from the analysis origins series
The catastrophic loss of the Piper Alpha platform claimed the lives of 167 people. On 6 July 1988, a small explosion caused secondary damage that resulted in a second larger explosion and then a sustained major fire. With a total insured loss of £1.7 billion (US$3.4 billion) it remains one of the costliest manmade disasters in history. Lord Cullen led the subsequent public inquiry into the tragedy, and his report outlined 106 recommendations for changes to safety procedures in the UK sector of the North Sea. In the aftermath of Piper Alpha and the Cullen Inquiry, a new focus was placed on modelling approaches for predicting fire and explosion events, including the use of computational fluid dynamics for mitigating the risk of fire and explosion damage. This feature considers the origins of two of the leading CFD codes for simulating fires and explosions in the offshore industry: KFX/Exsim and FLACS.
The Piper Alpha disaster was one that none of us wishes to repeat, but the lessons learned following the event through validation of numerical tools via experimental programmes have greatly improved our understanding of explosions and mitigation methods, which has helped to make our offshore facilities safer places to live and work. However, we mustn’t be complacent. As the physical size and complexity of some of the new offshore facilities grow, we need to be aware of new associated risks – specifically DDT (deflagration detonation transition), where a subsonic explosion (a deflagration) may accelerate to the point where it can transition to a supersonic detonation. This is what is understood to have happened at Buncefield in 2005, and it is important because the level of damage is much more severe for a detonation.
There remain important challenges for the industry, specifically for the simulation tools in terms of their predictive capability and how they are used in practice. None of the CFD tools can yet robustly simulate detonations, but there are at least some measures relating to local pressure gradient that can be used to check for the onset of DDT. It is important, as an industry, that we remember the lessons of Piper Alpha and continue to develop the simulation tools for the new challenges ahead.
Steve Howell is the Chairman of the NAFEMS Computational Fluid Dynamics Oil & Gas Focus Group and Technical Director at Abercus, a consultancy specialising in advanced engineering simulation in the energy sector.
In 1976 Professor Bjørn F. Magnussen and his first doctoral student, Bjørn H. Hjertager, presented a seminal paper on modeling of turbulent combustion for numerical simulation at the Combustion Institute. This paper introduced the Eddy Dissipation Concept (EDC) and is by far the most-cited paper on fire modeling from the Combustion Institute. The concept turned out to be a very efficient and robust model, which has subsequently been implemented in most commercial CFD codes dealing with turbulent combustion. The paper was the culmination of years of research at the Norwegian University of Science and Technology, NTNU (formerly NTH), in Trondheim, Norway. Since the 1960s, Magnussen had experimented with flames in the university laboratory, trying to understand the process of combustion and soot formation taking place in turbulent flows. After seeking an understanding of the physics, he started work on a mathematical model to calculate and incorporate the effects of turbulence.
Mathematically, the EDC is simple, though the physics are complex. It couples the turbulent flow with the combustion process through the reaction zones in the flow, the fine structures. The basic philosophic concept is that the turbulent eddies in the flow are broken down to fine structure zones where the chemical reactions take place. Magnussen’s original hand sketch of the reactive zones was later confirmed through laser technique in the laboratory, where it was shown that the reaction zones in the real flow appeared in a similar way (see Figure 1).
In the 1970s, Magnussen and Hjertager were early adopters of practical 3D CFD. The two Bjørns continued their work on turbulent combustion modeling until around 1980 when their paths diverged: Magnussen remained at NTNU in Trondheim focusing on fire modeling and the development of the Kameleon CFD code, while Hjertager returned to his hometown of Bergen to focus on explosion modeling and the development of the FLACS CFD code.
A substantial research group built up around Professor Magnussen at NTNU and SINTEF, the biggest independent research foundation in Scandinavia, also located in Trondheim. The group continued its research on fires and fire modeling. Collaboration with industry started in the late 1970s with simulation of flares, and continued with gas dispersion and fire development, including simulation technology for fire mitigation by various water-based systems. The research group provided advanced consultancy, working in close cooperation with oil and gas industry partners to solve specific problems, while also developing simulation tools. This close interaction between academia and industry is perhaps a significant reason for the industrial success of the methods and tools from this group, which has always had a practical approach.
Magnussen’s CFD code was named Kameleon to indicate its adaptability as a general CFD code that could be used for many different applications. Later, having developed into a dedicated simulator for fire analysis, it became Kameleon FireEx, and eventually KFX. ‘I had to decide where to focus,’ Magnussen says. ‘If I could make a contribution where I could save people’s lives, that was what I wanted to do. So I decided to focus on what we could improve on offshore facilities from a safety point of view.’
For industrial safety applications, a major challenge for the CFD approach is to capture both the complex physics of combustion and the important effects of congested complex geometries typically found in process and offshore facilities. In KFX this is achieved through the use of a structured orthogonal mesh, a distributed porosity technique and other sub grid models. This approach makes it possible to simulate complex combustion events very efficiently. The physics can include gas or multiphase leaks, dispersion, liquid spreading, droplet sprays with rainout, evaporation, combustion, soot formation and smoke dispersion and radiation in congested geometries, even with complex surrounding terrain (Figure 2).
Throughout its existence, KFX has been adapted and developed. ‘In the early years, models had to be built manually, a little like LEGO on a computer,’ says Trond Evanger, Managing Director of ComputIT. Simulations were necessarily coarse and simplified because of limited computing resources. Now the program is much more user friendly: it’s easy to import large CAD models of offshore platforms or electronic maps showing the terrain of a larger area to automatically create the CFD model, while modern computing power means the models can be much more refined. Development of the code is still ongoing, and the collapse of the oil price in the last few years has strongly actualized the technology, as optimized design that can be achieved by detailed modeling is a key to production at lower cost.
The code is validated against large-scale experiments in terms of flow, heat transfer and radiation. One of the major test sites is the RISE Fire Laboratory, formerly the Norwegian Fire Laboratory at Sintef in Trondheim, where a huge outdoor test rig built like an offshore module for large-scale fire experiments was built by ComputIT and the fire lab with industry funding. Measuring systems on the rig make it possible to compare realistic fire events with simulations, including also a real-scale deluge system. Such tests are important not only for validation, according to Evanger, but also so engineers who simulate fires can actually experience them firsthand and feel the heat for themselves.
Figure 1: Magnussen’s original hand sketch of the reactive zones (left) and the subsequent confirmation using laser photography (right).
Major industrial accidents onshore and offshore throughout the 1980s, including the loss of the Piper Alpha platform due to the devastating and sustained fire in 1988, focused attention on the development of more accurate simulation methods to better predict and understand the consequences of major toxic and flammable hazards in the process industries, in order to improve the accuracy of risk predictions and design of equipment, processes and safety barriers.
The petroleum industry realized the requirement for technology that could capture the interaction between accidental leaks and the complex geometries of industry plants both onshore and offshore, and that the rapid development of computer capabilities would facilitate this in the foreseeable future. The development of KFX since 1980 has been driven through JIPs with a total industrial funding of about US$20–25 million, in addition to a large number of related PhD theses. The impact of the CFD methodology on safety, design and cost in the petroleum industry has been astounding. In recognition of this, in 1995 Professor Magnussen was awarded the Statoil research price for ‘significant contributions to the Norwegian oil and gas industry’.
Figure 2: Fire simulation visualisation
When lawyers were looking for a technical expert witness on combustion in an insurance trial deciding liability for Piper Alpha, Professor Bjørn Magnussen was an obvious choice. ‘Even at that time I was a little bit famous,’ he says. Professor Brian Spalding of Imperial College London gave evidence on gas dispersion at the same trial.
The basis for Magnussen’s evidence was a series of photographs taken by a bystander on another platform. ‘The sequence of pictures could tell us the evolution of the fire,’ he says. ‘I used a certain technique to look at the pictures, using a magnifying glass in a particular way to restructure it into a 3D view.’ A key question at the trial concerned the size of a fireball: judging by the light emission on the rising structures that he could see in this almost-3D perspective, Magnussen says, it was by no means as big as it had been estimated.
The team at NTNU had previously been working with simulation of fires to calculate how long it might take to cause a rupture in a high-pressure pipe, but Magnussen’s offer to simulate the Piper Alpha incident was declined. ‘My vision has always been that if you have a real accident, you should go in and learn from what really happened,’ he says. ‘You should learn what to do in the future to make a safer structure and a safer operation.’
‘The importance of the Piper Alpha accident was to put more focus on safety for offshore workers and offshore constructions,’ says Magnussen. ‘There had been many early warnings about leakage of gas, which were not properly taken into consideration. Today there is no chance that so many warnings would be ignored.’
ComputIT was established in 1999 to industrialize KFX. Trond Evanger had joined Magnussen’s group at SINTEF in 1982, working in research and as a project manager on various joint industry projects. When Evanger learned of a business development opportunity, he contacted Magnussen and they seized the chance to make the KFX technology more widely available to industry. Since this time, KFX has taken a leading position internationally, especially for dispersion and fire simulations.
KFX now also covers gas explosions and structural integrity to explosion loads – and the fruitful collaboration between the two Bjørns has been reestablished. Bjørn Hjertager had continued with explosion modeling as a professor at Telemark University College, and together with Shell Research he developed the Exsim software, which has been Shell’s preferred explosion tool for more than 20 years. In 2014 ComputIT agreed with Shell and Hjertager to take over the full responsibility of Exsim, and has since then integrated Exsim into KFX as an explosion module.
The Exsim model is based on the Eddy Dissipation Concept, using the same modeling concept as KFX. The philosophy for KFX-Exsim is thus based on only one concept for modeling turbulent combustion covering both fire and explosion. This is important from a philosophic point of view, and provides assurance that industrial solutions are based on a consistent and coherent modeling concept.
A new chapter in the KFX history began in 2017 when DNV GL acquired ComputIT with the ambition to make CFD technology available for a larger part of the industry worldwide. KFX-Exsim is already being used by a large number of companies and universities around the world, but as a part of DNV GL new opportunities arise for CFD development and applications, and for the industry. The company’s industry-leading test facilities at Spadeadam in the UK also represent unique opportunities in this respect.
‘None of the codes can handle detonations at the moment,’ Magnussen points out, but he and Evanger believe the EDC could be well suited to handle detonation simulations. Although this is an avenue they would like to explore, it would require a great deal of funding. Bjørn Magnussen is close to 85 years old but still comes into the office every day and has an eye on the future, ‘because still there are a lot of things that can be improved.’
ComputIT would like to acknowledge Equinor (Statoil), Total, Eni, ConocoPhillips, Gassco, GRTgaz (Engie), and the Research Council of Norway for funding the development through many, many years. Article written by Trond Evanger with support from Fiona Shearer and input from Bjørn Magnussen. Magnussen and Evanger are the co-founders of ComputIT. Shearer is a writer and editor.
Figure 3: Piper Alpha Memorial Garden at Hazlehead Park, Aberdeen, Scotland
Interest in explosion research at Christian Michelsen Institute (CMI) in Bergen started in 1970 following a series of dust explosions at the Stavanger Port Silo, which was at the time Norway’s largest and most modern grain silo. A dust explosion laboratory was established under the leadership of Dr Rolf Eckhoff (now Professor emeritus at the University of Bergen) to investigate prevention and mitigation of dust explosions. In the following years, with the rapid evolution of the oil and gas industry on the Norwegian continental shelf, the scope of research was expanded to include accidental gas explosions.
The first experimental test program for large scale gas explosions was established for Statoil during the period 1978 to 1980, as part of the Sikkerhet På Sokkelen or SPS programme. This was followed by two major gas explosion programmes (GEPs) of research in the 1980’s. The first (GEP 80-86) ran from 1980 to 1986 and investigated pressure development due to flame acceleration by obstacle generated turbulence, the so called Shchelken mechanism, which was identified as primarily responsible for the intensification of explosions occurring in the complex geometries typically found on offshore platforms. This represented a significant investment, 70M NOK, and was sponsored by six major oil and gas operating companies: BP, Elf, Esso/Exxon, Mobil, Norsk Hydro and Statoil. The second (GEP 87-89) ran from 1987 to 1989 and focused on gas dispersion in complex geometries and on explosions in onshore plants. It was supported with an investment of 18M NOK, and was sponsored by BP, Mobil and Statoil.
GEP 80-86 was led by Dr Bjørn Hjertager. In 1980, Hjertager returned to his home city of Bergen, following the recent completion of his PhD studies with Professor Magnussen at the Norwegian Institute of Technology (NTU, now NTNU) in Trondheim, which had led to the development and publication of the Eddy Dissipation Concept (EDC) for turbulent combustion. The experimental data from the programme provided new insight about flame acceleration in complex geometries, and much needed data for the validation of a new CFD code, FLACS (FLame ACceleration Simulator), that was developed by Hjertager for the simulation of explosions as part of the programme.
At the end of the programme, FLACS 86 was released and the first FLACS training course was held (Figure 1). FLACS-86 was a finite volume CFD code that solved the compressible Navier Stokes equations, with a modified k epsilon model for turbulence and the eddy dissipation concept for combustion, where the rate of combustion is determined by the dissipation time of turbulent eddies. The code was based on the structured Cartesian grid approach and used a porosity/distributed resistance (PDR) approach for representing small obstacles typically found within the complex geometries on offshore platforms.
The programme identified the importance of capturing small scale congestion within the CFD model – indeed current guidance from Gexcon is that all geometrical details down to a size of one inch play an important role in the development of an explosion and should be captured within the CFD model. To capture such a refined level of detail explicitly within a body fitted mesh was at that time (and remains) prohibitive, so they were instead represented in terms of equivalent porosity, resistance and turbulence generation so that their effect is captured using a relatively coarse computational mesh with a fixed edge length of 1m.
The PDR approach has become widely recognized as a pragmatic approach for simulating explosions using CFD – it continues to be the basis of the FLACS code and it has been adopted by other CFD tools dedicated to the simulation of explosions including Exsim and AutoReaGas. Recent efforts by Shell Global Solutions to develop an explosion code using OpenFOAM has also used this approach, using PDRFoam as the underlying solver.
After the accident on the Piper Alpha platform in 1988, a wave of public outcry ensued that demanded answers. The report for the public inquiry, known as the Cullen Report, was published in 1990 and presented 106 recommendations for improving offshore safety. This led to the Offshore Installations (Safety Case) Regulations in the UK in 1992 which required the preparation of a safety case for all offshore installations. As a consequence of the event and the subsequent inquiry and legislation, explosion safety and explosion modelling received a renewed focus. Piper Alpha caused a re-evaluation of the level of our understanding of fire and explosion hazards and resulted in significant investment in and development of analysis software to understand why and how such situations could occur and how they could be minimised or prevented in future.
Figure 1: Physical testing (top), modelling (middle) and training (bottom) at the first FLACS course in 1986, around a standard congested region test configuration.
“Piper Alpha is always there, always at the back of your mind. It led to the creation of offshore regulation and standards that have been accepted around the world. However, the challenges that face us today are different – how to deliver improvements through innovation and the continuing quest to strike the right balance between cost and safety against a background of ageing assets and falling oil prices.” These are the words of Kees van Wingerden, CTO of Gexcon.
In the UK, large scale experiments were undertaken by British Gas at its Spadeadam test site in the north of England. The BFETS (Blast and Fire Engineering for Topsides Structures) joint industry project identified that the overpressures at full scale could be significantly higher than had previously been thought possible. It was realized that explosion phenomena are too complex for simple models and that advanced models based upon CFD would be required in future.
In Bergen, under the remit of the newly formed Christian Michelsen Research (CMR) and its subsidiary Gexcon (Global Explosion Consultants, originally Gas Explosion Consultants), three new gas safety programmes (GSPs) were completed through the 1990s, together with many joint industry projects funded by the EU and the Norwegian Research Council (NFR). The impact of Piper Alpha is clearly indicated by the sharp rise in the number of sponsors for the GSPs – there had been 3 sponsors for the second GEP starting before Piper Alpha in 1987, which increased to 13 for the first GSP starting in the aftermath of Piper Alpha in 1990.
Over the years, the numerous research projects have provided an increasing source of experimental data from CMR/Gexcon’s own dedicated test facility on the island of Sotra, near Bergen, which has improved the understanding of the physics of explosions. It has also led to the continual development of FLACS. In 1993 the requirement for a fixed mesh edge length of 1m had been overcome, which allowed the user more control over the meshing resolution. In 1996, FLACS became commercially available to enable other parties to undertake explosion simulations in line with the new safety case approach.
By the end of the 1990s, as a consequence of the continuing research efforts, an alternative approach for combustion modelling known as the β-flame model had gained favour over the eddy dissipation model and was subsequently implemented into FLACS. The β-flame model was proposed by Bjorn Arntzen as part of his doctoral studies at the University of Bergen which concluded in 1998, and is based upon correlations which link the local turbulent burning velocity to the local turbulence field.
Figure 2: Simulation of a gas explosion on an offshore platform
Over the past two decades, the offshore industry has progressively moved towards this probabilistic approach1. In the late 1990s NORSOK released the Z-013 standard for risk and emergency preparedness assessment. The standard described a new probabilistic methodology for a risk based design approach, which requires the simulation of a large dataset of possible events to be simulated at each stage of the sequence leading up to an explosion: background ventilation pattern under the influence of wind or mechanical HVAC during normal operations; dispersion of fuel and the accumulation of a flammable cloud following a loss of containment; and the explosion dynamics following a delayed ignition of accumulated hydrocarbons. If the probability of occurrence of each simulated event can be estimated, exceedance curves for the explosion load can be constructed, from which the design explosion load corresponding to an allowable level of risk can be determined. FLACS-RISK is essentially a simulation data management tool for handling the vast number of individual simulations required for a probabilistic explosion assessment in line with the NORSOK Z-013 standard, and has functionality to automatically create the exceedance curves that are the required output from the analysis. The development of RISK is recognition that the CFD solver is only one part (albeit an important part) of the analysis jigsaw that can simulate individual deterministic events, but it is also crucially important to understand how to use the code and compile the CFD predictions within a probabilistic framework.
Although FLACS was first developed as an explosion CFD code (Figure 2), it has subsequently evolved into a suite of solvers using common pre/post processing tools. A ventilation/dispersion solver allows FLACS to be used seamlessly for the three stages of the probabilistic assessment: ventilation, dispersion and explosion, and additional tools for calculating the release sources terms are also implemented within FLACS to provide a complete software solution for this application
Since 2000, numerous joint-industry projects have provided additional experimental data for validation of FLACS and continuing experimental work relating to dust explosions over the past four decades has culminated in a dust explosion solver. Very recently, in January 2018, Gexcon announced a new partnership with Shell to exclusively provide FRED, Shepherd and PIPA. FRED (Fire, Release, Explosion and Dispersion) is a consequence modelling tool that has been developed by Shell since the 1980s. With this new partnership, and other ongoing developments, Gexcon and FLACS continue to develop and help to make our offshore plants a safer place to live and work in future.
Gexcon would like to acknowledge Equinor (Statoil), BP, ELF, Esso, Mobil, Total, the Christian Michelson Research (CMR) Institute and the University of Bergen, Norway for their support, funding and contribution to FLACS development over the years. Article written by Mark Keating with support and input from Trygve Skjold and Djurre Siccama. Keating is the European Regional Sales Manager at Gexcon UK while Skjold is the FLACS R&D Director & Siccama the FLACS Technical Product Manager at Gexcon AS, Norway. The FLACS suite includes FLACS, FLACS-Fire and FLACS-RISK
Computational fluid dynamics is becoming increasingly used in the oil and gas sector as the benefits of the approach in terms of improved insight and better understanding of flow phenomena are realized. However, there is currently little detailed practical guidance for how to use CFD within the sector, so each practitioner has had to develop their own analysis methodology over the years. For many CFD applications that have become routine, whilst there may be broad agreement on how to undertake the analysis at the high level, the devil is very much in the detail. As a consequence, there are inevitably inconsistencies that already exist between the methodologies developed by different practitioners, and left unchecked this situation can only deteriorate as new entrants begin to use CFD too. In early 2016, several members of the NAFEMS CFD working group with a focus in the oil and gas sector, collectively recognized this risk and formed a new subgroup – the oil and gas focus group. The purpose of the group is to review industry guidance to identify where practical guidance already exists, and to develop new practical guidance where it is needed. The group is currently working across three themes: technical safety, flow assurance and subsea/hydrodynamics, and already has a number of publications in development across each of these themes. “By engaging as a group, we can move towards a consensus on the practical details of the analysis for each of the applications we’re working on,” says Steve Howell, current chair of the focus group. “And if we’re unable to reach consensus, then we’re able to identify some open questions, which in itself is progress – by clearly identifying what the open questions are, at least we’re in a position to do something about it in the future.” If you would like to get involved or contact the NAFEMS CFD Oil and Gas Focus Group please use the contact form.
Note from Technical Editor - For more information on Probabilistic Approach to Explosion Assessment in the Oil and Gas Sector see Steve Howells article in the April 2016 issue of benchmark