6. Response to fires 154
6.4 Methods of assessment
6.4.1 General
There are two possible approaches to carrying out structural analysis for the fire condition. Design codes such as BS5950-8 [6.14] and EC3-1-2 [6.15] offer simple ways of checking elements.
However, the codes were written for building structures and will not always be suitable for highly redundant offshore structures. Alternatively there are various types of finite element analysis available which are capable of analysing large substructures or even the whole structure. It is also possible to simply modify the existing “normal” or cold analysis by adopting elevated temperature material properties.
In order to analyse any structure in fire a thermal model is required. For simple linear elements, all that is required is the temperature distribution across the section at the mid point. This may be computed using a 2-D thermal analysis. For more complex elements and whole structures, ideally, the complete temperature history of all parts of the structure is required although some simplification may be possible.
In carrying out a thermal analysis, the modelling of proprietary fire protection is not straightforward.
For fairly simple insulating materials, it should be possible to obtain a reasonable estimate of the thermal properties. Intumescent materials behave in a very complex manner, as they react differently in different situations. The local thickness of steel and the heating rate are important.
When carrying out any analysis, it is necessary to establish the applied loads on the structure.
BS5950-8 [6.16] and EC3-1-2 [6.17] allow loads to be reduced below the normal design values in fire as it is considered that the probability of fire and full design load occurring at the same time is rare. BS5950 is slightly more conservative than the Eurocodes. For an offshore structure, the partial factors should be agreed between all parties.
It is also important to use appropriate mechanical material properties. BS5950-8 and EC3-1-2 effectively specify identical material properties for use in fire. However, BS5950 specifies different strain limits for different types of element and mode of behaviour, the elements referred to comprise composite structures not seen in the offshore industry (steel and concrete arrangements, see Sections 6.4.3.3 and 6.4.3.4 for more details).
6.4.2 Partial factors for fire
In determining the structural resistance required, the applied loads on the structure at the time of fire must be calculated. Both BS 5950-8 [6.16] and the Eurocodes allow reductions in some applied loads in fire reflecting the accidental limit state. These reductions, which are for buildings, are summarised in Table 6-5. For BS5950-8, the reductions are expressed as factors, for the
Eurocodes the reductions are expressed as Ψ1,1 factors. The use of Ψ1,1, rather than Ψ2,1, is expected to be recommended in the UK National Annex to EC1-1-2 [6.18].
In fire, the applied force or moment is given by:
1
k fi k,
G +Ψ Q where
Gk is the characteristic value of a permanent action
Qk,1 is the characteristic value of the leading variable action 1
ψfi is the combination factor for fire situation, given either by (frequent value) or (quasi-permanent value) according to paragraph 4.3.1(2) of EN 1991-1-2 [6.19].
It is expected that, in the UK, the more conservative frequent value, ψ1,1, will be used for ψfi Table 6-5 Applied load reductions in fire
BS 5950-8 Eurocode Type of load Location/type
γf ψ1,1
Office 0.50 0.50
Escape stairs and lobbies 1.00 0.70 Other (including residential) 0.80 0.50
Storage 1.00 0.90
Snow 0.00 0.20
Imposed
Wind 0.33 0.20
Permanent All 1.00 1.00
The values in the above table have been derived for buildings and may not be applicable to offshore structures. They are based on statistical evidence and are almost certainly conservative. It should be possible to derive similar information for offshore structures and subsequently eliminate possible costly over design.
As an illustration of what might happen consider wind loading. The design case for wind might be for a once in 50 year’s gust. During a fire, a structure might be vulnerable for a few hours. For the same level of reliability, the wind load might be only 20 % of the 50 year level.
6.4.3 Methods in structural design codes
6.4.3.1 Introduction
Many countries have structural design codes for fire and shortly the Eurocodes will be finalised.
Almost without exception, these codes are for building structures and may only be of limited use for offshore structures as building structures are generally much simpler than offshore structures with
The assessment methods can be used for beams and compression members. Little information is given for plated structures. It is normal to assume that members are unrestrained. Problems relating to expansion and restraint were discussed earlier in Section 6.
6.4.3.2 Member analysis
In a member analysis, the applied loads are calculated using the appropriate partial load factors.
The end reactions are generally calculated making the same assumptions that were made for the initial design. The effects of thermal restraint and any second order or P-delta effects are ignored.
Only load carrying ability is considered so deformations are ignored.
For compression members it is normal to consider the possibility that the degree of end fixity may increase in fire leading to a reduction in effective length. Codes such as EC3-1-2, allow the effective length in fire to be 50 % of the system length, although, in the UK this may be conservatively limited to 70 %. The reduction is based on two factors. Firstly, in a building a column will be constructed as a continuous member and secondly, it can reasonably be expected that the temperature at the ends will not be as high as at the mid-height position.
The method is useful for beams or columns which are not heavily restrained and for simple ties.
6.4.3.3 BS5950-8
BS5950-8 covers both non-composite construction and composite construction (steel acting with concrete). For non-composite all the guidance relates to beams, columns and tension members. It gives some guidance on unprotected steel but this is limited to 30 minutes fire resistance in the standard cellulosic fire and would not normally be applicable offshore.
For beams it gives two methods of assessment. The load ratio – limiting temperature method is largely based on fire resistance test results and is principally for I - section beams. The load ratio is the ratio between the member resistance in fire and the normal, cold, member resistance. The code assumes that the strength of a beam can be characterised by the temperature of the bottom flange and that, in some circumstances, a colder top flange will be beneficial. However, a colder top flange is assumed to be supporting a concrete floor. No guidance is given for beams supporting steel plated floors.
The second method is based on moment resistance. From knowledge of the temperature distribution across the section and the material properties at elevated temperatures, the plastic bending resistance may be computed. This method is useful for unusual sections but cannot be used without the temperature distribution. Where a comparison can be directly made, this method is slightly more conservative than the load ratio – limiting temperature method.
For members in compression, the only method given is the load ratio – limiting temperature method and the information is, again, based on standard fire resistance test data. For compression members with comparatively low slenderness, there is a built in assumption that the column will have an effective length in fire of about 85 % of the assumed cold effective length.
BS5950-8 gives simple interaction formulae to allow the load ratio to be calculated for both beams and columns.
A method for checking concrete filled structural hollow sections is given. However, the method given EC4-1-2 is more robust and is recommended.
In a useful annex, BS5950-8 gives guidance on re-use of steel following a fire and what one should look for when inspecting a building.
6.4.3.4 EC3-1-2
EC3-1-2 is for non-composite construction only. EC4-1-2 deals with composite construction.
For use in the UK (for buildings) both codes will have a national annex. All Eurocodes contain some nationally determined parameters. They also contain some informative annexes. For any country, the National Annex will give values for the nationally determined parameters and guidance on the use of informative annexes.
The structural Eurocodes are all written in the same format. The design methods start with tabular data. This is followed by simple design methods and finally there is some guidance on advanced methods.
EC3-1-2, however, has no tabular data as the only useful data would be on the protection of steels using proprietary fire protection materials. The bulk of the design information is in the form of simple calculation methods. It concludes with some guidance on advanced methods. The term
“simple” is sometimes a misnomer, as a small program or spreadsheet is required.
For beams in buildings, EC3 is generally less conservative than BS5950-8. However, for beams not supporting concrete floors it is very similar to BS5950 8. EC3 starts from the assumption that beams are uniformly heated. Their bending resistance is reduced by the reduction in yield strength.
It then allows an “adaptation” factor to be applied that may take into account of a temperature gradient and, for a continuous beam, colder support conditions.
For compression members, EC3 gives a simple method in which a non-dimensional slenderness is calculated which leads to a reduction in the squash resistances. The method is a modified form of all other Eurocode strut formula.
EC3-1-2 gives some guidance on members made from sheet steel with class 4 cross-sections.
These thin sections rapidly heat up and quickly lose strength. The guidance is for completeness and academic interest.
The strength of bolts and welds at elevated temperatures was given earlier in Table 6-2, EC3-1-2 gives some guidance on checking connections in fire. For example, for a bolt, EC3 states:
2
, , , ,
, v t Rd v Rd b M
M fi
F F k θ γ
= γ Where;
kb, θ is the reduction factor determined for the appropriate bolt temperature from Table 6-2.
Fv,Rd is the design shear resistance of the bolt per shear plane calculated assuming that the shear plane passes through the threads of the bolt
γM2 is the partial safety factor at normal temperature γM,fi is the partial safety factor for fire conditions
The important point to make is that although the reduction factor from Table 6-2 is lower than for structural steel, the partial factor at normal temperature, γM2, is 1.25 and the factor for fire, γM,fi, is 1.0. Thus the effect of the reduction factors is somewhat ameliorated.
Guidance is also given on advanced calculation methods. In this context, this refers to finite element modelling. It states that the model for mechanical response shall take account of:
• The effects of non-linear material properties, including the unfavourable effects of loading and unloading on the structural stiffness.
EC3-1-2 and EC4-1-2 have their roots in the ECCS Model code on fire engineering [6.20]. This code also includes information on fires, covered by EC1-1-2. It also contains a commentary on many of the clauses.
In due course, conflicting national standards will be withdrawn. At the time of writing, the situation is:
• The loading code, EC1-1-2 was published at a full European standard in 2002. EC3-1-2 and EC4-1-2 are undergoing final editing and should be available during 2005. For all three codes, the UK National Annexes are expected in 2007.
6.4.4 Finite element modelling
6.4.4.1 General
The use of Finite Element (FE) modelling is now becoming the norm. Packages exist which can carry out both thermal and structural modelling, incorporating Computational Fluid Dynamics (CFD), which will allow the growth and spread of fire to be modelled.
Finite element models can range from frame models with simple linear elements to complex models utilising a number of element types, some of these applications are discussed in the following sections.
In all examples and applications, the FE package being used should have been validated against test data and the engineers using the package should be trained and preferably experienced in the types of analysis being undertaken.
6.4.4.1.1 Frame models
Trusses comprising slender members or portal-like structures can be analysed as simple frames, however the analysis should be non-linear and capable of dealing with large displacements. Ideally the models should be 3D as 2D will not pick up some buckling modes.
6.4.4.1.2 Complex models
Complex finite element models should give the best prediction of structural performance. However, any model is only as good as its input data. There is little point carrying out an expensive FE analyses unless the thermal history is known with a degree of confidence and the design scenarios assumed are reasonable. For more information on FE modelling see Section 6.4.4.8.
6.4.4.1.3 Modified “cold” model
It is sometimes reasonable to use the same structural model as was used for the normal, cold, design in fire. Applied loads are appropriately factored and elevated temperature values for yield stress and Young’s modulus are used. For a structure, or parts of the structure, which are not highly restrained or which are not highly redundant the method may give reasonable answers but it is impossible to say whether the results from such an analysis are conservative or unconservative.
6.4.4.2 Structural modelling
Before any FE analysis is carried out the conceptual model of the structure should be carefully checked and possibly agreed with any potential certification authority. Consideration should be given to the need to include initial imperfections and whether a dynamic option should be included in the analysis. It is important that any analysis includes all non-linear effects and that it can model membrane action. The sensitivity of any analysis to the mesh density should be investigated
(although not necessarily for each job). Experience has shown that FE analyses of the same fire and structural scenario using the same software, carried out by more than one group, can produce widely different results. The differences are often due to differences in the conceptual model. The assumptions regarding boundary conditions must be justified. If a substructure is being analysed, the boundary condition assumptions regarding restraint thermal expansion can greatly affect results. Also, at junction between two elements is there a load path and should adjacent nodes be connected and in what way? Is the mesh sufficiently fine? Is the analysis being carried out by an experienced engineer? These are all very important considerations which must be addressed if the results are to be trusted.
In some areas it may be possible to carry out some preliminary “scoping” analyses to get some idea what answers might be expected from the FE.
Following any analysis, the results should be carefully examined and anything that looks unusual should be investigated. It may be correct or it may be due to an error in the conceptual model.
Compared with an elemental approach, any FE approach based on the same temperature distribution should give more reliable results. However, many FE models will not properly predict localised behaviour such as connection failure due to the need to refine the mesh density, unless the analyst is aware of the possibility of such failure and has made an allowance for it in the model.
The main problems in any FE modelling start with the fire. In order to get a reliable estimate of structural behaviour a reliable fire model is required. Often, designers will impose the Standard Fire (hydrocarbon, cellulosic etc) on the structure. This may meet any regulatory requirements but it can never model reality. In any real fire scenario, the heat flux impinging the structure will be different from place to place and will vary in time. Imposing the Standard Fire will not allow effects due to temperature differences to be modelled.
6.4.4.3 CFD
Computational fluid dynamics (CFD) can potentially predict the growth and movement of air, smoke, and flame. CFD is probably more complex than structural mechanics and although, researchers have been working on CFD for many years it is still in its infancy. In building design, it is used to predict smoke movement but many think it is not particularly good at predicting flashover fires. This should be less of a concern for any form of hydrocarbon fire as the pre-flashover phase will be less significant.
At present, the above cautionary advice for structural modelling, applies even more to CFD modelling. Knowledge of fire and an understanding of what a particular package is doing are paramount.
6.4.4.4 Eurocode requirements for advanced models
The structural Eurocodes all contain similar advice on using advanced models. The relevant parts are summarised below:
The analysis should include:
• The effects of non-linear material properties, including the effects of unloading on the structural stiffness and the effects of cooling;
• Validation of advanced calculation models;
Some of these requirements may appear to be very severe. It is recommended that they need not be followed for every structure analysed but they do emphasise the need to use validated software.
6.4.5 Definition and assessment of secondary steelwork
In deciding which structural members need to have their performance checked in fire the required performance for the structure for each particular limit state must be considered. All primary elements of structure will need to be assessed and will probably require some form of fire protection. A secondary member is one which, for the particular fire limit being considered, will not cause failure of a primary member or loss of compartmentation by its removal. All secondary members require assessment but may not require protection.
For example, a secondary beam, spanning between larger primary beams and supporting a plated floor may be sacrificial in fire. For the fire scenario under consideration, deformation of the floor may be unimportant. A steel plated floor system will often be able to act as a membrane and not require additional support. The beam may not be critical for giving restraint to the primary beam.
However, in a severe fire heat may be conducted along an unprotected beam into the primary beam and thus reduce the fire resistance of the primary beam. For practical reasons it might be better to protect the entire secondary beam rather than simple coating the ends.
Secondary members, which when cold, restrain a primary member may require fire protection to continue fulfilling this function when hot. However, experience has shown that at the reduced applied loads in fire, the restraint may not be necessary. For example, loads may be resisted by membrane action and the restraint may not be required. It is important to consider that it does not follow that a member which carries load will always be required in fire. The function of all members should be looked at. Only members which may fail or deform in fire leading to a performance requirement not being met should be considered for protection.
Simple design methods are not able to provide information on whether secondary members require special consideration. Only a full non-linear FE analysis will provide this information.