This calculation is made with the ReliabTbx-LCPC R.1.5 toolbox [7]. Fig.
2 gives the distribution
of the target reliability indexes given by the different designs. It can be noticed a rather narrow
variation and high values of these reliability indexes.
5. Calculation of a target reliability index
Conditional probabilities can be used for updating the reliability levels when new data become
available from laboratory or in-situ tests. Nevertheless, such probabilistic assessments usually
require specialist knowledge and expertise and are likely to be worthwhile only in exceptional cases.
To overcome such difficulties, especially for standard bridges, a semi-probabilistic assessment may
be easier to implement as far as guidelines for modifying the partial factors are proposed.
In this section, a general methodology is introduced to update (or recalibrate) the Eurocode's partial
factors by integrating information from inspections and investigations. These modified partial
factors must ensure the target reliability level as prescribed by the design standards [1].
5.1.1 Choice of partial factors to update
Not all partial safety factors deserve to be updated. This is only meaningful for partial factors
associated to variables that may be measured during inspections and investigations, or are the most
significant in the calculation of the reliability index. Let us note that some partial factors are
implicitly fixed to 1.00 and are therefore hidden in the calculation process. Consequently, if data are
available for the related variable, it may be necessary to introduce such a partial safety factor for
which the initial value in the design regulation is equal to 1.00. For the reinforced concrete slabs,
the partial factors of the yield stress f y , the reinforcement area As , the superstructure moment S
and the concrete density ρc are only considered for updating because the other variables have
smaller direction cosines and sensitivity measures than those issued from the other variables. Since
no partial factor on the steel area does not explicitly appear in the Eurocodes, a partial factor
γs =1.00 is conventionally introduced.
5.1.2 Partial safety factor updating
Theoretically, all the partial factors are affected when new data on a specific variable is accessible.
Nevertheless, since the set of partial factors is not unique, it is possible to only update the partial
factor related to this specific variable. When data from at least two variables are available, the
calibration becomes less amenable due to its numerical complexity and its applicability. Indeed, in
the case of reinforced concrete slabs, four variables are considered and all the combinations should
be analysed: this would require a large set of tables of partial factors for all these combinations.
The approach used in the present study is to simplify the problem by updating the individual partial
factors versus a range of mean values and standard deviations. This will lead to four tables: when
more than one variable is modified, it should be checked that the allowable reliability index
βa ({γˆ} ) for each bridge remains larger than the target reliability index β0 of the same bridge. At a
first stage the methodology is applied for each bridge. The modified partial factors for the family of
bridges is then obtained from a statistical analysis of all the individually modified partial factors, for
instance by choosing a specific fractile in order to eliminate "abnormal" case studies. In the present
paper, the open source ReliabTbx-LCPC R1.5 toolbox [7] (developed under Matlab© environment)
has been used for determining the updated partial safety factors.
5.1.3 Assessment rule
Let us assume the partial factor updating procedure to be achieved. For simplicity, let us consider
that in-situ measurements help to improve the knowledge about the bias ν S of the fixed load