Академический Документы
Профессиональный Документы
Культура Документы
INTRODUCTION
FAILURE MODES
In designing structures to resist explosion loading, decisions are
made under a great deal of uncertainty that may lead to a finite In a multi-level explosion resistant design [Yasseri and
risk of exceeding limit states of the structures. Classically, in Menhennett] a designer may pursue several goals. For instance
order to minimize the risks, conventional factors of safety based for explosion with higher frequency of occurrence it may be
on deterministic analyses are commonly used in the design. desirable to ensure the continuity of operation in addition to the
life safety, and for a higher overpressure, but less frequent
In common deterministic design procedures for explosion explosion, the goal is the life safety only. Thus the basic aim for
loading design parameters are set rather arbitrarily. Thus, for explosion resistant design is to contain the explosion, limit the
explosion resistant designs [SCI], it is not immediately obvious damage and finally to prevent progressive/catastrophic collapse.
how such choices relate to the required reliability level for the Therefore, depending on the goal, some level of inelastic
structural performance. In this study displacement and rupture deformation may be tolerated, provided rupture or the total loss
limit states are used for measuring performance. is not imminent. Thus, the acceptable behavior very much
depends on the targeted goals.
The application of probabilistic analysis in structural
engineering is still an emerging technology. Full experience In general, the design requirements to satisfy such goals are:
with such procedures remains to be gained. The reliability
methods are too complex for any reasonable size structure. The Deformations should be limited so that the target goals are
methods described herein should not be expected to provide
403
Paper No. 2006- JSC- 138 Yasseri Page 1
achieved. The acceptable transient and permanent deformation system or component. The capacity and demand can be
is obviously dependent on design goals (in a multi-level design) combined into a single function (the performance function), and
and on the layout and proximity of other structures so that vital the event that the capacity equals the demand taken as the limit
pipework and equipment can function (albeit in some damaged state. Reliability is the probability that the limit state will not be
state) and escalation is also avoided. reached or crossed.
Rupture of large parts of the structure is not acceptable, since it The performance measure is taken as the displacement
could lead to progressive collapse. This is only relevant to the normalized with acceptable level of displacement. Equivalent
higher level explosion when designing in a multi-level design performance functions and limit states can be defined using
format. For lower level of explosion only minimal damage is other measures, such as the plastic hinge rotation or the
allowed. The maximum total plastic strains are used to judge maximum plastic strain. However, deformation and the plastic
rupture. Thus, the resulting strains associated with acceptable strains are strongly correlated, namely both maximize at the
displacements should not cause brittle failure or ductile tearing same time.
[UMIST and TWI]. However, rupture may occur before the
allowable displacement is exceeded. The likelihood that a structure will meet or exceed a specified
level of damage for a given level of overpressure is:
For topside structures with Passive Fire Protection (PFP), if the Condional Probablity of Failure = P[R > LS | IM = y ] (2)
insulation is required to be in-place after the explosion, then a
lower limit of deformation or strain needs to be allowed so to
where R is the response measure of the system response or the
prevent large scale loss of PFP. Such limits should be
demand of explosion on the structure, LS is the limit state or
ascertained by test to check the de-bonding effect of large
damage level, IM the explosion intensity measure and y is the
deformations and strains on a PFP material. Past experience
shows that for strains up to 5%, the adhesion of the epoxy-based realization of the chosen explosion intensity measure. This
fire proofing material is not badly affected. For fire proofing probability of failure is represented by:
which is wrapped around the structural members, larger ⎡ E (R ) ⎤
Pf = P ⎢ ≥ 1⎥ (3)
deformations and strains may be acceptable. ⎣ E (C ) ⎦
where Pf the probability of is exceeding a specific damage-
THE CONDITIONAL PROBABLITY OF FAILURE
state, E (R ) is the expectation of explosion demand on the
For an existing structure subjected to an explosion, the structure and E (C ) is the structural capacity or damage state.
probability of failure Pf can be expressed as a function of the
explosion overpressure and other factors including duration, Assuming a lognormal distribution
structural strength, weld strength & quality, general ⎡ ⎤
ln (E (R ) E (C )) ⎥
workmanship, etc. This study will focus on developing the Pf = Φ ⎢ = Φ(Z ) (4)
conditional probability of failure function for the explosion ⎢ V 2
+ V 2 ⎥
⎣ R C ⎦
overpressure, which will be constructed using engineering
estimates of the probability functions or moments of the other Here E (C ) is the median value of the structural capacity defined
relevant variables. for the damage state, Vc is the dispersion or lognormal standard
deviation of the structural capacity, E (R ) is the explosion
The conditional probability of failure can be written as:
demand as a function of a chosen explosion intensity parameter,
Pf = P( failure OP ) = f (OP, X 1 , X 2 ,⋅ ⋅ ⋅, X n ) (1) VR is the logarithmic standard deviation for the demand and
In the above expression, the first term (denoting probability of Φ (∗) is the standard normal distribution function.
failure) will be used as a shorthand version of the second term.
In the second term, the symbol “|” is read given and the variable
Generally, maximum deformation is the controlling failure
OP is the Overpressure. In the third term, the random variables
mode, and it should not exceed a specified value. There are
X 1 through X n denote relevant parameters such as material cases where extensive deformations can be tolerated. In a
strength, explosion duration, mass, damping, section sizes etc. deterministic non-linear dynamic analysis, output is in the form
Equation (1) can be restated as follows: “The probability of of deformation, as well as other parameter defining the state of
failure, given an overpressure, is a function of the overpressure, the structure. In this work the maximum deformation
its duration and other random variables.” experienced by the system is used as a measure of performance
and it is normalized using the maximum acceptable
RELIABILIY INDEX deformation, and this ratio is termed R . According to our
definition R is acting as a utilization ratio, and E (C ) act as the
In the capacity-demand model, the probability of failure or
maximum available capacity with a maximum value of 1. In this
unsatisfactory performance is defined as the probability that the
study we use 0.85 so that the probability of failure is reduced.
demand on a system or component exceeds the capacity of the
200
Mean duration (ms)
100
The capacity-demand model, described earlier requires that the t = 190e
-0.205(OP)
engineer assign values for the probabilistic moments of the 50
random variables considered in analyses. This section reviews
0
information regarding the observed variability of design
0 2 4 6 8 10
parameters which can be used as a guide when characterizing
Overpressure (barg)
random variables for the analysis.
Four random variables are considered, the material strength, When the variance components are summed, the total variance
blast duration, damping and the plate thickness. The assigned will be 1.09546875. Taking the square root of the variance gives
probabilistic moments for these variables are given in Table 4. the standard deviation of 1.046646.
The maximum acceptable deformation anywhere on the wall is
120 mm. The Utilization Ratio is assumed to be log-normally distributed
random variable with the expected vale (first moment)
A finite element model of the wall was built for ABAQUS. E (R ) = 1.375 and σ R = 1.046646 . Using the properties of the
Parameters which influence the response are listed in table 5. As lognormal distribution, the equivalent normally distributed
Table 5 indicates four parameters were selected as the random random variable has the following parameters:
variable, thus for each level of blast pressure nine analyses were 1
performed. E (ln R ) = ln E [R ] − σ ln2 R = 0.089924
2
For the first analysis (RUN1), the four random variables are and
taken at their expected values. The utilization factor is 1.375 σ ln R = ln[1 + V R2 ] = 0.761197
(165/120).
The maximum allowable E(C) is assumed to be 0.85.
Results for ABAQUS non-linear dynamic analyses for 1.4-bar
overpressure are summarized in Table 5.
For the second, third and fourth analyses, one variable is taken
at a time and its value is assumed to be the expected value plus
0.60
0.40
Wiley, 1999.
0.20
Morgan, G.M., and Henrion, M., Uncertainty-A Guide to
0.00
Dealing with Uncertainty in Quantitative Risk and Policy
0.6 0.8 1 1.2 1.4 1.6 Analysis, Cambridge University Press, New York 1990.
Overpressure (bar) SCI, Guidance Note
UMIST and TWI, The Effects of Dynamic Loading on Structural
Figure 7: Conditional probability of failure Integrity Assessments, HSE Research Report 208, 2004.
Vose, D, Quantitative Risk Analysis: Guide to Monte Carlo
Simulation Modeling, John Wiley, 1999.
Yasseri, S. and Menhennett, P., “A methodology for
Performance-Based Explosion Resistant Design”, ERA Major
CONCLUSIONS Accident Offshore 2003, paper 1.3.
Yasseri, S. “Probabilistic Damage Analysis of Offshore
The conditional probability of failure was determined Installations”, 2nd ASRANet International Colloquium, Spain,
for a given level of explosion. This curve can be used June 2004.
to carry out detailed ALARP assessments of essential Yasseri, S. and Prager, J, “Explosion Recurrence Modeling”,
OMAE 2005, Paper 51048, Vancouver, Canada.
plant items, and structures that provide protection
against explosion-induced loss of essential safety
functionality. This approach identifies the load
8 thk
200x150x18R
SA
225 50
315