For the practical applications of probabilistic reliability methods it is important to make decisions about the target reliability level. Presently calibration to existing design practice seems to be the only practicable and politically reasonable solution to this decision problem. However, several difficulties of ambiguity and definition show up when attempting to make the transition from a given authorized partial safety factor code to a superior probabilistic code. For any chosen probabilistic code format there is a considerable variation of the reliability level over the set of structures defined by the partial safety factor code. Thus, there is a problem about which of these levels to choose as target level. Moreover, if two different probabilistic code formats are considered, then a constant reliability level in the one code does not go together with a constant reliability level in the other code. The last problem must be accepted as the state of the matter and it seems that it can only be solved pragmatically by standardizing a specific code format as reference format for constant reliability. By an example this paper illustrates that a presently valid partial safety factor code imposes a quite considerable variation of the reliability measure as defined by a specific probabilistic code format. Decision theoretical principles are applied to get guidance about which of these different reliability levels of existing practice to choose as target reliability level. Moreover, it is shown that the chosen probabilistic code format has not only strong influence on the formal reliability measure, but also on the formal cost of failure to be associated if a design made to the target reliability level is considered to be optimal. In fact, the formal cost of failure can be different by several orders of size for two different, but by and large equally justifiable probabilistic code formats. Thus, the consequence is that a code format based on decision theoretical concepts and formulated as an extension of a probabilistic code format must specify formal values to be used as costs of failure. A principle of prudence is suggested for guiding the choice of the reference probabilistic code format for constant reliability. In the author's opinion there is an urgent need for establishing a standard probabilistic reliability code. This paper presents some considerations that may be debatable, but nevertheless point at a systematic way to choose such a code.Keywords: Code calibration, Structural reliability, Decision analysis, Reliability index, Partial safety factors, Target reliability.
|Publication status||Published - 1997|