Use of CFD in the Performance-Based Design for Fire Safety in the Oil and Gas Sector Camille Azzi* and Lars Rogstadkjenet camille.azzi@gexcon.com GexCon AS Bergen, Norway ABSTRACT This paper addresses the current situation in the performance-based approach used to safeguard against fire in the oil and gas industry with focus on the modeling part. The characteristics of different fire types encountered in the petroleum industry are discussed. The paper reviews methods currently used to assess fire impact and discusses their advantages and disadvantages however the core of the article is devoted to discuss the use of Computation Fluid Dynamics (CFD). Predictions expected from simplified models and those of CFD simulations are also compared and discussed. Finally, the paper presents two industrial applications where the benefits of using CFD as a modeling tool in the performance-based design are illustrated. INTRODUCTION Fires and explosions are the most serious hazards in the petroleum industry and other industries storing or using hydrocarbons. Referring to accident statistics [1], there is no doubt that fire poses serious hazards on oil and gas installations, the personnel operating them and the environment. Safety has been traditionally handled by complying with the regulations settled by standards and legal requirements according to each country. Rules are commonly based on what is perceived as best design and may be motivated by historical habits updated or modified in the wake of large accidents where lessons are learned. Reactive and event-driven safety approach showed clear limitations and thus the industry has been moving towards proactive and performance-based design. The drive toward performance based design safety has pushed development of advanced modelling tools. Computational Fluid Dynamics (CFD) models are the most advanced tools available to predict the consequences of a fire event. These modelling tools provide the means to understand complex fire events and plan for appropriate mitigating measures. The accurate prediction of heat loading provided by these tools is increasingly becoming the basis for fire resistant design in the petroleum industry. The capability of these tools means that designers and safety responsibles can test the performance of certain designs through simulations rather than experience and knowledge gained from real accidents. Such hazardous events can be simulated with high reliability and the safety performance of a design can be evaluated before a fire event arising in the facility.
Field-based codes, as called in the fire protection society, have been extensively used in the residential fire safety to predict smoke propagation in complex residential areas such as hotels, hospitals, high-rise buildings and passenger ships etc. The petroleum industry has traditionally been using CFD software for numerous applications and hence CFD has more easily gained acceptance. Residential areas normally comprise a wide variety of fuel materials which characteristics are not always known. The relative simplicity of fuel characterization in oil and gas installations compared to the phenomenon of solid fuel combustion means that model prediction has been considered more reliable and their use have become widespread. The use of CFD modelling in the petroleum industry covers a wide range of applications. The impact of fire effluents on personnel is quantified and used to assess escape ways and temporary refuge safety. The integrity of primary structure, especially offshore, is of paramount importance and fire simulations are extensively used to evaluate the survivability of load bearing structures. CFD is also used extensively to assess escalation to hydrocarbon containing equipment and critical safety functions. In these situations, modelling is effectively used to determine the required level of passive fire protection where excessive application of PFP introduce unnecessary costs in construction, operation and maintenance. The subsequent sections of the article provide an overview on the events leading to and characteristics of fire types in the oil and gas industry, discuss the common methods used for fire impact evaluation and provide a brief comparison highlighting main discrepancies between predictions. The last part of the article provides industrial applications using CFD simulation in fire protection and lists the conclusions drawn. CHAIN OF EVENTS LEADING TO FIRE SCENARIOS AND THEIR SPECIFICS IN THE PETROLEUM INDUSTRY Loss of containment of hydrocarbon is in general the main initiating event that leads to explosion and fire. Discussion of the root causes leading to loss of containment are outside the scope of this article. The chain of possible events following hydrocarbon release are illustrated in Figure 1 below. In summary, a release that is not ignited at all will generate an explosive / toxic atmosphere. A release that is immediately ignited will result into fire while a delayed ignition will lead to an explosion / flash fire followed by fire provided fuel release is sustained. The type of fire is rather depended on the state of hydrocarbon storage and the way it is released. The different fire types can be summarized as follows. Fire types in the petroleum sector are largely dominated by how the hydrocarbon is stored and, upon release, flows and spreads in the area. Liquid pool fires normally occur as a result of a liquid spill ignition. The pool shape will depend on the ground where the spill occurs. A spill collected in a bunded area will normally result in static size of pool while a continuous spill on an open floor will result in a running pool fire. A pool fire is normally fueled by the evaporation of its liquid governed by the heat transfer from the flame to the pool. The aspects that tend to have high impact on the potential hazards of the pool fire are; size and shape of pool, size of flame, combustion properties affecting burning rate and heat and smoke production.
Liquid spray fires normally follow an ignition of pressurized liquid. The high momentum release tends to break the liquid into fine particles (mist) that are easy to evaporate and burn in form of jet fires. When liquid spray and flashed gas are both present, the fire is called two-phase jet fire. Rainout of large droplets and formation of pool fire can accompany a spray fire as well. Gas jet fires follow ignition of pressurized gas leak or release of completely flashing liquids. The severity of a jet fire depends strongly on the jet release rate and momentum of the jet. The highspeed flow in jet fires result in elevated convective heat when impinging on a target. These fires can be considered to have the most severe effects on passive fire protection due to combination of heat fluxes and erosive forces. A gas jet which momentum is dissipated under water for example will burn as a diffuse gas fire on the sea surface. Flash fires are the result of a delayed ignition of flammable cloud. The term refers to a situation where combustion is too slow to result in pressure build up. Flash fires are short lived. A release of gas or flashing/evaporating liquid that is not immediately ignited normally results in the formation of flammable vapor cloud. The flammability and reactivity of the cloud strongly depends on the concentration of fuel-air mixture. Should the formed vapor cloud be in a confined and/or congested area, the flash fire could become an explosion resulting in significant pressure build-up. Flash fires or explosions are also followed by jet fires provided the fuel release is sustained. No ignition Gas dispersion Toxic / Explosive atmosphere Loss of containment / Release Immediate ignition Delayed Ignition Fire Explosion Explosion followed by Fire Fire impact on personnel material and environment Explosion impact on personnel material and environment Explosion / Fire impact on personnel material and environment Figure 1 Potential events following hydrocarbon release and leading to fire
THE VARIOUS METHODS TO EVALUATE FIRE IMPACT In the performance-based approach, there are numerous methods used in estimating fire impact on critical elements. The common methods range from recommended values based on experiments to simplified and advanced modelling. Standard guidance Experimental values in tabulated form are provided for use in references such as in FABIG technical notes [2] and NORSOK standards [3]. The tabulated guidance values are available for various jet and pool fires. The values are reported for different fuel material and sizes. Heat load, smoke concentration, flame length and temperature are reported for different pool fire sizes. Similarly, the heat loads, smoke concentration and jet flame length and temperature are reported for a range of release rates. Effects of confinement and deluge on some of the scenarios are also described. These values are mostly useful to estimate heat loads on engulfed objects and crude evaluation of the smoke concentration. Simplified models The simplified modelling is normally conducted with empirical correlations based on experimental measurements and physical principals Simplified modelling is widely used due to its simplicity and fast computation time. There are various correlations for pool and jet fires to estimate the flame length, temperature, smoke concentration and heat loads. The far-field heat load is commonly predicted using these correlations. Some of the correlations are applicable within one flame length but no or little information is provided about targets inside the flame. There are various in-house and commercial tools which provides the set of correlations in a computerized from with graphical user interface. CFD models CFD tools have been used in the oil and gas sector for various safety applications such as dispersion, fire and explosion. These models are used for a wide range of jet and pool fire applications to predict heat loads and temperatures inside the flame, in the near and far field. Impairment of escape routes and temporary refuge are also assessed by predicting thermal, toxicity and visibility thresholds using CFD tools. There are several CFD software used for fire simulations. Some are general purpose CFD software which can also be set up for fire simulations and others are dedicated software for fire calculations. Conducting a review of the available tools and listing their capabilities is outside the scope of this work. The paper rather focuses on a CFD software that is dedicated for dispersion, explosion and fire simulations. FLACS [6] is developed, maintained and used in GEXCON. The code includes various models to solve flow, combustion, turbulence and heat transfer equations. FLACS software uses two different combustion models, one for the premixed mechanism used in explosion and flash fires, and another non-premixed combustion model used for fire simulations. The soot formation and radiative gas properties are computed and used in the radiation transfer calculations. Fuel properties and their influences on the fire are taken into account in the simulations. Most fuels
with known material properties can be considered and their characteristics are reflected in the combustion and effluent production. FLACS predicts the fire consequences in a three-dimensional environment where all objects in the domain are taken into account. The induced turbulence due to interaction of flow and objects is computed by solving the transport equations for turbulent kinetic energy and its dissipation. Both heat convection and radiation transfer are computed accounting for the scattering and reflection induced by the presence of obstacles in the simulations domain. Most of the fire applications are implemented in full scale and thus entail large simulation domains. This normally requires numerous finite volumes (cells) to discretize and resolve the flow in a domain. FLACS uses a porosity concept that allows modeling of small objects as partial obstacles in the cell volume [6]. Therefore, the cell volume size does not need be very small to resolve all the small objects which helps keeping the number of finite volumes in a large simulation domain reasonable. The reasonable number of finite volumes make conducting a large number of full scale fire simulations feasible for the purpose of fire consequence and risk analyses. Comparing CFD to simplified methods The advantages and disadvantages of using CFD as opposed to standard guidelines and simplified modeling for fire safety design in the oil and gas industry are highlighted in this section. Firstly, conservatism is a common approach in simplified modelling and the use of tabulated guidelines in order to compensate for the uncertainties residing in these practices. Conservatism is not always a bonus for the design as it induces drawbacks such as extra unnecessary costs, volume and weight. A common application where conservatism is a bad design practice, is the implementation of passive fire protection (PFP). PFP application is very useful in protecting against fire but its excessive use would imply extra costs for the implementation and maintenance of PFP. That is in addition to the adverse effects of added congestion which increases the explosion severity. The use of CFD simulations in this case can provide reasonable estimation of the impact of potential fires on the element that is subject to PFP implementation. Assessment of the impact predicted by CFD and response of the element allows to apply the amount of PFP required for the structure or equipment to withstand the potential fires in the area for the needed duration. Therefore, using CFD in this application provides means to apply only the required PFP coverage instead of excessive. Secondly, a known issue in the use of standard guidelines and simplified modelling is the range of applicability of these tabulated values and correlations. The values are normally documented and the correlations are derived based on a set of experiments with defined conditions. The use of these values and correlations outside this range of conditions can nullify their validity. Based on experimental data, the empirical correlations are normally able to provide the flame characteristics (length, size, temperature) for both jet and pool fires within a reasonable accuracy. However, most of the data assume no interaction of the flame with obstacles. Therefore, these type of obstructed fires are only very crudely represented by empirical correlations. On the other hand, CFD simulations can provide a good representation of the interaction between fire and obstacles and the resulting flame shape. The flame shape of a 2 kg/s release of methane is predicted via
empirical correlations [ref. 5] and shown in Figure 2. That is a reasonable prediction in calm wind and in absence of obstacles. However, in a compartment, a horizontal jet fire will eventually impinge on the wall and change shape. It can be acceptable in this case to assume that the impact of jet fire on the wall has a diameter similar to that of the jet at the point it contacts the wall. Figure 2 shows that the jet diameter at 10 meters from the leak is about 1.9 meters. Now considering the same scenario assessed with CFD, a jet fire impinging on a wall (10 meters away) will be similar to the left side pictures in Figure 3. Moreover, in case there are obstacles such as small equipment or piping in the way of jet, the fire shape will be drastically affected. Fire in this case will be similar to the right picture in Figure 3. The jet loses of its momentum at the pipe rack and then the buoyancy-driven flame impinges on the ceiling instead of wall. This simple scenario shows a great discrepancy between the prediction obtained from using different models and adopting different assumptions. Figure 2 Flame shape prediction by empirical correlations [5]
Figure 3 Two (top) and three-dimensional (bottom) temperature contours of a 2 kg/s methane jet fire impinging on a bulkhead (left) impinging on a pipe rack in front of the wall (right)
Response to fire impact Whether it is tabulated values, empirical correlations or CFD, the main reason of predicting fire impact is to anticipate the response of main critical elements such as major structures, walls, decks and hydrocarbon containing equipment. Numerous structural response codes are used to predict the deformation of failure of critical structures. Similarly, there are other tools that computes the heat transferred to pipes/vessels containing fluid and account for fluid flow and pressure relief. The possible failure of structure of equipment is predicted in order to assess potential escalation and evaluate the need for application of passive fire protection (PFP). CFD outcomes are normally post-processed in the form required by the response software and fed into the software to compute structural response. Direct coupling of the CFD tool with response software has also been implemented, in the oil and gas industry, where the transfer of data between the CFD and structural tools has been automated. However, this topic is not further elaborated in this paper as the aim is to shed light on the use of CFD for fire impact calculations in the petroleum industry. INDUSTRIAL APPLICATIONS Some practical applications are shown in this paper. The geometric models are slightly modified and anonymized for confidentiality reasons. The legends are also hidden for the same reasons mentioned. Facility siting and emergency preparedness As part of various authority requirements as facility siting has to be conducted for hydrocarbon storing plants in order to ensure that accidental release will not go beyond the plant properties and reach unwanted places such as residential areas or neighboring plants. The release impacts can be flammable or toxic mixtures, heat fluxes from fires or overpressures from explosion. Therefore, all three impacts, dispersion, fire and explosion need to be simulated and assessed. The site geometry was built in the preprocessor of FLACS and the various scenarios of dispersion, fire and explosion were defined and simulated in FLACS. Jet and pool fire impact were both calculated and assessed. The advantage of using CFD instead of simplified method in this case was tremendous in reducing the unnecessary conservatism incurred by the use of simplified methods and yet shedding lights on some hazardous impacts that would have not been identified by simplified models. During the facility siting it became clear that the topography in the area have great impact on flames, heat fluxes and smoke spreading. Figure 4 clearly shows the combined effects of topography (geometric model) and wind conditions on smoke propagations its impact on the neighboring areas. On the other hand, Figure 5 shows a jet fire resulting from the ignition of an accidental natural gas release from the pressure relief valves located on the top of a pressurized storage tank. The picture shows wind effects on the flame and resulting heat fluxes. The wind flow pattern was heavily affected by the obstacles in this scenario especially the storage tank.
WIND FIRE Figure 4 WIND FIRE Combined effects of topography and wind direction on smoke propagation towards neighboring areas. Figure 5 Top view of an ignited accidental release from a pressure relief valve on top of an LNG storage tank.
Fire impact on the structure of an offshore facility As mentioned earlier in this article, the application of passive fire protection (PFP) have always been essential in protecting critical structures. However, its excessive application incurs additional costs and drawbacks such as corrosion and increased congestion. In this application, the impact of fire on critical structural elements was calculated and the results outsourced in order to calculate the structural response of the critical elements. The use of CFD together with structural response provided means to assess the performance of structures under fire impact and identify the elements which require passive fire protection. The release profiles were defined based on the inventories existing in the area. Figure 6 shows a jet fire resulting from ignition of a natural gas release with a flow rate of 5 kg/s. The horizontal jet is impinging on the support beams which are required to be intact in order to prevent equipment collapse and escalation. The picture shows both the flame and spatial distribution of heat fluxes on the beams. The spatial resolution of heat fluxes in this case is crucial as it allows the application of PFP only on the required areas. Figure 6 Three-dimensional temperature contours of a 5 kg/s natural gas jet fire impinging on crossing beams of a main structure.
CONCLUSIONS Fires and explosions are the major hazards encountered in the oil and gas industry. Fires are frequently encountered as they occur following an immediately ignited release and following an explosion due to delayed ignition. The various types of fire were described and the available methods to assess fire impact were discussed. The paper stressed on the fact that the various assessment methods of fire impact should be well understood and applied only in their range of validity because applying the correlations outside their range of defined conditions can provide invalid results. The tabulated experimental values are mostly used to estimate heat loads and temperatures on flame-engulfed objects. The experimental values can be conservative when using the same flux impact for the entire element in question. If the conservatism of using these recommended values is affordable therefore the use of guidance can be considered sufficient. Some empirical models using idealized flame shape are applicable for radiative heat exposure outside the fire and within one flame length. Other simpler empirical models are applicable in the far field beyond two flame length away. These models can provide very reasonable predictions where their range of applicability stands such as no obstacles between the flame and target. These models are also useful for screening purpose were a large number of scenarios is required to be assessed in order to identify a selected number of design scenarios. However, caution should be exercised to compensate for the assumptions made in these models. Compared to the simplified models, CFD tools can be used in a wider range including most of the applications encountered in petroleum fire accidents. CFD simulations are relatively effortdemanding and expensive compared to the simplified models. However, in some applications the use of CFD is verified and beneficial when the expenses induced by conservatism out-weigh those of the CFD simulations or where simplified models fail to reflect an accurate picture of the fire impact. FLACS is a CFD software used in the oil and gas industry to conduct dispersion, explosion and fire simulations. The code is suited for applications with large domains that are normally encountered in the oil and gas industry. Two applications were presented in the paper illustrating the benefits gained of using CFD. In the first application, the use of CFD was crucial to avoid the conservative extents of dispersion, fire and explosion hazards in an LNG plant siting where the conservative impacts outside the facility properties could have resulted in stopping the whole project. In the second application, the optimized application of PFP for critical elements was assisted by the use of CFD to predict fire impact on the elements and their structural response. When the performance-based approach is used to replace the regulation-compliance attitude, one has to ensure that the performance is predicted using the correct tool for the application. A thorough understanding of the tools and their range of validity is crucial in the performance-based safety approach. The simplified methods can be useful when applied in their range of validity however in many of the practical cases more advanced tools are required in order to reduce conservativism or capture the complete picture of fire impact. CFD tools, although more expensive, have been found to provide more accurate and scenario-specific results compared to the other simpler methods.
REFERENCES 1. MARSH Energy Practice, The 100 largest losses 1972-2011 Large property damage losses in the hydrocarbon industry 22 nd Edition, 2012 2. FABIG, Design Guidance for Hydrocarbon Fires, Technical Note 13, September 2014 3. NORSOK STANDARD Technical Safety, S-001, Rev. 3, January 2000 4. Blast and Fire Engineering for Topside Structures, Phase I and II, Steel Construction Institute 5. SINTEF & SCANDPOWER, Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry, April 2005 6. GEXCON FLACS, A CFD Software for Dispersion, Fire and Explosion Simulations