ultimate strength of structural members with …

10
ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH ATTACHMENTS DURING ACCIDENTAL FIRES Jørgen Amdahl 1 , Tore Holmås 2 and Bjørn Skallerud 1 1 Norwegian University of Science and Technology 2 SINTEF, Trondheim, Norway ABSTRACT A fire test is carried out with a HEA-240 girder with unprotected, secondary stiffeners penetrating the passive fire protection on the girder. The temperature development in the girder and the mechanical response up to collapse are measured. The test results are compared with theoretical analysis of transient heat transfer and strength degradation. The effect of fire protection on the time to failure is estimated in order to assess coat-back requirements. INTRODUCTION Accidental fires are events with a severe consequence potential, demonstrated by the Piper Alpha catastrophe in 1986. Adequate design against fires is therefore essential in the offshore industry as well as in many other industries and sectors. Traditional fire design is based upon several assumptions, which are often of simplistic nature. Standard fire scenarios are typically used (standard temperature-time curves or heat flux histories), and the temperature evolution and the resistance of the load-carrying members with and without passive fire protection (PFP) are estimated by means of standard tables. Neither the fire nor the heat transfer and the mechanical response reflect the real processes in a fire. Consequently, the design often becomes overly conservative, notably with respect to excessive application of passive fire protection. In the joint industry project Integrated Analysis of Steel and Aluminium Structures Exposed to Fire, conducted by SINTEF in the period 1990-2002, it has been the goal to contribute to development and application of modern and advanced tools for fire design, in which the governing phenomena are modelled by a fundamental description of the physical processes. This applies to all three steps of fire analysis and design: the fire process, the heat transfer and the mechanical response. In the first step the combustion process is simulated on the basis of the actual amount of combustible material, air supply, ventilation etc. The program utilised in the present project is KHAMELEON FIREX [1]. In the second step the temperatures and fluxes obtained in the simulation of the fire process are used in a time domain finite element analysis of the heat transfer taking into account the effect of any PFP. Finally, on the basis of the temperature evolution in the structure, the degradation of the resistance of the structure system is simulated in a nonlinear finite element analysis. A key issue in the project has been to develop efficient interfaces between the three analysis steps with respect to data transfer as well as finite element modelling. The integrated program system has been used in several offshore projects in the North Sea and elsewhere. The experience from these projects is that significant cost savings with respect to PFP can be obtained, when the design is based upon reality engineering, i.e. simulation of realistic fire scenarios based on fundamental principles.

Upload: others

Post on 25-Feb-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH ATTACHMENTS DURING

ACCIDENTAL FIRES

Jørgen Amdahl 1, Tore Holmås 2 and Bjørn Skallerud1

1Norwegian University of Science and Technology

2 SINTEF, Trondheim, Norway

ABSTRACT A fire test is carried out with a HEA-240 girder with unprotected, secondary stiffeners penetrating the passive fire protection on the girder. The temperature development in the girder and the mechanical response up to collapse are measured. The test results are compared with theoretical analysis of transient heat transfer and strength degradation. The effect of fire protection on the time to failure is estimated in order to assess coat-back requirements. INTRODUCTION Accidental fires are events with a severe consequence potential, demonstrated by the Piper Alpha catastrophe in 1986. Adequate design against fires is therefore essential in the offshore industry as well as in many other industries and sectors. Traditional fire design is based upon several assumptions, which are often of simplistic nature. Standard fire scenarios are typically used (standard temperature-time curves or heat flux histories), and the temperature evolution and the resistance of the load-carrying members with and without passive fire protection (PFP) are estimated by means of standard tables. Neither the fire nor the heat transfer and the mechanical response reflect the real processes in a fire. Consequently, the design often becomes overly conservative, notably with respect to excessive application of passive fire protection. In the joint industry project Integrated Analysis of Steel and Aluminium Structures Exposed to Fire, conducted by SINTEF in the period 1990-2002, it has been the goal to contribute to development and application of modern and advanced tools for fire design, in which the governing phenomena are modelled by a fundamental description of the physical processes. This applies to all three steps of fire analysis and design: the fire process, the heat transfer and the mechanical response. In the first step the combustion process is simulated on the basis of the actual amount of combustible material, air supply, ventilation etc. The program utilised in the present project is KHAMELEON FIREX [1]. In the second step the temperatures and fluxes obtained in the simulation of the fire process are used in a time domain finite element analysis of the heat transfer taking into account the effect of any PFP. Finally, on the basis of the temperature evolution in the structure, the degradation of the resistance of the structure system is simulated in a nonlinear finite element analysis. A key issue in the project has been to develop efficient interfaces between the three analysis steps with respect to data transfer as well as finite element modelling. The integrated program system has been used in several offshore projects in the North Sea and elsewhere. The experience from these projects is that significant cost savings with respect to PFP can be obtained, when the design is based upon reality engineering, i.e. simulation of realistic fire scenarios based on fundamental principles.

jamda
Note
Int. Conf. Response of Structures to Extreme Loading, Toronto; Canada, 2003-08-03/06
Page 2: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

SCOPE OF PRESENT WORK The main purpose of a direct design against fire is to ensure that the main load-carrying members maintain their function during the fire duration. For offshore deck structures important load-carrying members are plate-girders and I-profile girders, to which the deck plating is attached. The stiffeners of the deck plating are typically welded to the web plate of the girder. If the girder is exposed to substantial heating in a fire situation passive fire protection must be applied in order to keep the temperatures below acceptable limits during the fire time. The acceptable temperature depends upon the tolerable strength reduction of the plate girder as determined in a systems analysis. Frequently PFP is not applied to the stiffened plating, and the stiffeners penetrate therefore necessarily the PFP. As the unprotected material often will be heated to substantial temperatures within a few minutes, the risk exists that excessive heat will be transferred to the girder through conduction in the stiffeners. In order to avoid this, it has been common practise to apply PFP also on the secondary stiffeners in the vicinity of to the attachment to the girder. This so-called coat-back length varies, but has as a rule-of-thumb been taken as 450 mm. Application of coat-back PFP is a tedious and costly process. Any reduction of coat-back requirements may therefore leads to substantial cost-savings and improves the economics of the project. The purpose of this work is to gain further knowledge of the heat conduction from secondary members and its effect on the resistance on the main load-carrying members. This will, in turn, contribute to development of more appropriate design guidance for application of PFP and notably coat-back requirements. EXPERIMENTS Test specimen The test specimen is shown in Figure 1. The girder is an HEA-240 profile with a length of 4700 mm. The secondary stiffeners consist of HEA-100 profiles at a spacing of 600 mm. They are included at a length of 280 mm to each side of the girder. The secondary stiffeners are welded to the web of the girder as close as possible to the top-flange. The girder is insulated with steel mesh reinforced epoxy coating, with an average thickness of 6.2 mm. The insulation is applied in a recognised yard by experienced personnel according to standard procedures for offshore structures. The secondary stiffeners are left unprotected.

Figure 1 Test specimen

Page 3: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

The girder is simply supported so that the free span is 4340 mm. A single pin load is applied through a hydraulic cylinder at mid-span (pt. C, refer Figure 1). The girder is located on the top of the furnace allowing exposure to heating from three sides. On the top flange, which corresponds to the horizontal opening of the furnace, lightweight concrete (SIPOREX) blocks are placed in order to close the furnace. The concrete blocks follow the deflections of the girder. The furnace is heated according to the standard hydrocarbon fire specified by the Norwegian Petroleum Directorate [2]. The furnace temperature is routinely controlled by thermocouples at various locations in the furnace. In addition, the temperature close to the specimen is measured by means of thermocouples located 100 mm from the lower flange of the girder at four different positions. Steel temperatures are measured by means of sheeted thermocouples with a diameter of 1.5 mm peened into drilled holes with a depth of half the metal thickness. The temperatures are recorded at three sections (C, L1, R1) where a secondary stiffener is attached and two sections (C-L1, C- R1) midway between secondary stiffeners. Measurements are conducted at the bottom flange, top flange at web mid-height and in the web of the unprotected secondary stiffeners. The vertical deflection of the beam at the pin load is also recorded along with the magnitude of the pin-load. Further details of the test are given in [3] TEST PROCEDURE The test starts by applying the pin-load of 92 kN. This corresponds to a bending stress of 150 MPa and a utilisation of 42% with respect to first yield. Next, the heating is started. The pin-load is held constant until noticeable vertical deformation in way of the pin-load is observed, indicating that the girder is close to collapse. Then the load is reduced to a lower level and held constant. When the vertical deformation once more has reached certain level, the load is reduced. The procedure is repeated a number of times. The idea with this procedure is to prolong the collapse process considerably, so as to obtain more information about the behaviour of the fire protection during long heat exposure and finite deformations. In addition, the approach provides more interesting data for verification of the mechanical response simulations. Test observations The surface of the coating starts to react within a minute of the fire time and a protective layer of char is formed. After 4 minutes a minor damage in the char on the bottom right flange is observed. This damage develops further, and after 6 minutes a piece of approximately 100 x 100 mm falls off close to mid-span. The steel is still covered by the reinforcement and some epoxy. The remaining epoxy reacts and fills the damage part. From 12-30 minutes observations are made difficult due to the flames. After 40 minutes deep cracks are observed in the expanded coating on each side of the bottom flange. Large parts off the char hang from the lower flange making a wide-open crack into the steel. The steel mesh and some char are still protecting the flange to some extent. The partial loss of PFP affects the temperature of the bottm right flange at all measurement positions between sections L1 and C-R1. After 50 minutes the vertical deflection has reached approximately 30 mm, and the load is reduced to 80 kN. The load is then reduced in steps of 10 kN till a load level of 30 kN. The test is terminated after 75 minuets. No further change to the PFP is apparent during the load reduction phase. Temperatures typically recorded during the tests are plotted in Figure 2. It is observed that the temperature rises significantly faster in the web at the attachment than between the attachments. This is attributed to the effect of heat conduction. The HEA-100 profiles have no coat-back and attains therefore rapidly a high temperature. There is a significant gradient towards the attachment to the web, which is a clear evidence of heat being conducted into the web. The temperature in the bottom flange is probably higher than the ones measured in the web. It is also noticed that the temperature increases faster in the left part of the bottom flange of the HEA-240 profile. This is due to the partial loss of PFP early in the test. The relative small temperature difference illustrates that the remaining part of the PFP (inside the reinforcement mesh) has a very significant protection capability. The most severe heating occurs at mid-section, where the loss of PFP is

Page 4: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

most substantial. The steel is however, far from being unprotected. If this were the case, the temperature profile would be similar to the one recorded for the unprotected HEA-100 profile. All the heat increase in the right bottom flange is primarily due to loss of PFP and not due to conduction. This is evidenced by the fact that the left flange follows the “normal” temperature evolution everywhere, except for the right flange at mid-span after 50 minutes of meeting. This temperature rise is probably largely due to conduction caused by the temperature difference between the right and left flange

0

200

400

600

800

1000

1200

0 10 20 30 40 50 60 70 80Elapsed time [min]

Tem

pera

ture

[0C

]

Betw. stiff.

At stiffener

Figure 2 Temperature development at attachment and midway between attachments.

It is interesting to observe that the temperature rise in the top flange is fairly close to the one for the bottom left flange. This shows that the SIPOREX blocks on the top flange have a significant insulation effect. An important observation is that minor cracks in the PFP are not very harmful with respect to the temperature development. Figure 3 witnesses of considerable cracks in the PFP after testing, but the temperature rise is not affected significantly. The reason is that the steel material, which is uncovered locally, has a relatively small view factor to the environment, so that the resultant heat exposure is moderate

Figure 3 Test specimen after test – observe the crack in PFP on the right photo

Page 5: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

TEST SIMULATION Transient heat transfer The purpose of the transient heat transfer analysis is to obtain a temperature profile for the entire beam in the region close around mid-span. The analysis is calibrated such that the temperatures in the measuring points are close to the measured values. The final temperature field history is then used as input to the mechanical response analysis. The mid-section of the beam is modelled by means of quadrilaterals, as shown in Figure 4. It is a demanding task to perform rigorous conduction analysis of steel with passive fire protection due to very different properties; insulation is very slow heat conductor with and has a small diffusivity compared to steel. Insulation is therefore modelled using the effective heat transfer conecpt. The insulation is then simply represented by an effective heat transfer coefficient (k = c/tinsul [W/m2K]) related to the conduction coefficient (c ) and the temperature difference over the entire thickness of the insulation (tinsul) at any time. This accounts also for the fact that the insulation properties vary over time, during the ignition and expansion phase through burn-off. The effective heat conduction coefficient has to be determined from experiments and is strictly valid for a given thickness of the structure it protects. In the “steady state” phase, when the outer layers of the PFP has reacted, the heat transfer coefficient is typically ~ 5 W/m2K. The temperature fields predicted with the heat transfer analysis program FAHTS [4] after 20 minutes and 60 minutes of fire time are shown in Figure 4. The effect of heat conduction from the attachment into the web of the HEA-240 profile is observed.

Figure 4 Temperature field after 20 minutes and 60 minutes fire time

The temperature histories in the measurements points are compared with test results in Figure 5. The agreement with experimental data is very good for both the web and the lower flange. The temperature in the HEA-profile close to the web seems to be overpredicted in the first minutes of, but it should be taken into account that the predicted temperature is the average temperature for the element, while the measurement is performed in a region with a large temperature gradient. The temperature at the upper flange midway between attachments is underpredicted by approximately 150 0C after 60 minutes. This is because the insulation effect of the SIPOREX blocks is underestimated by the modelling. Further improvement is possible, but is not pursued in the present work.

Page 6: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

0

200

400

600

800

1000

1200

0 10 20 30 40 50 60 70 80Elapsed time [min]

Tem

pera

ture

[0 C]

-center top flange -side top flange

11

2

2

3

3

5

4

4

4

66

5

Test Simulation

At stiffener

Betw. stiff.

Figure 5 Temperature histories – numerical predictions versus test measurements.

The temperature in the top flange at the attachment is not recorded in the test. Numerical predictions are, however, indicated in Figure 5 by means of open squares for the flange centre and open triangles fore the flange wings. It is reasonable to assume that there is a similar underprediction of the temperatures in the top flange at the attachment because of the SIPOREX blocks, so that the true temperature is probably in the range of 450 0C – 600 0C. MECHANICAL RESPONSE The mechanical response analysis is conducted with the nonlinear finite element program USFOS [5] using the finite element mesh and the temperature histories from the heat transfer simulation. It is only necessary to include the mid section of the beam where the bending moment is largest. The ends of the modelled section are assumed to remain plane, and this feature is simulated using the bicycle wheel concept, i.e. rigid beams connecting the shell nodes to the “master” node at the elastic neutral axis. The force from the hydraulic jack is best represented as two equal concentrated loads located 50 mm to each side of the mid node. The missing length of the girder is accounted for by applying a concentrated moment proportional to the jack loads at the two “master” nodes. The degradation of material strength with temperature is assumed to follow the reduction curve for the effective yield strength according to Eurocode 3 requirements. The analysis is performed in the following way: First the temperatures incremented according to the heat transfer simulation up to a certain time instance. Then, the external loads are incremented until collapse of the section. This is carried out for various elapsed times. The procedure adopted gives more information than a simple simulation of the test, because the continuous degradation of the capacity is highlighted. Figure 6 illustrates the collapse mode using the temperature distribution after 57 minutes of fire. For the present configuration of the girder, the governing failure mode is the formation of a plastic hinge in bending. The upward deformation of the lower flange is noticed. The results of the analysis are shown in terms ultimate bending moment at mid-span as a function of elapsed time in Figure 6 along with the recorded force history (demand) from the experiment. According to the simulations the girder will fail after 61 min (when the resistance becomes smaller than the demand) while the critical load level, 63 kNm, could be maintained until 63 minutes in the test. This is considered a very satisfactory agreement.

Page 7: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

0

50

100

150

200

250

0 20 40 60 80Time [minutes]

Mom

ent K

Nm

Resistance - NLFEMDemand - test

a) b)

Figure 6 a) Collapse of mid-section using temperatures after 57 minutes of fire (fringes show temperature distribution) b) Resistance and demand versus elapsed time in test

It should also be considered that the time for collapse is not unique. For the high temperatures encountered in the bottom flange and the web close to the attachment creep is becoming significant. If the load had been kept constant for a sufficient period, assuming that the temperature could have been held constant, failure could have occurred at a higher load level and smaller time. THE EFFECT OF COAT-BACK ON THE RESISTANCE What is the effect of coat-back on the ultimate strength of the girder? This can be studied using the steel temperatures simulated in the transient heat transfer analysis. Suppose that coat back is applied to such an extent on the stiffeners that there is effectively no heat transfer from the stiffeners to the girder. It is then reasonable to assume that the temperature distribution in the girder at the stiffeners is fairly equal to that between stiffeners. Consequently, the resistance of the girder is analysed applying this temperature distribution uniformly in the longitudinal direction. This is denoted “full coat back”. Figure 7a shows the degradation of the resistance versus time for the I-girder in test. At a given time instant the resistance in case of full coat-back may be considerably larger than that with no coat-back. However, the difference in time to failure is considerably smaller. It is particularly noticed that for the load applied in the test (100 kNm), use of full coat-back will only give three minutes longer time to failure. For short beams the shear strength may be the governing parameter at the beam-ends. Analyses are also carried out for the same component subjected to predominantly shear. The same temperature distributions considered above are applied. The results are displayed in Figure 7b. As expected, the resistance in shear is more vulnerable to the heat transferred into the web, particularly in the early stage of resistance degradation. However, also in this case the time to failure does not deviate by many minutes for a utilisation in the range of 40-50%. These analyses show that coat-back has to be considered in relation to the actual application of PFP on the main load-carrying member. Heat conduction in steel is by very nature a slow process, and the amount of PFP used has to be such that the temperature development is retarded for a long period, if the heat transferred through attached stiffeners shall dominate the temperature escalation in the member. In the present case, the temperature rise in the region between stiffeners is so fast when the critical time is approached, that the heat trough the attached stiffeners becomes less important. This is also a result of the fact that the temperature difference between the stiffener and the girder reduces as the temperature in the web escalates.

Page 8: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

0

50

100

150

200

250

0 20 40 60 80Time [minutes]

Mom

ent K

Nm

Resistance - no coat backResistance - full coat back

0,00

0,10

0,20

0,30

0,40

0 20 40 60 80Time [minutes]

Shea

r for

ce [M

N]

Resistance - no coat backResistance - full coat back

a) Bending resistance b) Shear capacity

Figure 7 Resistance versus time for I-girder – with no coat back and full coat back THE EFFECT OF END CONNECTIONS ON THE RESISTANCE Girders carry their ordinary loads by bending moment and shear forces. The design of end connections is typically governed by the effects of these ordinary loads. During accidental fires the resistance with respect to bending and shear will degrade as discussed in previous sections. If the connection fails the girder will most often also fail, even if it may not be fully utilised. On the other hand, if the connections remain intact, the mode of resistance may change favourably during large deformations. By for example designing the connections to resist also a significant axial load (virtually equal to zero for ordinary loads), members deforming by bending may redistribute a significant part of the load effect to membrane forces. This is discussed in detail in a companion paper [6]. The same effect may also be identified for members with predominant shear load effect for ordinary loads. In order to illustrate this potential resistance reserve, the I-girder subjected to predominant shear load is analysed using the temperature distribution after 51 minutes. The end of the girder is clamped against rotation, but has some freedom to move in axially. The axial restraint is modelled by means of a spring to ground at the right end of the analysis model. The results of the analyses are depicted in Figure 8. Spring1 and Spring2 relates to a spring with a stiffness of 10% and 2.5% of the axial stiffness of the girder section. It appears that axial fixity may increase the resistance significantly, even for the case with low spring stiffness (Spring2). Even for the case with free axial motion the girder has almost 50% larger resistance than that corresponding to shear yield over the cross-section (using the actual yield stress reduction) over the web height. This is due to the development of a tension-compression field similar to a truss-work as illustrated in Figure 9. This compensates partly for the reduced resistance in the high temperature region close to the stiffener. How much of this additional resistance can be called upon? If rupture occurs, the beneficial effect of membrane strengthening is undermined. Figure 10 shows the strain at 0.15 m deformation. The strains are very large in the web close to the stiffener. However, this strain is dominated by compression. Obviously, the element mesh is too crude to ensure a reliable assessment of critical strains and therefore critical deformation. It is observed that the web is subjected to local buckling for Spring2, similar to what is observed in the case of free ends. For spring1 and full axial fixity, lateral buckling of web does not occur due to the fact that membrane tension prevents the web from deforming laterally.

Page 9: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

0,00

0,10

0,20

0,30

0,40

0,50

0,00 0,05 0,10 0,15 0,20Displacement [m]

Forc

e [M

N]

FixedSpring1Spring2FreeShear yield

0,0

0,4

0,8

1,2

1,6

0,00 0,10 0,20 0,30Displacement [m]

Axia

l for

ce [M

N]

spring1

spring2

a) b)

Figure 8 Shear resistance versus displacement for girder with attachment after 51 minutes of fire a) The effect of axial restraint on resistance b) Axial force in springs Yield force normal temp = 2.47 MN

a) b)

Figure 9 a) Illustration of tension-compression field in girder at ultimate resistance b) von-Mises stress

a) b)

Figure 10 a) Collapse deformation mode b) Strain distribution at 0.15 m deformation (Spring1 case)

Page 10: ULTIMATE STRENGTH OF STRUCTURAL MEMBERS WITH …

CONCLUSIONS In a redundant structure partially failed members may contribute substantially to the global resistance to accidental fires. A prerequisite is that any passive fire protection (PFP) applied to member does not crack and fall off during finite deformations. This is achieved if the PFP is reinforced. Even if small pieces of PFP fall off exterior to the reinforcement mesh, a significant residual protection remains. The test conducted with the girder with attachments and the analysis show that the need for coat-back is very little for the given test conditions. The reason for this is the slow rate of heat conduction is steel. The governing factor is the ratio of heat transmitted through the PFP versus that conducted through the attachment. Attachments are also often located at the ceiling of the fire compartment with less exposure to the fire. Even though the tests results cannot be directly transferred to real fires, they demonstrate clearly that the amount of coat-back can often be reduced considerably compared to standard practise, when focused is placed on ultimate resistance rather than steel temperatures. Careful design of member connections may increase the robustness to fires substantially. If axial restraint is provided structural members, which carry the load by bending, may redistribute to membrane action during finite deformations. Members with predominant shear action may redistribute the stresses over the cross-section, even if the shear capacity is exceeded locally and there is no axial fixity. If the member ends are restrained, the capacity increases further. Hence the design of connections should be done paying attention to requirements imposed by accidental fires. REFERENCES 1. Vembe, B.F., Lilleheie, N.I., Holen, J.K. and Magnussen, B.F. (1998) KAMELEON FireEx - A

Simulator for Gas Dispersion and Fires. 1998 Int. Gas Research Conference.

2. Norwegian Petroleum Directorate (1992) Explosion and Fire Protection of Installations in the Petroleum Activity, Norway.

3. Danielsen, U. and Sommerset, Ø. (2000) Ultimate Strength Testing of Structural Members with

Attachments. Report No. STF22 F00836, SINTEF, Trondheim, Norway. 4. Holmås, T. and and Amdahl, J. (1998) Advanced Structural Fire Design of Offshore Structures. 2nd

Int. Conf. On Thin-Walled Structures, Singapore

5. Eberg E., Hellan Ø. and Amdahl, J. (1993). Nonlinear Reassessment of Jacket Structures under Extreme Storm Cyclic Loading- Part III: Development of Structural Models for Cyclic Response, 12th Int. Conf. Offshore mechanics and arctic Eng. , OMAE ’93, Glasgow

6. Holmås, T., Amdahl, J., Skallerud, B. And Langhelle, N. (2003) Behaviour of Beam-Column Connections Subjected to Extreme Loads during Fire. Response of Structures to Extreme Loading, Toronto, Canada