History of PIPESIM, part 2: PIPESIM's legacy in the Schlumberger Era - by Mack Shippen

Mack is a Flow Assurance Consutant with Petroflow Inc. and serves as a Visiting Scholar at the University?of Tulsa.? Previously, Mack was with Schlumberger (SLB) for over 23 years, where he played a pivotal role in the development of PIPESIM, managing the product for over a decade and driving significant advancements in core science, user experience, and integration of GIS and Python extensibility.

This tribute is released as part of the Olga and Pipesim 40th anniversary celebration.

See also: PIPESIM: A story from the early years - by Colin Watters

I hired on to Baker Jardine fresh out of graduate school at Texas A&M in January 2001. Baker Jardine was headquartered in London and I joined the Houston office to look after the growing market in the US. Bryant Mueller ran US operations, and we were all in our twenties – young yet motivated and thankfully Alan trusted us to get the job done!

As a small company, my role focused on technical support and consulting but stretched into marketing and sales. We didn’t have anyone dedicated to manage the IT system so Bryant placed me in charge of that as well!

Assimilation

During my first week on the job, I learned that Baker Jardine was being acquired by Schlumberger. In the years that followed, in addition to adding customers, Schlumberger itself rapidly adopted PIPESIM for many internal purposes ranging from artificial lift, completion design, production log interpretation to well and network performance studies.

At the time of the acquisition in April 2001 Baker Jardine had grown to nearly 50 people with the majority concentrated at the London office and about six of us in Houston and two other small offices in Caracas (Venezuela) and Villahermosa (Mexico). A majority of BJA staff are in the photos that follow.

Move to Abingdon

In 2003, about two years after the acquisition, the PIPESIM development team was moved to Abingdon, a small town near Oxford England and home to the ECLIPSE Reservoir Simulator. Naturally there was keen interest towards integrating PIPESIM and ECLIPSE. Baker Jardine had developed a coupling, called “Field Planning Tool”, however with the development teams under the same roof, there was potential to take this much further. As a first step, the PIPESIM engine was componentized, such that calculation routines were made more modular and efficient. Next, the PIPESIM engine was ported to Linux to enable coupled runs to be executed on high-performance Linux clusters. While originally developed on Unix, PIPESIM had mostly been used in DOS/Windows environments since the late 1980’s. The integration environment was commercialized as IAM (Integrated Asset Modeler) in 2005 and has been used extensively in the years since.

Advanced Multiphase Flow Models

In the early days of PIPESIM, there were just a few multiphase flow models including one developed by Alan Baker for condensate flowlines and published in 1988. Over the years, customers have requested various additional models to adapt for a wide range of flow conditions and incorporate the latest science. After forty years, there is now a plethora of models available in PIPESIM and the science has advanced significantly.

OLGA-S

PIPESIM first incorporated the OLGA steady-state mechanistic model in 1994 through an arrangement between Dick Shea and Alan Baker. There is a long and colorful history of the development of this flow model, the story of which is told in the history of OLGA section. Today, OLGA is considered the gold standard multiphase flow model having been extensively validated and refined with both large-scale experimental facilities and many field studies. With the incorporation of OLGAS in PIPESIM, many operators and E&C companies standardized on PIPESIM and OLGA for detailed steady-state and dynamic analysis of production systems respectively with the confidence of common underlying physics.

University of Tulsa Collaboration

In 1973, Professor Jim Brill started TUFFP (Tulsa University Fluid Flow Projects) to experimentally investigate and model multiphase flow in pipes and the research consortia is still going strong having celebrated its 50th anniversary last year. Baker Jardine joined in 1993 and incorporated their latest models into PIPESIM including the Xiao and Ansari mechanistic models. Schlumberger joined the consortia in 2002, shortly after Cem Sarica became Director of the project upon Jim Brills’ retirement.

It was this year that Tulsa unveiled the TUFFP Unified Mechanistic Model, a culmination of years of research and applicable for all pipe angles and fluids. This model became integrated into PIPESIM starting in 2007. For over twenty years now, SLB has advised on the code structure, provided test data and feedback on its development. Additionally, the experimental data generated at Tulsa is used to improve both OLGA-S and the TUFFP Unified model alike. Today, the latest update (2024) of the TUFFP Unified model sits alongside OLGA-S in PIPESIM, adding confidence to critical designs. Likewise, PIPESIM serves as an important tool used by TUFFP to validate and compare field-scale flow models.

Technical Support Adventures

Working on the front lines for a few years exposed me to the many nuances of PIPESIM. I remember one day I received a call from a customer (circa 2002) who had been furiously running many cases and was in a panic when he discovered the “charge units” listed in the PIPESIM output file followed by a large number:

I rang up Colin Watters who explained to me that this was a remnant from a time long ago when we did actually charge for runs (which were originally quoted in British Pounds but changed to “units” as BJA expanded internationally) and promptly removed this message from the engine output for the next release.

Another time, I had to explain to a customer why he was getting an engine error message that stated “Adrian said this could never happen” which I can only imagine was the outcome of an unresolved argument between Colin and Adrian.

One day about twenty years ago, I received a call from another customer in a panic. They were operating an offshore gas field with a very long subsea tieback. Methanol was injected through a dedicated injection line that ran many miles in parallel to the tieback to transport methanol from the platform to the subsea wellhead manifold. The problem was that the injection pressure on the methanol injection pump rapidly dropped from a couple thousand pounds of pressure to roughly four hundred, indicating that the injection line experienced a serious leak. The entire field had to be shut in or else hydrates would surely form and block the tieback. Production losses were on the order of six figures USD/day, so the situation was urgent as they could not locate the leak to repair it. I quickly built a PIPESIM model for the methanol injection line given the known trajectory, inlet pressure and flowrate and sensitized on the length of the pipe that would result in the seawater gradient outlet pressure. The client then directed the ROV to inspect that location and located the leak a short distance from where PIPESIM predicted it would be. I was thanked profusely for this pro-bono work, though the client was understandably reluctant to promote the experience as a case study!

Evolution of the User Experience

For the first nine years, building a PIPESIM model meant using a text editor to type out a keyword input file.

PIPESIM Keyword input file (1984-present), 1987 User Guide

Back in the early 1990’s though, Windows was gaining widespread popularity and in 1993, the first graphical user interface was released for PIPESIM single branch modelling. It was 16-bit and followed a year later by a separate user interface to create networks:

First Generation, 16-bit Graphical User Interfaces for PIPESIM and PIPESIM-NET (1993-2000)

The desire for a more integrated environment for wells and networks coupled with the need for integration with external programs led to the development of the second-generation user interface, named “PIPESIM 2000” though it was actually released in late 2001. The 32-bit COM-based framework could be controlled by the Openlink API which enabled integration with IAM, Process simulators and end-user scripting.

Second Generation, 32-bit User Interface (2001-2012)

In the first few years after the Schlumberger acquisition, there was quite a bit of interest in expanding on the well modeling capabilities, especially considering the breath of applications in well modeling within Schlumberger itself, particularly the completions and artificial lift segments. Inflow Performance models were greatly improved to account for a variety of scenarios such as frac pack and tight gas fractures. Gas lift and ESP design procedures were dramatically enhanced with close collaboration with the Artificial Lift group.

Around 2005 a team in Abingdon developed an interactive well schematic to make building and viewing well models much more intuitive. I joined the product development team in 2007. From my own experience on projects, I found that building network models could be quite laborious. I recall one particular project for a gas storage field in southern California very close to where the film ET was shot. Driving to the facility I could see pipelines clinging to mountain sides with very steep and chaotic trajectories. To model these effects, scripts were created to transform pipeline trajectories from CAD drawings to PIPESIM and a consultant was hired specifically for this purpose. At the time, GIS was becoming fashionable, and this experience served as the inspiration to natively integrate GIS into PIPESIM.

So, over the next few years, we set out yet again to rebuild the user interface, moving the development team to Houston in 2010 for this task. With the 2013 release, PIPESIM redefined the user experience (now 64-bit) offering a GIS network canvas, an interactive wellbore schematic and many other customer-driven usability improvements.

Third generation, 64-bit User Interface (2013-present)        

One such enhancement was a tubing and flowline catalog. I recall one time several years before giving a customer a "PIPESIM" ruler as a gift that contained a handy reference for tubing and pipeline tables. He thanked me but then asked why we just didn’t put the catalog in the user interface itself. This was such an obvious solution that I promised if given the opportunity I would do it and so we did.

We also added a new GUI validation layer. The engine itself has data validation but it is quite forgiving, and I had seen too many cases where users ran into problems based on simple errors with input data. The task of defining upper and lower limits for validation fell to me and I had a lot of fun with it, leaving a few easter eggs behind for users to discover. Just try defining a land flowline with an elevation higher than Mount Everest, a subsea flowline with a depth lower than the Mariana trench, or an ambient wind speed for anything stronger than a category 1 hurricane! Still, one day a user complained because a model he imported from an earlier version of PIPESIM would not run because it flagged an invalid ambient temperature that was higher than the upper limit I had defined (134°F) which was the world record!

The engine made significant advancements during this time as well. One notable achievement was the parallelization of the network solver in 2012. This effort fell to Rod Lessard, Carlos Boneti and Jun Li. Most laptops computers these days have 4 to 8 cores which correspond to 8-16 processing threads. When a network is solved today, each branch is solved independently and in parallel and the results are assembled together to progress through the iterations. This effort resulted in up to an 8-fold speedup (depending on the case), greatly enhancing the performance of intensive tasks such as network optimization and reservoir coupling. Network optimization capabilities first introduced with GOAL in 1991 evolved exponentially and leverage advanced optimization algorithms developed by the SLB Cambridge Research Center in Boston.

To further enable automation and leverage advancements in the rapidly evolving field of data science, the PIPESIM Python Toolkit was introduced in 2017, empowering engineers to innovate at a level not possible before.

Despite these advancements, the user interface today still generates a keyword text file which is processed by the engine (.pst for single branch wells and pipelines and .tnt for networks). The separation of the PIPESIM engine and user interface is quite convenient for evolving each independently and (as with ECLIPSE and OLGA) is a key advantage of a calculation engine born in the pre-windows era of the 1980’s. Indeed, for some software products developed natively in Windows, untangling the GUI and calculation engines has proven a daunting and sometimes futile task.

For certain advanced applications even today, the GUI is bypassed entirely, and the engine keyword file is modified (often with clever Python scripting) and run in batch mode. Such is the case with “micro-PIPESIM” where the PIPESIM engine is deployed as a lightweight Linux executable on edge devices for individual well simulation and control. At the opposite extreme, PIPESIM is run remotely on high-performance Linux clusters in the cloud for large-scale, automated well and network digital twin implementations focused on real-time optimization. This latter scenario is eerily reminiscent of the mid-1980’s when PIPESIM was run remotely on a mainframe!

Industry Recognition

As PIPESIM celebrated its 30th anniversary in 2014, its technical achievements combined with many successful field applications led to special recognition by industry. These include “Best Production Technology Award” from World Oil and Hart’s E&P Meritorious Award for Engineering Innovation.

SPT Group Acquisition

In 2012, Schlumberger acquired SPT Group, primarily to add the advanced OLGA dynamic multiphase flow simulator to the software portfolio. However, two years prior, SPT group had acquired Neotec, which was a direct competitor to PIPESIM with a product suite including PIPEFLO, WELLFLO and FORGAS. Neotechnology Consultants Ltd. (Neotec) was formed in Calgary in 1972 by Dr. Garry Gregory and Dr. Khalid Aziz, professors at the University of Calgary who specialized in multiphase flow research. PIPEFLO was first released in 1974 as a simulator focused on modeling multiphase flow in pipelines and was primarily used in onshore field development projects in Western Canada, often involving multiphase transport across long distances. Dr. Gregory, a pioneer in the field of multiphase flow research, focused on advancing the methodology by which more accurate multiphase system designs could be achieved, and is widely recognized for his many contributions in this field of study.

After the acquisition of SPT Group, the PIPEFLO and PIPESIM development teams were gradually merged together. This afforded an illuminating opportunity to discover nuances and reveal insights that were previously cloaked in proprietary code that had evolved independently over many years. Some of the key insights discovered by Neotec were ported into PIPESIM code base including the Gregory multiphase flow model for wells and improvements to pigging calculations. Bryan Leppard in particular, lead engine developer at Neotec since 1989, has contributed many ideas and has served as lead engine developer since Colin retired in 2016. Additionally, PIPESIM and OLGA were more technically aligned, particularly so for heat transfer calculations and fluid modeling.

Move to Oslo

In 2018, with a goal to consolidate the flow assurance simulators to one technology center, PIPESIM was moved once again, this time from Houston to Oslo. A wealth of expertise is now concentrated in the “Flow Assurance Center of Excellence” with cross-pollination of ideas and people among the PIPESIM and OLGA development teams.

Quality assurance has progressed a bit since the fast and loose Baker Jardine days. Each night over 1200 GUI regression tests and 640 engine regression tests are automated in parallel on the cloud representing hundreds of hours of runtime. This is on top of unit tests that cover the majority of the code. PIPESIM has achieved notoriety as the most thoroughly test-automated application in SLB!

A key theme over the past couple of years has been advanced well modeling involving intelligent completions and complex flowpaths in downhole networks. There are many other exciting developments in the works (no spoilers here!), driven by the next generation of engineers both within SLB and among the many loyal customers.