Professional Documents
Culture Documents
1 Introduction
2 Target reliabilities for limit state design
Structural fire design defines the performance of the struc-
tural system in case of fire. This performance has many di- 2.1 ALARP and cost-optimization
mensions, and relates e.g. to lives lost, extend of non-availa-
bility (downtime) in the wake of the fire event, direct material The societal ALARP requirement applies once tolerability
losses to the building, loss of cultural value, loss of network has been demonstrated for all relevant performance
functionality, and pollutants released into the environment. measures. While this is commonly done for each perfor-
For a given performance dimension of interest, the adequacy mance measure in isolation, the tolerability of the design is
of the design is conceptually determined by its position on more reasonably determined by the interaction, i.e. joint oc-
the corresponding Frequency-Consequence diagram, Fig 1 currence, of different measures.
[1]. If the design meets the tolerability prerequisite, its ade-
quacy is determined by the ALARP criterion, requiring that The ALARP assessment itself should be considered as a cost-
efficient safety measures are implemented. benefit assessment (CBA), where the initial safety invest-
ment and maintenance costs are balanced against the present
Full evaluation of the ALARP acceptability requires a (costly net value (PNV) of risk reduction [1], in principle consider-
and complicated) project-specific assessments of the societal ing all performance measures simultaneously. As discussed
costs and benefits of investing in structural safety. Therefore, in [4][5][6], the societal CBA entails the maximization of the
the ALARP assessment is often substituted by the specifica- lifetime utility Y, specified by Eq. (1), with B the benefit de-
tion of a target reliability index, e.g. the target reliability in- rived from the structure’s existence, C the construction and
dices listed in EN 1990:2002 [2] for normal design condi- safety investment cost (including discounted maintenance),
tions. Appropriate target safety levels for fire exposure are A the obsolescence cost, and D damage costs associated with
however not clearly defined [3]. The absence of targets for the performance objectives. Commonly, structural optimiza-
structural fire design is even more pronounced when looking tion is performed considering a renewal process, assuming
to aspects of resilience (e.g. downtime and loss of function- continued need for the structure’s functions [5]. When con-
ality). In the following, the specification of reliability targets sidering the benefit B as independent of the safety investment
as a proxy for the ALARP evaluation is revisited for struc- scheme (i.e. when the time between failures is large relative
tural fire design, and the concept of a resilience target ex- to the eventual reconstruction time), the maximization of Y
plored. corresponds with the minimization of the lifetime costs K.
1|P age
International Probabilistic Workshop 2019
2|P age
International Probabilistic Workshop 2019
Table 2 Optimum reliability index β. Value obtained from Eq. (8) with VR = VE = 0.2 /
ω γ 1 1
value listed in ISO 2394:2015 [9] / value resulting from Eq. (12) with b = 0.2 Pf , opt = b 1 + = = (12)
Consequences of failure γ ξλ DII λ DII 0
High (ε = 0.1) 3.4 / 3.1 / 3.4 3.5 / 3.3 / 3.5 3.7 / 3.7 / 3.7
Moderate (ε = 0.01) 3.9 / 3.7 / 3.9 4.1 / 4.2 / 4.1 4.3 / 4.4 / 4.3
Low (ε = 0.001) 4.5 / 4.2 / 4.5 4.6 / 4.4 / 4.6 4.7 / 4.7 / 4.8
- The costs of failure and of improving safety may differ. Eq. (12) specifies the optimum failure probability irrespec-
- The assumption of lognormality may not be appropriate. tive of distributional assumptions. The crucial assumption for
- The uncertainty (i.e. VR) may be larger in case of fire. the validity of this formulation is the safety cost specification
- The effect of fire-induced restraining forces is not clear. of Eq. (10), which Fig 2 shows to be compatible with Rack-
witz [6]. Optimum failure probabilities and corresponding re-
liability indices are visualized in Fig 3, for DII ≥ 5 (lower DII
Van Coile et al. [3] further observe that the optimum (target)
result in Pf > 0.2). Specific values are listed in Table 2 (third
failure probability defined by Eq. (8) is in principle not in-
entry in each cel; b ≈ 0.2 for VR = VE = 0.2)
versely proportional to λ. This is contrary to the normal de-
sign target reliability index scaling applied in the Natural Fire
Safety Concept (NFSC) [11], i.e. (9), with Pf,t,fi the target fail-
ure probability given the occurrence of a significant fire, pfi
the significant fire occurrence rate, and Pf,t,EN the target fail-
ure probability for normal design according to EN 1990 [2].
Pf ,t ,EN Φ ( − β t ,EN )
Pf ,t , fi = = (9)
p fi p fi
(
C1 = ε C 0 c − d ln ( Pf ) ) = C ( a − b ln ( P )
under the (not likely) condition however that DII0 is the same
0 f (10) for the fire design situation as for normal design situations.
) ) 1 + ωγ + ξλγ P
(
min 1 + a − b ln ( Pf
f (11) λ fi Pf ,t , fi =
1 ?
=
1
DII 0, fi DII 0, EN
= λEN Pf ,t , EN (13)
Pf
3|P age
International Probabilistic Workshop 2019
Fig 4 gives a conceptual representation of a possible down- Fig 5 Event tree of direct and downtime losses incurred in case of an adverse event
time loss percentage τ in function of tD. The upper part of the
graph represents the CDF of downtime (i.e. a fragility curve
3.2 Lifetime cost optimization
given an adverse event), while the lower part visualizes τ. Fig
4 suggests that 3 ranges of downtime loss can be discerned:
Representing the cost model of Fig 5 in the lifetime cost Eqs.
an area of minimal loss associated with practical inconven-
(2)-(4), results in (15) for the total expected cost. Further con-
ience (area I), an area with considerable losses associated
sidering the safety investment cost C1 as a function of both
with a temporary relocation (area II), and an area of large
Pf,c and Pf,r (the decision variables), the optimum design is
downtime losses corresponding with a permanent relocation
specified by the set of equations (16)-(17).
to another building in the vicinity (area III). This subdivision
is included here for illustrative purposes only. Limiting com- E[K ] C ω
plexity, a single limit state is defined by Z = tDlim – tD, repre- = 1 + 1 1 +
senting a transition between situations where the downtime C0 C0 γ (15)
λ
(
+ Pf ,c (ξC + ξ D ) + (1 − Pf ,c ) Pf ,r (τ LS ξ D ) )
losses are considered negligible, to situations where down-
time losses are considerable. The losses incurred when ex- γ
ceeding tDlim can be represented as τLS·ξD, where τLS is calcu-
∂ 1
lated as the expected loss quantile when exceeding the limit C
state, i.e. Eq. (14). Together with the model of losses incurred C0 1 + ω
in case of structural collapse, the limit state loss model can
∂Pf ,c γ ξC + ξ D (16)
be represented by the flowchart in Fig 5, where Pf,r is the + = Pf ,r ,opt
λ τ LS ξ D
(τ ξ )
probability of exceeding the downtime limit state, condi-
tional on no structural collapse.
γ LS D
t Dmax
∂ 1
C
∫ τ (t ) f D tD dt D
C0 1 + ω
τ LS = ∂Pf ,r γ
t Dlim
t Dmax (14) (17)
+ 1 = Pf ,c ,opt
∫ f tD dt D λ
τ ξ
t Dlim
γ LS D
4|P age
International Probabilistic Workshop 2019
ω
−c 1 +
1 γ +1 = P (19)
Pf ,r ,opt λ f ,c ,opt
τ ξ
γ LS D
1
Pf ,r ,opt =
(1 − P
f ,c , opt ) DII r
(20)
5|P age
International Probabilistic Workshop 2019
Eqs. (25) and (26) are in essence identical to Eq. (12) which [3] VAN COILE, R.; HOPKIN, D.; BISBY, L.; CASPEELE, R.:
has been evaluated in some depth in 2.3 above, resulting in The meaning of Beta: background and applicability of the
the optimum reliability indices listed as third entries in Table target reliability index for normal conditions to structural
2. Note however that Eq. (25) in conjunction with (22) clearly fire engineering. Procedia Engineering (2017), 210: 528-36.
indicates that optimum investment in collapse prevention
should be assessed taking into account the downtime (i.e. ξD) [4] VAN COILE, R.; JOMAAS, G.; BISBY, L.: Defining ALARP
incurred in case of structural failure. In other words, invest- for fire safety engineering design via the Life Quality Index.
ment in collapse prevention should be done considering the Fire Safety Journal (2019).
full range of consequences.
[5] FISCHER, K.: Societal Decision-Making for Optimal Fire
The optimum failure probability for the downtime limit state Safety. Doctoral dissertation (2014), ETH Zurich, CH.
as specified by Eq. (26) specifies optimum resilience targets
in function of DIIr (Table 3). By investigating DIIr values for [6] RACKWITZ, R.: Optimization – the basis of code-making
different types of structures (taking into account differences and reliability verification. Struct Saf (2000), 22, 27-60.
in function, location, etc.), optimum ‘resilience reliability in-
dices’ can be linked to structure types as target values. While [7] LANGE, D.; DEVANEY, S.; USMANI, A.: An application of
derived from very simplified considerations, these target val- the PEER performance based earthquake engineering frame-
ues can inform design rules for new structures and assess- work to structures in fire. Eng Struct (2014), 66, 100-15.
ment methodologies for existing structures, similar to the tar-
get reliability indices for normal design conditions which [8] ISO 2394:1998.: General principles on reliability of
have been derived from similar considerations. structures. International Standard.
Table 3 Optimum resilience target (reliability index) in function of DIIr [9] ISO 2394:2015.: General principles on reliability of
DIIr [-] 2 5 10 102 103 104 105 106 structures. International Standard.
βr,opt [-] 0 0.8 1.3 2.3 3.1 3.7 4.3 4.8 [10]JCSS.: Probabilistic Model Code: Part 1 – Basis of De-
sign. The Joint Committee on Structural Safety (2001).