You are on page 1of 21

J. Construct. Steel Res. Vol. 37, No. 2, pp.

93-113, 1996 Copyright 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved

0143-974X(95)00026-7 ELSEVIER

0143-974X/96 $15.00 + 0.00

Non-linear Analysis of Blast Walls and Stiffened Panels Subjected to Hydrocarbon Explosions
L. A. Louca, a M. Punjani b & J. E. Harding c
"The City University, Department of Civil Engineering, Northampton Square, London EC 1V 0HB, UK ~q'he University of Sheffield, Department of Computer Science, Sheffield S 1 4DP, UK "University of Surrey, Department of Civil Engineering, Guildford, Surrey GU2 5XH, UK (Received 15 August 1995; revised version received 10 November 1995; accepted 4 December 1995)

ABSTRACT This paper presents results on the response of a typical blast wall and a teestiffened panel subjected to hydrocarbon explosions with geometries typical of those used in current offshore structures. The panels have been modelled using a non-linear finite element analysis package (DYNA3D) with thin shell elements for both the plate and stiffener components accounting for the effects of plasticity, strain-rate and buckling. Correlation with previous experimental work is presented together with a parameter study highlighting the effect of an increasing peak pressure and the shape of the pressure-time curve. The effect of boundary restraints is shown to have a significant influence or,! both the torsional response of the stiffeners and the overall panel deformations. The results are also compared with a simple single degree of freedom (SDOF) dynamic model where good correlation is obtained in the region where only limited plastic deformations occur. Copyright 1996 Elsevier Science Ltd.

NOTATION

k M Y Ym R Rm

Stiffness of equivalent system Mass of equivalent system Transverse panel displacement Maximum transverse panel displacement Resistance Maximum resistance 93

94

L. A. Louca, M. Punjani, J. E. Harding

1 INTRODUCTION

The behaviour of stiffened plate components under static loading has received considerable attention in the past due to their buckling efficiency in resisting in-plane and out-of-plane loading. In offshore applications, stiffened panels are used extensively as floorplating and bulkheads acting as fire and blast walls. For many existing platforms, topside structures have been designed using codes of practice based on static loading. For instance, floorplating on an upper deck would have been designed to resist static imposed loading from equipment with no allowance for any upward pressure caused by an explosion below. Similarly, bulkheads which are used to protect equipment and personnel from extreme load situations may have only been designed for fire loading but not for combined fire and explosion loading. Since the publication of the Cullen repoW on the Piper Alpha tragedy, safety cases are required by the Health and Safety Executive (HSE) to demonstrate that the offshore installation meets stringent safety criteria. This involves assessing the structural response of stiffened panels to a given blast loading with the aim of containing or venting the blast to prevent an escalation of the consequences of the event. As a result of a Joint Industry Project on Blast and Fire Engineering for Topside Structures, Interim Guidance Notes2 (IGN) have been issued by the Steel Construction Institute suggesting techniques for assessing stiffened panel behaviour including the use of transient non-linear finite element techniques and simple single degree of freedom (SDOF) models consisting of a spring mass system based on an approach developed by Biggs. 3 Leach4 has recently extended the Biggs method to panels with different boundary conditions and has provided design charts for the elasto-plastic design of cladding panels subject to blast loading. The work also highlights the need for calibrating the approach with non-linear finite element analysis. This present study presents some results in this area. The effect of torsional buckling of the stiffener outstand components on the response has been highlighted by Louca and Harding 5'6 in situations where stiffened panels have been loaded statically in uniaxial compression. Walker 7 has recently shown the effect of stiffener tripping under dynamic loading and has highlighted the need for careful modelling of the outstands in order to pick up this local buckling mode as the stiffness and strength of a stiffened panel can be seriously reduced if this failure mode occurs. A series of blast tests carried out by Schleyer8 for British Gas on bulb-flat stiffened panels found that when the stiffeners were facing the shock front the maximum central displacement obtained was only 15% less than an unstiffened panel of the same overall slenderness, due to stiffener buckling (tripping) reducing the beneficial effect of the stiffener. When the panel was tested such that the

Non-linear analysis of blast walls and stiffened panels

95

stiffeners were in tension, no tripping occurred and the central deflection was almost reduced to half that of the equivalent unstiffened panel. This paper presents results highlighting the effect of boundary restraints, modelling and variation in the pressure time history of the load on the overall and local response of a typical blast wall. A simplified model of the wall consisting of an individual stiffener and associated plating has also been used to highlight the effects of boundary restraints on the tripping behaviour of the stiffener. The results are also compared with the single degree of freedom dynamic analysis developed by Biggs which is commonly used in the offshore industry, and its limitations are highlighted.

2 NUMERICAL MODELS

In order to investigate the response of the blast wall and idealised stiffened panel subjected to dynamic loading due to a hydrocarbon explosion, a nonlinear finite element (FE) code DYNA3D has been used to carry out the study. The package is capable of modelling geometric and material non-linearity, local buckling and strain rate effects. The package is an explicit code and is highly suited to modelling transient problems where high rate dynamics and stress wave propagation effects are important. The code does not function as most FE codes which are based on an implicit formulation. A large stiffness matrix does not have to be solved and does not require iteration at each time step consequently placing less demand on computer resources. In order to initiate buckling, 'crimp' imperfections were applied to the stiffener webs in the transverse direction at a single node at the stiffener midspan and a lateral imperfection was applied across the width of the stiffener flange components at the same location. The effect of imperfections has received a great deal of attention in the past, highlighting the sensitivity of the response to varying imperfection mode. 9 Most primary structural members used in modem offshore structures comprise of high strength steel although mild steel is used for secondary members. Most existing blast wall structures consist of grade 50 steel and this has therefore been used as the material for both the plate and stiffener components. It is well known that steel is strain-rate sensitive with high loading rates significantly enhancing the yield strength, particularly for mild steel. In order to account for this effect a rate-dependent isotropic elastic-plastic material model was adopted using an expression given in the IGN, 2 which is an extrapolation of the Cowper Symonds model.

96

L. A. Louca, M. Punjani, J. E. Harding

3 CORRELATION WITH EXPERIMENTAL WORK A recent joint industry project ~ published by the Steel Construction Institute gives results for a simple problem which was analysed by a number of different companies using a range of available FE packages. A considerable scatter of results was obtained highlighting the importance of establishing reliable results by validating the package used with experimental or previous reliable theoretical results. It was therefore felt essential to correlate DYNA3D with previous experimental data to ensure its validity as a design and assessment tool for blast loading problems. A test on a large scale tee-stiffened panel reported by Houlston and DesRochers ~ has been assessed for this exercise. The geometry and the static material properties are shown in Fig. 1, together with the position of accelerometers and displacement gauges A2, A3, DI2 and D13, respectively. The numbering reflects that used in the original research and has been retained for cross-reference but has no direct significance to the results presented in this paper. The panel is made up of a 4.572 m 2.438 m steel plate with a thickness of 6 mm stiffened by four tee stiffeners, with flange widths of 76 mm, web heights of 152 mm, and thicknesses of 13.3 and 7 mm, respectively. The stiffeners were cut from rolled sections and welded to the panel at a spacing of 0.914 m. The edges of the panel were cast into concrete which provided a fully fixed boundary condition in all directions. The pressure on the panel was recorded at a number of points and the results were averaged to produce a single pressure time history with a peak pressure of 3.45 bar, ignoring any spatial variation of the load across the panel. This was not considered to be significant in this case. The loading curve is indicated in Fig. 2. In order to simplify the FE model, the response of the multi-stiffened panel
914 CLAMPED

K T

ARIES

+ A2., D12

A~, D13 8

Yield strength=310N/mm2 Youngs modulus=207000N/ram2 Poissons ratio=0.3 Modulus density=7850Kg/rd 3

4572

Fig. 1. Sketch of test panel used for correlation exercise.

Non-linear analysis of blast walls and stiffened panels


4.00 - - r

97

--

Pressure (bar)

3.00

Q..

2.00

1.00
0.00

'
10.00

I
20.00

'

I
30.00

'

I
40.00

I 50.00

T i m e (ms)

Fig. 2. Experimentalpressure-time curve. was idealised using a single stiffener with associated plating. Previous studies on the response of multi-stiffened panels using this idealisation for modelling static tests has shown that it gives good correlation with experimental results although it must be acknowledged that with a limited number of stiffeners the overall panel restraint can have a significant effect. The mesh adopted is shown in Fig. 3. The longitudinal supports (parallel to the stiffener) of the panel assume restraint in the transverse in-plane direction and restraint of rotation about the two axes. The edge is free to displace vertically (out-ofplane) and no shear is allowed to develop along the longitudinal edges. The two ends of the panel, including the ends of the stiffener, are fully clamped providing restraint in all directions. The variation of displacements with time at the points shown in Fig. 1 obtained from the experiment are summarised in Fig. 4 and compared with the equivalent DYNA3D results, together with

98

L. A. Louca, M. Punjani, J. E. Harding

Fig. 3. Mesh of idealisedpanel.

results obtained from an implicit FE package (ADINA) by the authors of Ref. 11. In general, the correlation is excellent between the DYNA3D results and the experiment considering the complex nature of both the loading and the response and the simplifications involved in the modelling. The ADINA results are also reasonable. The curves corresponding to the displacements D12 and D 13 are more accurate than those relating to A2 and A3, the latter being the displacements obtained from manipulating the acceleration responses. The experimental failure was concentrated in the central part of the panel at 5 ms from the initiation of loading. This also occurred in the FE model. The panel and stiffener then oscillate elastically about the deformed shape. The ADINA results show larger oscillations than both the experimental and DYNA3D results. This may be due to the fact that coarser meshes are

Non-linear analysis of blast walls and stiffened panels


A

99

E E

::

~-20 E O m -40 ZL ._~ c) -60


O

AD NA

-2
A,

/Tx
I J , I '

,--..,-x.~-'~.%.3._
| I I I I ' I l I I I I I

-6 '*

10

15

20

25 T i m e (ms)

30

35

40

45

50

+ + DYNA3D results 0 E E -10


A C

~-20
O O

' I ~

0
v

..j~,.

ADINA
/

-2
Q)

D13

-3

E ~

--~-30 ._~ a -40 0

-5 ~
I I ' I ' I ' I l I ' I ' I
I --6

10

15

20

25

30

35

40

45

50

T i m e (ms)

Fig. 4. Comparisons between experimental and theoretical displacement-time histories. DI2 and D13 are displacements measured directly during the experiment. A2 and A3 are displacements calculated from the acceleration readings. Positions of D12, D13, A2 and A3 are shown in Fig. 1.

generally adopted with implicit FE codes, which may not pick up the higher frequency modes accurately.

4 BLAST W A L L MODEL In order to investigate the response of a typical blast wall for an offshore installation, two models were investigated. The first was a complete model of a typical bulkhead consisting of longitudinally stiffened panels supported by transverse girders and the second a single stiffener with associated plating. The complete wall consists of vertical transverse girders, with 900 m m deep webs of thickness 15 m m and 400 m m wide flanges with a thickness Of 35 m m which support stiffened panels 12.5 m m thick with six longitudinal tee stiffeners, cut from universal beams. The resulting dimension of the Ts are a web depth and thickness of 229 and 9.1 mm, respectively, and flange width and thickness of 191 and 14.5 mm, respectively. The wall is 27 m long overall but in order to reduce the size of the model symmetrical boundary conditions

100

L. A. Louca, M. Punjani, J. E. Harding

have been used to model half of the wall. The model width is 5 m. Figure 5 shows the mesh adopted for the wall which consists of 2472 shell elements.

4.1 Effect of pressure-time history


Typical pressure-time histories resulting from a hydrocarbon explosion tend to exhibit an appreciable rise time and can have a relatively long duration. The response of stiffened panels can be sensitive to the shape of the pressuretime curve depending on the natural period of the structure in relation to the duration of the loading. A previous study 12 has suggested using three different load curves which are shown in Fig. 6. The first is a symmetrical triangular pulse which has been commonly adopted in the past. The second assumes a sharper rise time, with the total impulse energy and peak pressure being the same as the first curve. The IGN suggest that the structure can be quite sensi-

Longitudinal stiffeners

Fig. 5. Blast wall model.

Non-linear analysis of blast walls and stiffened panels


1.10 - . 1.00 - - 0.90 --0.80 ---

101

/
'

/'\

"\
"\,' /

,'/
/

."~

x ",
\ ",

",

__Load curve 1 _.Loadcurve2 ---Loadcurve3

0,70 --.O

//

P
D

0.60---

0.50
0.40 0.30 0.20 0.10

// / //./
0 10 20
30 40 50 Duration

', "'..
I

0
60
(ms)

70

80

90

100

110

Fig. 6. Idealised pressure-time curves.

five to the rise time if the response is in the dynamic or quasi-static range. The third load curve has a slightly higher total impulse energy but is an idealisation of a double peak often observed in experimental results of pressuretime curves of a gas-type explosion. The effect of using such a curve compared with load case 1 can be quite significant if the structure just reaches yield with load curve 1, as a significant increase in the deformation may result from load curve 3. Initially, a natural frequency analysis of the blast wall was performed to establish the lowest fundamental period and mode shape to obtain an indication of the response under a given duration of load. Figure 7 shows the lowest mode which gave a period of 113 ms. The mode shows torsional and lateral deflections occurring in the two longitudinal stiffeners at the lower boundary of the panel. From this, it was decided to use a total loading duration of 100 ms which would produce a response in the dynamic range. Figure 8 shows a summary of the peak displacement of the blast wall obtained with increasing pressure for the three load curves shown in Fig. 6 for a total loading duration of 100 ms. All three load cases give a distinct decrease in stiffness beyond a pressure of 1 bar. At 4 bar the peak displacement of load case 2 can be seen to be significantly greater than the others. This can be explained by the fact that the natural period of the wall is 113 ms placing the response in the dynamic range which can be significantly affected

102

L. A. Louca, M. PunjanL J. E. Harding

Fig. 7. Lowest mode of vibration correspondingto a period of 113 ms.

by the rise time. This is despite the fact that load curve 3 imparts a higher total impulse energy to the wall. The displacement-time history of the wall when subjected to different peak pressures using load curve 3 is shown in Fig. 9 with the load uniformly applied over the plate such that the longitudinal stiffeners were placed in tension. Clearly under a 1 bar peak pressure the response is purely elastic as the curve shows no permanent deformation at the end of the loading duration. Detailed study of the results shows that the largest stresses occur at the ends of the transverse girders and at their intersection with the longitudinal stiffeners. This

Non-linear analysis of blast walls and stiffened panels


200 - - L o a d curve 1 - - ' - L o a d curve 2 - x- Load curve 3
/ / / / / x

103

E E
e-

.
"/
100

,Y

E
0 0 O.

=5
a) (.!.

T
0

I
1.0

I
2.0

'

I
3,0

l
4.0

'
5.0

Peak pressure (bar) Fig. 8. Effect of pressure-time curve on peak response.

suggests that geometric stress concentrations are occurring at these points which may lead to brittle weld failures. Increasing the peak pressure to 2 bar causes yielding to occur at 35 ms from the onset of loading at the ends of the transverse girders. This spreads along the girders and causes yielding at the junction of the outer longitudinal stiffener with the transverse girder. A maximum displacement of 61 mm occurs in the central panel of the wall between the two transverse girders. The transverse girders also undergo a significant deformation of approximately two thirds of the panel value. This exceeds current static design criteria (span/200) given in BS 595013 for the structural use of steelwork in buildings. No yielding occurs at the edges of the panel. At 3 bar peak pressure, yielding again occurs at the ends of the transverse girders and spreads along the two outer stiffeners at an earlier stage. At 55 ms the yielding on the transverse

104

L. A. Louca, M. Punjani, J. E. Harding


200

- ----+r .... ---.x-

4 bar 3 bar 2 bar 1 bar

150 --

E
c

E
rj (/J

100

E3

50

--

I
0

I
50 Time (ms) 75

25

Fig. 9. Response of blast wall model to increasing peak pressure for load curve 3.

girder spreads along the second longitudinal stiffener, but more importantly the longitudinal edge supports and the panel between the longitudinal stiffeners begin to yield and this becomes quite extensive at 60 ms. While peak displacement is large at 117 mm, provided adequate fixity can be provided at

Non-linear analysis of blast walls and stiffened panels

105

the supports so that the plate can develop its membrane capacity, the load may be resisted. The blast wall model was also uniformly loaded on the stiffener side such that the longitudinal stiffeners were placed in compression. Surprisingly little change in response occurred in this instance with only an overall increase of about 5 0 in the peak panel displacement due to some earlier yielding, which as before occurred at the junction of the outer longitudinal stiffeners with the transverse girder. However, more importantly, no torsional deformation of the stiffeners occurred.
4.2 Effect of transverse membrane restraint

If adequate in-plane restraint is provided to a steel plate component at its edges, high out-of-plane forces can be resisted by membrane action. In offshore construction blast walls are normally attached to a plate girder and the fixity provided can vary. Figure l0 shows the effect on the response of the blast wall if no transverse in-plane restraint is provided to the top longitudinal panel edge. Load case 3 has been applied with a total duration of 100 ms. At 1 bar the response is wholly elastic and the difference in restraint has no effect. Even at 2 bar the effect is only slight. However, beyond this point a marked change in the overall response occurs with the unrestrained edge suffering severe in-plane and local lateral displacement.

5 SIMPLIHED HNITE ELEMENT MODEL Although there are limited data available on the failure criteria for welded joints subjected to blast loading, experience suggests that the weld failure stresses at points of stress concentration may be a limiting factor. Failure of these welds may effectively change the boundary condition at the end of a stiffener from being built in to having little or no support at all. The results of the previous section suggest that loss of restraint at the stiffener end due to such a weld failure could be important and in order to investigate this further and to carry out a more detailed study a simpler FE model has been set up. This consists of an individual longitudinal stiffener with associated plating representing the longitudinally stiffened panel of the blast wall between two transverse girders. Figure 11 shows the idealised model. As with the experimental correlation, the longitudinal edges have symmetrical boundary condition imposed on the model.

106
1000

L. A. Louca, M. Punjani, J. E. Harding

900

--

Restrained

,7

800

700

E
o

600

E o t~ .--. "o ,.~

500

400

300

200

100

1.0

2.0

3.0

4.0

5.0

Peak pressure (bar)

Fig. 10. Effect of transverse membrane restraint.

5.1 Effect of in-plane boundary restraints


The boundary restraints at the end of the panel have been assumed to be simply supported with two different inplane stiffener restraints. The first assumes the stiffener is restrained in-plane (SR) at one end whilst the other assumes the stiffener is allowed to displace in-plane at both ends (SU). The panels have been loaded with the plate facing the shock front with load case 3 such that the stiffeners were placed in tension. Figure 12 shows a summary of the results for a peak pressure up to 4 bar and a total loading duration of 100 ms. Up to a peak pressure of about 2 bar the difference in response with boundary restraint is fairly negligible with the response predominantly in the elastic region. However, beyond this point the

Non-linear analysis of blast walls and stiffened panels


Transverse girders

107

i
SR/SU t -

i /

Lon itudinal stiffener

,4. ~,~

SO

Longitudinal edges- symmetrical boundary conditions


Fig. 11. Sketch of idealised panel between two transverse girders.

response shows a distinct variation with boundary restraint with the unrestrained stiffener showing significantly larger deformations. The panel was then reversed, such that the stiffeners were facing the shock front. In this case the stiffeners suffered severe tripping deformation despite the fact that the section is classified as plastic. This behaviour was not picked up in the blast wall model. The tripping deformations occurred at a peak pressure of 3 bar for the SU condition whilst in the SR state it did not occur until the peak pressure reached 4 bar. Figure 13 shows the deformed mesh for the SU condition highlighting the tripping failure in the tee stiffener using load case 3 with a peak pressure of 3 bar, together with the displacement-time history. The tripping appears to be the primary failure mode with yielding beginning in the stiffener at mid span. The peak displacement in this case was 475 mm compared with a value of 341 mm for the previous case described with the stiffener in tension. The difference in the overall panel displacement is almost certainly entirely caused by the tripping behaviour.

6 SINGLE DEGREE OF FREEDOM DYNAMIC MODEL In of to of using an FE code to analyse a blast wall, even with the savings from use an explicit code, significant computer resources and man hours are required carry out such a complex analysis. If a sensitivity study is required as part a preliminary design, simpler and faster techniques are required which can

108
2000

L. A. Louca, M. Punjani, J. E. Harding

-NLFEA-SU --+- NLFEA-SR

E E

E
o

1000

t
0

.S
I
1

'

'

]
4

i
5

Peak pressure (bar)

Fig. 12. Effect of in-plane stiffener restraint on peak displacement.

more readily be incorporated into a design procedure. As noted previously, one of the most common approaches used in design is the Biggs method, which idealises the structure into either a single or multi-series of springs and masses, approximating the stiffness and mass of the real system. Figure 14 shows a single degree of freedom idealisation together with the resistance function for the spring. The latter is assumed to be a bi-linear response typical of that used to represent an ideal elastic-plastic material such as steel. The model is assumed to undergo rebound when the maximum displacement is reached (zero velocity). To use this model meaningfully, it is necessary to determine the static resistance curve of the structure in order to obtain the static stiffness and maximum resistance of the panel including the effects of membrane action. Biggs provides tables which list these values for beams and slabs. The IGN allow stiff-

Non-linear analysis of blast walls and stiffened panels

109

--

NLFEA

-1 O0

~-- - 2 0 0 E E
0 0,.

.~

-300

-400

-500

L
25

'

I
50

'

I
75

L
100

T i m e (ms)

Fig. 13. Displacement-time response of tee-stiffened panel and deformed mesh showing tripping.

ened plates to be idealised as beams but an initial study carried out by the authors found that the resulting displacements were not comparable with the dynamic FE analysis. It was found that using static non-linear FE analysis to establish the static resistance curves substantially improved the correlations. In order to establish the minimum resistance of the blast wall, a number of failure modes would normally be examined. For the purpose of this exercise, where the effects of boundary conditions and validation of the Biggs method is being considered, the simplified model presented in section 5 has been adopted. This assumes that the stiffened panel between two transverse girders initiates failure. In order to establish the static resistance curve, the panel was loaded statically with uniform pressure, with the panel facing the pressure loading, placing

110
J J J J

L. A. Louca, M. Punjani, J. E. Harding

R m Resistance

-f

Deflection

Ym

'~ F(t)
Fig. 14. Idealised structure and resistance function.

the stiffener in tension. The resistance curve obtained is characteristic of a plastic section, i.e. a constant resistance with increasing displacement which can be approximated as a bilinear response. In order to account for the effect of strain rate, the static yield stress was increased by 20%, which is the maximum value recommended in the IGN. Figure 15 shows summary curves comparing the results of the Biggs analysis using the static non-linear curve with the transient non-linear finite element analysis for the three load curves of Fig. 6 with a loading duration of 100 ms and an increasing peak pressure. Up to 2 bar peak pressure the correlation is excellent, considering the simple nature of the method. At 3 bar the correlation is again good although for load curve 3 the Biggs model is beginning to give higher deflections than the FE results. At 4 bar the predictions show no correlation for any of the three load cases. This appears to be due to the widespread plasticity that occurs in the model with the higher peak pressures which is not reproduced by the Biggs representation. As noted previously, the transient non-linear finite element analysis indicates that at peak displacement the stiffener becomes plastic at mid span and part of the plate also becomes plastic. This suggests that the Biggs model is only accurate up to a point where a limited amount of plasticity occurs. Although the resistance curve has been assumed to be bilinear in the static non-linear finite element analysis, under dynamic loading the spring mass model will reach a peak displacement at some stage on the resistance curve as shown in Fig. 14. Beyond this point unloading occurs, which is elastic and parallel to the loading curve. This is generally known as rebound and is sometimes neglected as only the peak displacement is normally required. Figure 16

Non-linear analysis of blast walls and stiffened panels


2000 --+-- Load curve 1 - F E - 4 Load curve 1 - B i g g s -Load curve 2 - F E - - - Load curve 2 - B i g g s Load curve 3 - F E - - ~ - Load curve 3 - B i g g s

111

E E
p. 0

/ / / 1000/ I / / / / /

E
(D 0 1

0,.

/f
//

iI / / ] ~

I /I/!/
/

I /// ///

1.0

2.0

3.0

4.0

5.0

Peak pressure (bar) Fig. 15. Comparisonof Biggs model with simplifiedfinite element model. shows the full displacement-time history of the spring mass model for a peak pressure of 3 bar using load case 1, comparing the response both with rebound and without. Clearly the effect of the rebound is significant in determining the permanent deformation that results in the panel and a very good correlation with the FE result is obtained if rebound is included.

7 CONCLUSIONS Results based on non-linear finite element analysis and a simple single degree of freedom dynamic model have been presented for the response of a complex blast wall consisting of stiffened panels to hydrocarbon type explosion loading. The finite element study has highlighted the need for careful modelling

112
50

L. A. Louca, M. Punjani, J. E. Harding

40 m

30--

E E

20--

E
O tZ

10 m

t:3

- ....

NLFEA Biggs NL-wilh rebound

~j

t
-10 m

Biggs NL-no rebound

-20

I
25

I
50

I
75

I
100

Time (ms)

Fig. 16. Comparison of Biggs model with transient non-linear finite element analysis.

of boundary conditions as these can have a major impact on the ability of a panel to withstand the dynamic forces involved. The effect of restraint to the stiffener was most dramatic in that the overall buckling mode can change from a column mode to a tripping mode depending on the degree of restraint

Non-linear analysis of blast walls and stiffened panels

113

offered if the stiffener outstand is in compression. However, if adequate inplane support is provided and the surrounding structure can resist the membrane forces in the panel, very high pressures can be resisted. Comparison of the finite element results with a Biggs type model was very good provided that a realistic static resistance curve was adopted accounting for in-plane effects and that only limited plasticity occurred. This provides limited data in an area where few comparisons have been published against a full transient FE system.

REFERENCES 1. Cullen, Lord, The Public Inquiry into the Piper Alpha Disaster. HMSO, UK, 1990. 2. Interim Guidance Notes for the Design and Protection of Topside Structures against Explosion and Fire, SCI-P-112. The Steel Construction Institute, UK, 1992. 3. Biggs, J.. M., Introduction to Structural Dynamics. McGraw-Hill, New York, 1964. 4. Leach, P., Design charts for elasto-plastic design of cladding panels subject to dynamic loading. Proc. Inst. Civ. Engnrs Structs & Bldgs, 99 Feb. 1993, pp. 55-61. 5. Louca, L. A. & Harding, J. E., Torsional and local buckling of stiffener outstands. Behaviour of Offshore Structures Conference, MIT, USA, 3 (July, 1994). 6. Louca, L. A. & Harding, J. E., Torsional buckling of outstands in longitudinally stiffened panels. Thin-walled Structures, 24 (1996) 211-229. 7. Walker, A. C., Non-linear analysis of blast walls to evaluate safe ultimate loads. 2nd Int. Conf. on Offshore Structural Design Against Extreme Loads, ERA, London, 1993. 8. Schleyer, G. & Mihsein, M., Development of mathematical models for dynamic analysi,; of structures. 1st Int. Conf. on Structural Design Against Extreme Loads, ERA, I,ondon, 1992. 9. Dow, Ft. S. & Smith, C. S., Effects of localised imperfections on compressive strength of long rectangular plates. J. Construct. Steel Res., 4 (1984) pp. 51-76. 10. Blast and Fire Engineering for Topside Structures, Joint Industry Project, Steel Construction Institute, UK, February, 1991. 11. Houlston, R. & DesRochers, C. G., Non-linear structural response of ship panels subjected to air blast loading. Proc. 6th Conf. Non-Linear Analysis ADINA, 1987. 12. Van Wees, R. M., Design of stiffened panels to withstand explosion loads. 2nd Int. Conf. on Offshore Structural Design Against Extreme Loads, ERA, London, 1993. 13. British Standards Institution, BS 5950, Structural Use of Steelwork in Building, Part 1, 1990.

You might also like