Passive fire protection – how much do we really need?
01 December 2021
When a steel structure is subject to the high temperatures of a fire, it starts to soften. Left long enough and the structure will eventually collapse. Applying sufficient passive fire protection (PFP) on topside structural steel members is critical.
(Click here to view article in digital edition)
Its purpose is to prevent or mitigate the serious consequences from a fire, such as to:
- Prevent escalation of fire from one area to an adjacent area
- Ensure the temporary refuge is intact for the time necessary
- Protect personnel from the fire (heat and smoke) and make escape or evacuation possible
- Protect systems and equipment of essential importance for safety
- Maintain structural integrity for the required period of time
Figure 1 – Bringing together the CFD, structural and software expertise ‘under one roof’ creates a formidable force to tackle PFP optimisation problems.
The material acts as a robust thermal insulator, delaying heat transfer to primary and secondary structural steel, hydrocarbon-equipment supports, escape routes, and pressurised hydrocarbon pipes and vessels. However, as the PFP application has not been optimised, it may not be essential on all parts of the structure.
Applying the correct amount is important. Too little and you do not go far enough to protect the people and the asset in the event of a fire. Too much and you increase costs and weights during design and create potentially unnecessary maintenance costs during operations.
As part of a recent PFP optimisation project on the topside of an FPSO, DNV endeavoured to better understand the thermal loading through probabilistic fire analysis and to calculate the structural resistance through non-linear progressive collapse analysis. This used the best tools and practices available to inform decision-making and provide an optimised solution for the PFP.
Rigorous analysis required
While rare, given the potentially serious consequences of a fire offshore, why not put PFP on everything whether it needs it or not? Aside from the obvious cost implication of over specifying PFP, there is the need to ensure it is not only applied to essential structures but also that it is regularly maintained, repaired and inspected.
The key to determining if and where PFP is really needed, is to fully understand the loads and strength of the structure to protect against fire induced collapse. PFP is normally applied in the yard during construction. To fix it offshore requires considerably more time and effort, meaning that repairs must be carefully considered. If the working model is to put PFP on whether its needed or not, then knowing where repairs are essential is unlikely. It is also worth considering that PFP can create a barrier for inspecting steelwork.
Figure 2 – The main objective of 3D fire simulations is to determine the consequences of the fires in terms of the thermal loads.
The key to determining where PFP is really needed and where it is not, is to develop the best understanding of the two key aspects common to any structural problem: what are my loads and what is the strength of the structure?
DNV uses probabilistic fire analysis to give the best prediction possible of the thermal loads, while progressive collapse analysis provides the most realistic prediction of structural capacity. The company’s best-in-class software is used to undertake both tasks and seamlessly transfer the data between the different simulation models.
How do you predict something as unpredictable as a fire? You start with what you know and work towards the answer that has the lowest, most acceptable level of uncertainty. The process starts by selecting several representative cases to cover the whole scenario with simulations through a computational fluid dynamics (CFD) application. This gives answers for a broad spectrum of possible scenarios. The results can be used to perform a probabilistic assessment to determine the most likely loadings for the desired level of safety.
As most offshore structures are large and complex with different processes in each of the areas, to make the remainder of the analysis process more manageable, a number of representative failure cases are selected. This is done by first splitting the structure into smaller areas.
For each area the hydrocarbon inventory, the probability and size of possible leaks, and probability of ignition are considered to determine a risk index against a list of possible scenarios. In simple terms, the case with the biggest risk is selected for each area. However, using that scenario in all locations within that area will be overly conservative and as such, it is sometimes necessary to sub-divide areas further to refine the scenarios.
This is the most critical part of the process as it depends greatly on the skill and expertise of the engineer to ensure that the correct balance is achieved between selecting a small enough group of scenarios to make the analysis manageable but a big enough group to ensure they are not overly conservative.
Fire simulations using CFD
The main objective of 3D fire simulations is to determine the consequences of the fires in terms of the thermal loads. This is used as input for the probabilistic assessment. Using the representative failure cases, a list of targets can be developed for the CFD assessment. This not only encompasses the nature of the fire itself but also the elevated temperatures.
There are two different types of fire to consider: a pool fire and a jet fire, the latter being the most aggressive of the two and the most complex to model.
Figure 3 – As part of a recent PFP optimisation project on the topside of an FPSO, DNV endeavoured to better understand the thermal loading through probabilistic fire analysis and to calculate the structural resistance through non-linear progressive collapse analysis – Representative image: Shutterstock.
A jet fire presents a dynamic and ever-changing set of data. The point in space where the highest radiation value is reached is not fixed but generally moves as the leak rate changes. For example, the maximum radiation on an un-impinged jet fire starts away from the source of the leak but moves closer as the leak rate decreases. As a jet fire can move in varying directions, several possible scenarios need to be considered.
As a consequence of different fire types, leak rates and directions, and the need to determine the thermal loading over a large enough area, numerous simulations should be run and a significant amount of data collected. For a typical topside structure, this may run into several hundred scenarios.
Probabilistic fire assessments
While fire modelling will create a comprehensive picture of the possible outcomes, it will not show which is most likely to occur. To take the worst case of all these cases and provide an absolute answer would be excessively conservative so as not to be worthwhile. Instead, the likelihood of each of the variables is considered in the form of a Monte Carlo simulation.
This is a mathematical technique that allows people to account for risk in quantitative analysis and decision-making. Each variable is assigned a probability function, for example a normal distribution. Thousands of cases are then run, with each variable assigned a random value based on their probability distribution. If enough cases are run, eventually, a pattern develops, and the most likely answers occur more frequently.
The outcome of a Monte Carlo simulation is not a single deterministic answer but is presented in terms of an exceedance curve.
For any given fire load, the likelihood of that occurring in any given time period is determined. If the level of safety required is known, then this can be worked back to the required fire load. Once fire loads are selected, the next step is to assess the resistance of the structure to these loads.
Non-linear progressive collapse analysis of structures
Non-linear progressive collapse analysis is a powerful tool to build a realistic picture of the capacity of a structure in situations where some damage of the structure can be tolerated. To appreciate the benefits of using non-linear collapse analysis to optimise PFP, it is first necessary to understand the behaviour of the material itself as well as that of structures built with steel.
As a ductile material, steel can stretch and deform before it breaks. This means failure is more predictable and less sudden than its predecessor, cast iron. The more load (stress) you apply, the more it deforms (strains).
In designing a structure for normal operational loads, each part of the structure needs to meet with a specified criteria from a design code. As no part can fail under the load the structure remains generally elastic. By ensuring that each piece in isolation meets the specified criteria, a structure can be designed that actually has a greater overall factor of safety than any individual element.
Progressive collapse analysis looks at the structure as a whole assembly, rather than as a series of individual parts. As the first part of the structure reaches the yield point, rather than immediately failing, it enters the plastic region and starts using the extra capacity available. This part will eventually get to the point where it is too damaged to continue and stops being effective in the overall structure. As parts of the structure start to become weak through overloading, the load will find alternate paths through stronger parts of the structure, until those parts become weak also. This continues until the structure eventually collapses. By taking this approach, we simulate as closely as possible to how the structure reacts in real life and maximise the capacity of the structure.
Gabriele Ferrara and Steven Coull, DNV
Following thorough analysis, computer models are used to test which members are integral to the overall integrity of the structure and which are not. Members which, if they were to fail during a fire, did not cause the overall structure to collapse, do not need to be protected with PFP. The concept, known as structural redundancy, can be used to determine the application of PFP accordingly. This methodology can be applied to many types of extreme loading, not just from fires.
To ensure that each aspect of the modelling is correct, the behaviour of steel under extreme temperatures must be explored. When it is subject to high temperatures, steel will become less stiff, meaning more elastic deformation and the yield point reduces, quickly moving the stress strain of the steel into the plastic zone. Both effects are important, but it is the latter that has the greatest affect.
Calculating and applying reduced material properties due to a specific member temperature is relatively simple but it is important to know the temperature of the steel, not what the temperature of the fire is. The amount that a steel section will heat up depends on numerous factors such as the shape, the thickness of the section, and the duration of exposure. Eurocode 3 (BS EN 1993-1-2:2005 Eurocode 3: Design of steel structures – Part 1-2: General rules – Structural fire design) provides guidance on how to calculate the internal temperature of the steel.
It is important to note, that PFP will not endure forever in a fire, nor will it completely remove the thermal loading. The role of PFP is to reduce the temperature for a long enough duration to provide sufficient time for either the fire to be brought under control or for personnel to be evacuated. So, while thermal effects are applied to unprotected steelwork, there will be a reduced set of thermal effects to be applied to the protected steelwork.
Accurate, reliable and holistic evidence
Bringing together the CFD, structural and software expertise ‘under one roof’ creates a formidable force to tackle PFP optimisation problems. This seamless, strategic approach is particularly important when large amounts of data need to be transferred between the fire and structural simulations.
The fire simulation software, KFX, is a DNV developed CFD code. The progressive collapse software, USFOS is developed by USFOS AS, but marketed and supported by DNV. DNV has developed interfaces between these two analysis codes that have greatly aided the efficiency of the data transfer.
About the authors:
Steven Coull is a structural engineer with over 20 years of experience in Structural Integrity Management working at Aker, AMEC, Atkins, ClerkMaxwell and currently DNV. He joined DNV in 2012 and is currently Team Lead for the Fixed and Floating Structures team who deliver client-centric solutions to structural and naval architecture problems across both the traditional hydrocarbon and renewables sectors. Over the course of his career, Steven has worked in all areas of structural integrity management from structural analysis to inspection planning and management.
Gabriele Ferrara is a Chemical Engineer with 20 years’ experience in Technical Safety and Loss Prevention. He has previously worked as a senior Loss Prevention specialist in Saipem and has joined DNV since 2013. Since joining DNV, he has worked on the quantitative evaluation of risk to support design, advise clients, and improve safety both on- and offshore across the globe. He holds a PhD on Gas Explosions and is author and co-author of many papers on international journals and/or conferences about the experimental and numerical study (CFD) of explosions and fires.
Contact Details and Archive...