Safety First

Taking oil and gas samples can be a hazardous process and needs to be a managed accordingly. With sometimes having to break into high pressure and temperature systems. This may result in the sampling process requiring a work permit / risk assessment or as a minimum be supervised by an Operator. Things to look out for are:

1. Venting of hydrocarbons
   This may cause small volumes of gas to be released, so this gas needs to be vented safely.

2. Adequate training for the person taking the sample
   The person taking the samples must know the HSE risks and follow properly established protocol to remove or minimize risks to a manageable level.

3. Sample point integrity
   Sample points should have double valve isolation and bleed valve assembly. i.e. double block and bleed with ideally a pressure gauge.

   

4. Personal Safety Requirements
   This involves the use of safety glasses/goggles/face shield, hard hat, appropriate gloves, safety shoes, clothing and breathing apparatus if gas samples with high levels of H₂S are to be sampled.

5. Impact of Static Electricity
   Static electricity is a risk during sampling. Appropriate earthing procedures should be carried out. Refer to ASTM D4865-09 Standard Guide for Generation and Dissipation of Static Electricity in Petroleum Fuel Systems for details.

6. Correct Sample Labelling and Sample Containers
   For transportation it is important to make sure any sample taken is labelled correctly and is in the right size and type of container for the transportation method used. Proper labelling in the US, for example, includes DOT hazard diamonds with symbols and numbers, alongside GHS pictograms, signal words, and statements.

7. Radioactivity of the Sample
   For most liquid and gas samples this is not an issue due to the low levels of radioactivity present but if a solid sample has been taken, it is important to understand whether the solid contains NORM (Naturally Occurring Radioactive material).

As well as taking samples safely, you also need to take them correctly. There are good and bad ways to take a sample. Ideally you want to make sure:

• The sample bottle is compatible with the sample to be taken. E.g. oil samples should be taken in lined metal cans or glass bottles and not plastic containers as the oil can react with the plastic container depending on the type of plastic used. (We’re looking at all you milk carton users!!).

  

• The sample bottle is clean and uncontaminated.

• The sample line from the sample point to the sample container is adequately flushed to an appropriate drain or waste container beforehand to ensure a homogenous sample is obtained.

I once went offshore to survey several unmanned and manned platforms in the North Sea. The operator had reported some stubbornly high OIW figures on a couple of platforms. On one of the manned platforms, I went out on the plant with the operator, who normally took the water samples for OIW analysis, and got him to take a sample as he normally would. Straight away he took the clean sample bottle, rinsed it out with the first water that came out of the sample point and then disposed of that water. Once that was done, he started taking the actual sample to be analyzed.

I immediately realized why the OIW levels were higher than they should be. This was due to the operator washing the clean sample bottle with the sample line flush. This water wash step left a film of oil on the inside of the bottle even though it was emptied out prior to taking the sample. The analyzed water sample produced an OIW value that was a combination of the fluids from the flush and the actual water sampled. It turned out this increased the OIW value by 20-30%. The corrective action was to flush the line into a suitable disposal bucket or waste stream and then collect the water sample with the clean, previously unused collection bottle. The OIW values immediately returned within required specifications.

So what samples should you take

Once you know where and how the sample will be taken, it is important to understand why you want to take the sample in the first place. Most samples are taken to verify that the oil, gas, and water are within specifications. It is important to note this often requires more than one analysis per fluid, as several parameters need to be monitored. The following sections give an indication of what samples should be taken and what analysis to perform.

Oil Samples

The main variables requiring monitoring in crude oil producing operations are:

With respect to sand, measuring the amount of sand being produced is a lot more difficult than you would think. It needs a specialized sand sampling system. See the picture below:

Samples of sand need to be taken from sand removal systems to measure the amount of entrained oil before the sand is discharged overboard. There is a risk for oil sheens if oil levels on the discharged sand are too high similar to the one in the picture below.

Gas Samples

As with oil, there is a lot of effort put into monitoring the quality of the gas to make sure it is suitable for export. To get the right gas specification there may be various gas treatment systems being used such as:

1. Gas dehydration

2. Acid gas removal (i.e. CO₂ and H₂S)

3. Condensate dew-pointing

4. Mercury removal

To ensure these gas processes are working as effectively as they should be, gas samples need to be taken and analyzed at regular intervals. This may involve the use of online sampling and analysis systems.

Even with online systems being in place, there is a need for manual sampling and analyses to be performed offline to validate the performance of the online systems.

Operations that are relatively simple and/or where plant upsets are infrequent or low impact may require only periodic sampling and analysis of the gas streams to be performed. For example, this may mean only a simple chromatographic analysis to C8 and/or a dew point measurement is sufficient.

Where more in-depth analysis is required, gas samples may be further tested for the following gas properties:

• Calorific value

• Relative density

• Hydrocarbon and water dew points

• Mercury

• Concentrations of sulphur compounds

• BTEX

• Inerts

As well as the above gas properties periodic monitoring of the glycol used for gas dehydration or hydrate inhibition may also be required.

Finally, some of the above gas properties can be used for monitoring the gas process performance or for quality control of the gas being exported. These properties are used as key features in the quality specifications attached to gas sales agreements, so any analysis needs a high degree of accuracy.

Water Samples

Although water is not a revenue generator, there are multiple system health checks dependent on proper and routine water analyses. These are:

It is worth noting that when it comes to water, the environmental analysis requirements are based on where the water is being discharged. If it is offshore (>12 miles from shore), well away from the coast line, then the analysis requirements are much less onerous compared to water discharged onshore, in a river/lake or near-shore (<12 miles from shore). This is because the impact the discharged water will have on humans and the indigenous flora and fauna is likely to be negligible for offshore scenarios thus requiring less tests. However, when the discharge is onshore, near-shore and in rivers or lakes then you will likely need to analyze the composition of the water far more closely. You can see these differences in the lists below .

Environmental requirements - offshore

• Total Oil and Grease (TOG) - aka Oil-in-Water (OIW)
  Depending on the global location that can just be measuring the dispersed (insoluble) oil in the water or the total hydrocarbon content of the water that includes both the dispersed and water-soluble hydrocarbons

• Monitoring “grey water” / sewage discharges offshore

Environmental requirements – onshore / near-shore

• Total Oil and Grease (TOG) - aka Oil-in-Water (OIW)

• Total Suspended Solids (TSS)

• Total Dissolved Solids (TDS)

• Total Organic Carbon (TOC) and/or Chemical Oxygen Demand (COD)

• Total Nitrogen

• Total Phosphorous

• Escherichia (MPN/100ml)

• Faecal enterococci (MPN/100ml)

• pH

• True colour (Wavelength/ Maximum spectral absorption coefficient)

• Acute toxicity (UT)

It is worth noting that the above list of tests are some of the most common tests that need to get performed. Depending on where you are in the world will determine how many of them you will need to do.

Process Performance requirements

• Total Oil and Grease (TOG) - aka Oil-in-Water (OIW)

• TSS

• Bacteria ATP

• Bacteria qPCR

• CI Residuals

• SI Residuals

• Ca/Mg ratio Mn²⁺ and Fe²⁺ conc.

• pH

• Water Composition

The above analyses are the requirements for the produced water when managing things like corrosion, scaling and water quality.

If you are also using seawater injection into the reservoir to help enhance production, then there is a whole new set of analyses that need to be performed to ensure the seawater is the right quality for injection. Specifically, the seawater injection stream should be tested for:

• Dissolved Oxygen

• Free and Total Chlorine

• TSS

• Water Quality

• SO₄^2- (this is particularly important if sulfates are a concern and/or a Sulfate Reducing Unit (SRU) is in place)

These are the sampling and analysis requirements for the main production and injection streams but there are a number of other parts of the process that need to be sampled on a frequent basis. These include but are not limited to:

• Off Spec Oil Tanks

• Fuel Gas

• Diesel Storage Tanks

• Closed Loop Cooling Medium

• Closed Loop Heating Medium

• Lean TEG

• Rich TEG

• Fresh Water Generators

• Slop Oil Tanks

• Slops and Bilge Water

• Subsea Chemicals

• Subsea Hydraulic Control Fluid

• Sewage Treatment Package

As you can see most oil and gas facilities need to take and analyze large numbers of samples just on a routine basis. This increases further when something unexpectedly goes wrong and may even start to include solid samples but how to deal with solid samples is probably another Substack post in itself.

Sample Data Management – It’s our Thing!!

As we just described, there are a multitude of oil, gas and water samples that need to be taken on a routine daily, weekly or monthly basis. As well as the manual samples there are online analyzers that you can add to the system that will analyze in real time things such as

• BS&W

• OIW

• Water content of gas

• O₂ content of injected seawater

• Corrosion

• Particle counts

Even when you have online analyzers, manual samples are still required to ensure the analyzers remain calibrated.

When you have this instrumentation and its combined with the manual sampling and analysis data this provides integrated data feedback with the data generated directly via the analyzers and instrumentation, which is then used to maximize the efficiency of the overall system by managing the data generated.

Doing something useful with the data to improve processes is really the point to all the hard work of sampling and analyzing properly. The following two examples highlight the difference between collecting samples in a single point in time to develop a risk analysis compared to a monitoring system that delivers trends and actionable insights over time for an entire system. The table below shows 7 risks identified from sampling a Permian system. Framing those risks in a likelihood and consequence (or level of urgency) matrix provides a framework for creating an effective mitigation or risk removal program. Corrosion and souring risks are identified and a path forward is recommended, however it will be up to the field to maintain these programs and to set threshold levels for a monitoring program.

In contrast, online dashboards that trend systems continuously are able to identify risks as fluid properties change. In the example below, 7 wells are monitored on a continuous basis with plots of P/T/chemical injection rates highlighted on a dashboard, along with thermodynamic risks for asphaltene instability and a simulated deposition rate. The system provides real-time feedback, and in the case of W-BF-2 well the dashboard is recommending a 25ppm increase in AI to reduce the risk from medium to low and examine BH pressure data to recalibrate the system.

Yeah, But I Don’t Have Time For All That…

We get it: sampling, analyzing, and processing all that data can feel daunting and cumbersome. And it can often appear that a system is running smoothy. Going through the motions without a real understanding of what is truly happening will eventually bite you. The US bobsled team just found this out at the World Cup Bobsled event while tuning up for the Winter Games. Good timing for our post…bad timing for our Olympic hopefuls!

Without those leading and lagging indicators that will crop up in routine sample analyses, bigger problems will cause massive issues, making it more difficult and more expensive to right the ship.

The good news is you don’t have to go it alone. Partnering with Pontem Analytics is a sure fire way to set up and establish a robust testing regime that includes all the metrics you will need without the ones you don’t. Reach out and start the conversation. We’re happy to help.

Thanks for reading Data and Discipline! Subscribe for free to receive new posts from Pontem Analytics.

Last year, we shared round 1 of Pontem x SPE, which was a piece showing our longstanding involvement with SPE and a few recent announcements at that time. If you missed it, you can read that piece here:

At Pontem Analytics, we rely on the same spirit that SPE was built on: sharing hard-won lessons, staying current on the best technical thinking in the industry, and helping the next generation of engineers grow into confident professionals.

That is why we are excited to share two more updates that deepen Pontem’s involvement with SPE—one in the United States and one in the Eastern Mediterranean.

Andrew Yule joins the SPE Fort Worth Section Board as Communications Chair

Andrew Yule has joined the SPE Fort Worth Section board as Chair of Communications and Outreach.

The Fort Worth Section is a highly active technical community, with a clear focus on:

• Strong member engagement through well-attended technical talks and events.

• High-impact technical programming, including the Distinguished Lecturer series that brings leading experts to local SPE sections around the world.

• Community-building events, including a December happy hour that keeps the section connected even outside the standard luncheon format.

• Meaningful student support, with scholarship funding that helps petroleum engineering students pursue their education.

In the communications role, Andrew’s focus will be to help make sure members hear about what is happening, why it matters, and how to get involved—whether that is attending a luncheon, supporting scholarships, volunteering for open board roles, or simply showing up and participating.

Kostas Poursanidis steps into Chairperson role at the SPE Cyprus Section

Kostas Poursanidis has been elected Chairperson of the SPE Cyprus Section, continuing his service in the Section’s leadership.

The Cyprus Section’s goals are centered on building and sustaining a unified petroleum engineering community in the country. In particular, the section exists to:

• Enable technical knowledge sharing within the Exploration and Production sector.

• Support professionals and students through learning, connection, and development opportunities.

• Operate as a non-profit, apolitical, and nonreligious professional community.

• Advance SPE’s broader mission to collect, disseminate, and exchange technical knowledge across the exploration, development, and production of oil and gas resources, for public benefit.

• Provide opportunities for professionals and students to enhance both technical and professional competency.

That combination of technical rigor and community support is exactly the kind of ecosystem that helps good engineers become great ones—and we are proud to see Kostas stepping into a leadership role that will help strengthen that ecosystem.

What this means for Pontem

These two board appointments reflect something we care about deeply: building organizations that keep the industry learning.

SPE is one of the few professional societies that consistently delivers real technical value at every career stage—from early-career engineers looking for mentorship and exposure, to seasoned practitioners who want a strong community and high-quality technical discourse.

We are looking forward to supporting both sections in practical ways—by showing up, contributing to technical discussions, helping promote events and knowledge-sharing, and supporting opportunities for students.

If you are based near Fort Worth or Cyprus (or you just want to get more involved in SPE), we would love to see you at an upcoming event.

Don’t be lulled into a false sense of security with respect to water samples just because water isn’t usually seen as a revenue stream. It doesn’t mean that you don’t need to make sure you have got the right water composition.

But right from the get go, it can be difficult sometimes persuading your drilling team the importance of taking a water sample because any sampling exercise takes time and that can mean a lot of money especially for deepwater wells in harsh environments. The whole sampling exercise could run into a few million $’s for a handful of water samples. Taking that into account be prepared to do a cost benefit analysis showing the drillers why the water samples need to be taken.

The reality of the situation is that the composition of the water can make or break an oil and/or gas development. So, with that in mind you want to make sure the water composition used for design purposes is correct. If you get it right this is what happens:

1. Identification of the most suitable chemical injection locations

2. Selection of the right materials

3. Determination of the most optimum treatment methods

4. Optimum MEG Reclamation Unit (MRU) design and hydrate management strategy

5. Identifying the correct baseline from which operations can begin trending data

If you don’t get the water composition right, there are a number of things that can go wrong. Such as:

1. Inorganic scaling (mainly associated with CaCO₃, BaSO₄ and Halite scale formation)

2. Corrosion

3. Calcium Naphthenates and Carboxylate Soaps

4. MEG reclamation unit fouling

5. Emulsions

Obtaining Clean water samples

Hopefully it is now clear why using the right formation water composition for design and operational purposes is important. Therefore, the next step is to know how to take a downhole water sample with the least amount of drilling fluid contamination in the shortest period of time possible.

This is important as the drilling fluid, especially water-based drilling fluids are likely to have components which will contaminate any formation / aquifer water sample taken. Potential contaminants are NaCl, KCl, CaCl₂, acetic acid, BaSO₄^2-, etc. To minimize drilling fluid contamination there are downhole tools available that identify when the fluid sample to be taken has been cleaned up enough to minimize drilling fluid contamination. One example of this type of tool is the “Quicksilver probe” supplied by SLB. Other service companies have their own version of this type of tool. You just need to ask them about it.

These tools are not always 100% efficient so you can still get a small amount of contamination that needs to be taken into account. To understand how much the sample is contaminated one can ask for a chemical tracer to be added to the drilling fluid, with the aim for it to show up in the water sample. By measuring the chemical tracer, you can accurately determine the amount of drilling fluid contamination there is. If you know the level of contamination and the composition of the drilling fluid, you can then back calculate the contamination out of the analyzed water sample.

If you are to use a chemical tracer, it is important that the tracer has a chemical composition that would not naturally occur in the water sample being taken. One example of a tracer the author has used in the past is sodium thiocyanate (NaSCN).

Once you have gone through the effort of getting a nice clean water sample, the last thing you want to happen is for the sample’s integrity to be compromised during its journey to the laboratory for analysis. In many instances, the samples age and that process can potentially cause a change in composition even for pressurized samples in many instances. Taking that into account when a pressurized downhole water sample is taken it is recommended that this is blown down offshore and preserved for transport. In the case where you want to keep a water sample long term then it is probably better if you keep the sample pressurized.

For me if I have a choice I prefer to analyze the water sample as quickly after it has been taken and preserved. This way you minimize the chance of the sample aging and affecting the results.

With sample preservation being so important Pontem has developed a comprehensive water sample preservation process.

Water Sample Preservation Requirements

When taking water samples from any oil or gas well it is important that they are appropriately preserved offshore to maintain their integrity before they are analysed. This is because without sample preservation, changes in temperature and ageing of the sample may cause the composition of the water to change causing misleading results to be obtained.

To help with the above process we have provided some information on the recommended water preservation techniques to use

Water Preservation Techniques

1) WATER SAMPLE PRESERVATION FOR CATIONS (METALS) ANALYSIS

If possible, use an acid-washed glass bottle for this sample. This can be done on-site, by rinsing the bottle with deionised water with acid added.

Immediately after sampling, add 3 mL of 35% nitric acid (1 part conc. acid to 1 part water) to 1 litre of sample, or 2 ml to 500 mL, or 1 mL to 100 mL.

Check the pH if possible using indicator paper – it should be below pH 2. If not, add another 1 mL of acid.

2) UNPRESERVED WATER SAMPLE FOR OTHER ANALYSES (pH, ANIONS, etc)

Fill the bottle with sample (leave minimal air space). Keep cool (below original sample temperature, if possible). Submit for analysis without delay. A minimum volume of 100 ml is desirable.

To keep water samples cool I have used an Esky to great effect. They can be used for storing water samples as well as beer 😊(probably best not to do it at the same time though).

3) WATER SAMPLE PRESERVATION FOR VOLATILE FATTY ACIDS (VFA) ANALYSIS

Add 1 mL of 6N sodium hydroxide solution to XXX mL water sample (labelled 6N NaOH) – CAUTION: STRONGLY CAUSTIC, WILL BURN SKIN.

Cap the bottle tightly and invert it several times to mix the contents.

Check the pH if possible using indicator paper – it should be above pH 9. If not, add another 1 mL of NaOH and shake again.

Only a small volume of water (10 ml) is needed for VFA analysis. However, if the sample volume is less than 100 ml, write the actual volume of NaOH solution added on the label of the bottle. A correction can then be made to the result.

Once in the hands of the laboratory that is performing the water analysis. It is important to use the correct analytical methods to minimize any analytical errors in the final concentrations measured.

Now for the Science Bit

Once you have acquired the water samples, minimized contamination, preserved the water for transportation, and got them to the lab for analysis. The next thing we need to do is work out what analytical chemistry techniques to use. For Chemistry Nerds like me I would say this is the fun part.

Figure 1: Young Stephen Hamilton…not on Halloween.

Strap-in for an analytical chemistry lesson.

The analysis of the formation / aquifer water samples involves a number of sophisticated analytical techniques that are not available everywhere so it is important you know what is required so you can go and find a lab to do the analysis that is needed. The main companies that do this in the oil and gas industry in alphabetical order are:

1. Corelab

2. Expro

3. Intertek

4. SLB

However, you might not be able to always access this expertise so you may need to look closer to home. If that is the case, then in-country Universities and NOC / Oil and Gas Industry Research labs are good alternative options.

Here are the things that you would want to analyze for to get a workable formation water composition:

• Cations (Metal Ions such as Na⁺, K⁺, Ca²⁺, Mg²⁺, Ba²⁺, Sr²⁺, etc)

• Anions (eg. Cl^-, SO₄^2-, HCO₃^-)

• Volatile Fatty Acids (e.g. formic acid, acetic acid, propanoic acid, butanoic acid, etc)

  • VFA’s are probably one of the most overlooked and under reported analytes in proper brine characterization. Without these values in hand, several risk analysis profiles are left with wider uncertainties, which can impact start-up costs and future OPEX.

• pH – Ideally the pH should be performed onsite as soon as the water sample has been blown down as changes in dissolved CO₂ and exposure to the atmosphere will change the pH over time.

• Salinity / Conductivity

• Density

Cations Analysis Method

Metal ions such as Na⁺, K⁺, Ca²⁺, Mg²⁺, Ba²⁺, Sr²⁺, Fe²⁺ are analyzed with Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP AES).

ICP AES is an analytical chemistry technique used to detect and quantify the concentrations of chemical elements in a wide variety of samples. The method is known for its high sensitivity, accuracy, and ability to analyze multiple elements simultaneously, from trace to major component levels.

How ICP-AES works

The ICP-AES process involves several key steps:

1. Sample preparation: The sample must first be converted into a liquid form. Solids are typically digested using a strong acid or fusion techniques, while liquids may be diluted or used directly.

2. Aerosol generation: The liquid sample is pumped into a nebulizer, which turns it into a fine mist or aerosol with the help of a flowing argon gas.

3. Introduction to plasma: The aerosol is carried into the core of an argon plasma, a superheated (6,000–10,000 K) ionized gas that serves as the excitation source.

4. Excitation: Inside the plasma, the heat evaporates the solvent, breaks molecular bonds, and excites the atoms and ions, raising their electrons to higher energy levels.

5. Emission: As the excited electrons relax and fall back to their ground state, they emit light at specific wavelengths characteristic of each element.

6. Separation and detection: The emitted light is directed into a spectrometer, which uses a diffraction grating to separate the light into its individual wavelengths. Detectors, such as charge-coupled devices (CCDs), then measure the intensity of the light at each specific wavelength.

7. Quantification: By comparing the intensity of the measured light to calibration standards with known concentrations, the instrument calculates the concentration of each element in the sample.

Key features and advantages of ICP AES

• Multi-element analysis: Can measure up to 70 different elements in a single analysis run, which dramatically speeds up testing.

• Wide dynamic range: Can accurately measure elemental concentrations from parts-per-billion (ppb) to high percentages within the same sample, reducing the need for multiple runs.

• High sensitivity: Offers excellent detection limits, especially for metals and semimetals.

• Minimal interferences: The high-temperature plasma minimizes the chemical interferences that affect other techniques like atomic absorption spectroscopy (AAS).

The use of ICP takes care of the metal cations and some of the other elements such as phosphorous (P), Sulfur (S), Arsenic (As), etc. Now we measure pH and conductivity of the water. This can be done using ASTM D1293 and ASTM D1125 respectively.

Anions Analysis Methods

The next in line should be the analysis of the anions such as Cl^-, SO₄^2- and HCO₃^-. The first anion, we will discuss is HCO₃^-. This can be determined by measuring the total alkalinity of the water sample. This is measured using a potentiometric titration by titrating hydrochloric acid into the water sample until the equivalence point is reached.

Total alkalinity is usually reported as HCO₃^- concentration. However, you need to be aware that although most commonly reported as HCO₃^- , it is actually a mix of carbonate, carboxylate and hydroxide species.

Total alkalinity equation is

A_T = [HCO₃^-]+[CO₃^2-]+[Acetate]+[OH^-]-[H⁺]

In most cases the vast majority of the alkalinity is associated with the HCO₃^-ion but where there are high levels of volatile fatty acids you may get excess HCO₃^- causing the formation water to be supersaturated in CaCO₃ at reservoir conditions. If that happens you just need to remove enough HCO₃^- ions from the water so that at worst CaCO₃ is at equilibrium saturation at reservoir temperature and pressure.

Although a tried and tested method, the use of a titrator does provide a greater spread of results than compared to the cation results you get from the ICP. There is an alternative method that is not so commonly used and that is the use of Ion Chromatography (IC) which can be used to differentiate between the various alkaline species thus providing some added confidence in the HCO₃^- results obtained. There are some advantages in knowing the “true” bicarbonate numbers, however the main point of emphasis is to understand what values the laboratory is reporting back so subsequent calculations and modeling will be appropriate and accurate. We have provided some additional information on the use of Ion Chromatography later in this post.

Therefore, it is important to look more carefully at the results and not just assume the number generated by the lab is the correct value. Where possible it is good to get a result from a second analysis technique or lab.

Finally, one little tip I think is worth sharing regarding the HCO₃^- concentration is that if your analysis gives you > 2,000 mg/l of HCO₃^- then there is something probably wrong.

For the other anions (Cl^-and SO₄^2-) that are needing analyzed; it is recommended to use ASTM D4327 or better which uses Ion Chromatography (IC also known as Dionex). Ion chromatography can also be used to analyse for Volatile Fatty Acids but first they must be converted to their carboxylate salts to make sure they do not flash off from the water sample during the sample transportation process. (See Water preservation procedure)

Ion Chromatography (IC) separates ionic substances by exploiting their different charge properties. A sample is introduced into a column containing a stationary phase with fixed charged groups. Ions in the sample then compete with the stationary phase’s counterions for binding sites, retaining them based on the strength of their ionic interactions. A mobile phase (eluent) then washes the ions through the column, with different ions eluting at different times, allowing for their separation and identification.

Here’s a breakdown of the key principles:

1. Charged Stationary Phase: The column is packed with a material that has immobilized charged groups.
   Anion Exchange: The stationary phase contains positively charged groups to attract negatively charged ions (anions).
   Cation Exchange: The stationary phase contains negatively charged groups to attract positively charged ions (cations).

2. Sample Introduction: The sample, containing the ions to be separated, is injected into the column.
   

3. Ion Exchange: Ions in the sample compete with exchangeable counterions for binding sites on the stationary phase.
   Ions with stronger coulombic (ionic) interactions with the stationary phase will bind more tightly.
   Factors such as charge density and ion size influence the strength of these interactions.
   

4. Elution: A mobile phase (eluent), which is a solution of ions, is passed through the column.
   The eluent displaces the ions bound to the stationary phase, causing them to separate. Ions that bind weakly are washed away first, while more strongly bound ions require altered conditions (e.g., higher ionic strength) to elute.
   

5. Detection: As the separated ions exit the column, they are detected, most commonly by conductivity, which measures how well the solution conducts electricity.

If an Ion Chromatogram is not available, the following methods to analyse for SO₄^2- and Cl^- can be used:

SO₄^2- - APHA 4500-SO4 E (colorimetric) or another option is to use ICP-AES to measure total sulfur content in the formation water and then assume that all the sulfur is associated with SO₄^2-. This is not ideal as you can overestimate the level of SO₄^2- ions present if there are other forms of sulfur ions present such as sulfides.

Cl^- - APHA 4500-Cl B (titration). Although a commonly used method, it is probably less accurate than ion chromatography.

Evaluation of the Results

All of the above analytical methods are effective at analyzing the various components in a formation water sample. However, it is the authors opinion that ICP-AES is the most reliable method for measuring the various cations and elemental species in the water and that Ion Chromatography is the most reliable method for analyzing the anions compared to the various titration techniques.

It is also worth noting that the quality of the results can vary from lab to lab so it is important to have a reasonable amount of skepticism when reviewing the results of any analysis that has been performed. When it comes down to it you can send exactly the same water sample to three different labs, and you will probably get three different answers. Trying to fathom what is right in those instances really comes down to experience and understanding what “good” should look like.

Once all the analysis is done you usually get a results table that looks a bit like this. This table has the minimum data you need to allow you to do the analysis such as determining the scaling and corrosion risks associated with the water. In many instances you can get a lot more cations and anions analyzed depending on your requirements. If you are worried about heavy metals such as lead, cesium, vanadium, etc. They can be analyzed but you may have to tell the lab that you would like those included as they don’t always get automatically measured / reported.

Quick QA/QC checks

Once you see the water analysis results from the lab there are a few things you can do to check if the results look “right”. These are:

1. Check what type of geology the water has come from. For example if it is a lacustrine environment, the water salinity may be very low (TDS < 5,000 mg/l) or if it is near a “salt dome” the formation water may be virtually saturated in NaCl (TDS >250,000 mg/l).

2. Are the K⁺ and/or Ca²⁺ concentrations in the water similar or higher than the Na⁺ concentration? If they are, check what brine was used in the drilling fluid. There is a good chance the sample has been contaminated.

3. Unusually high levels of VFA usually may mean some contamination from acetic acid contained in the drilling fluid or well clean-up fluids but worth checking whether other fields in the same area have high levels of VFA because there are formation waters that do have high levels of VFA but they aren’t very common. I talk a little bit more about this later in this post.

4. Does the water sample have a high concentration of Fe²⁺ (>20 mg/l)? If it does it may be caused by corrosion of the sample bottle or well tubing. Also worth checking if the rock, the water has come from contains high levels of iron ore just in case the Fe²⁺ values are correct.

Finally, as well as the above, one of the best things you can do is to make sure that the water composition makes sense from a geological perspective. This involves determining whether there is any scaling components either supersaturated or undersaturated in the formation water that in reality should be saturated in the water at worst. You do this because at reservoir temperatures and pressures experienced, water that has been sitting in the rock for millions of years at pretty much the same conditions should not be supersaturated. That is because any inorganic scale that could have dropped out would have dropped out and left whatever dissolved salt that was left as saturated in the formation water. Also, for CaCO₃ the expectation is that in the CO₂ rich environments we have in the oilfield, the default is to expect any formation water to be saturated in CaCO₃ at reservoir conditions. Sometimes this can be a fairly straight forward process, or it can take some time to get right but it is certainly worth doing. If you want to learn more about this process Pontem have already written an excellent Substack post looking at this exact process.

This process gets easier the more experienced you get because after seeing hundreds, if not thousands of different formation / aquifer water compositions you get to understand what “good” looks like. But it is still not easy as you have such a wide variation of water compositions. These can range from virtually mineral water samples with Total Dissolved Solids (TDS) no more than 1,000 – 2,000 mg/l to formation waters that are nearly saturated in NaCl with TDS contents of up to 300,000 mg/l.

In fact, one case I looked at last year was for a formation water that had >1,000 mg/l of acetic acid in its composition. Even though I had worked on fields that had these levels of acetic acid previously I still thought the acetic acid concentration was so high that there must have been some drilling fluid contamination as the drilling / clean-up fluid did contain acetic acid.

However, I knew I had to get this right because at these levels of acetic acid this could have a significant impact on the inorganic scaling (in a good way) and corrosion risks (in a bad way) associated with producing the formation water under evaluation.

Consequently, I did some more homework and fortunately for me I had formation water compositions for two nearby fields to compare against. Both fields had formation water compositions that had levels of acetic acid that were also > 1,000 mg/l. When you see that you know it isn’t drilling fluid contamination but a real effect as it is just not plausible that all three fields could have the same level of drilling fluid contamination when taking their water samples over a time period of multiple years.

Finally, once you are confident in the water composition you can start to use it to determine whether inorganic scale will deposit in the system. The level of the scaling ions like Ba²⁺, Ca²⁺, SO₄^2- and HCO₃^- are critical components from the water analysis, and with respect to CaCO₃ scaling, the pH is very important so you need to look at the HCO₃^- and VFA concentrations along with the CO₂ content at the various pressures and temperatures in the system.

When assessing the corrosion risks the most important parameters taken from the water sample are; Chlorides, VFA concentration, pH and CO₂. It is also worth noting that it is not the end of the world if you don’t have a reliable pH measurement because as long as you have the CO₂ mol% in the gas phase and the system temperatures and pressures there are plenty software packages that can work out what the system pH is.

Usually for scaling, the more CO₂ and less HCO₃^- in the system, the better it is for CaCO₃ scaling as the pH is more acidic. For corrosion, the opposite is true, as low pH drives higher corrosion.

When it comes to evaluating whether a system has a calcium naphthenate or carboxylate soap problem. Here are a couple of helpful rules of thumb that can be invaluable when assessing a new field’s formation brine for Calcium Naphthenate and Carboxylate soap risks. A few include:

• Calcium Naphthenates and Carboxylate Soaps can become an issue if the pH is greater than 6. Solids and soaps will begin to form in these higher pH scenarios.

• Water compositions with high HCO₃^- and VFA concentrations (>1,000 mg/l, each for example) are indicative of oil with ARN acids that form the Calcium Naphthenates and Carboxylate Soaps. This combination of high levels of HCO₃^- and VFA frequently occurs when you have problems with Calcium Naphthenates and Carboxylate Soaps. It is not infallible but shouldn’t be ignored.

Hopefully this shows how much goes into obtaining a representative water samples and what sort of information you can get from it once it has been analyzed properly. If you want help from Pontem Analytics feel free to contact us.

Subscribe now