| Model Uncertainty | |||
Model uncertainty in this type of problems is coming from two sources. First, the number of basic physical variables has been limited to a finite number n leaving out possibly an infinite set of parameters that in the model idealization process have been judged to be of secondary or negligible importance for the problem at hand. Second, model uncertainty is due to the mathematical idealization (by example: the use of a linear elastic finite element, linearised limit states, . . . ). Besides these two simplification causes, there is lack of knowledge about the detailed interaction between the different masonry properties. There are three ways in which model uncertainty can be described (table 1): introducing an extra basic variable (1) - proposed in ISO 2394: "General principles for reliability analysis of structures" -, multiplying the model with an unknown coefficient, to be determined from test results (2) - proposed in EC 1 - or introducing a random vector field (3). In these cases, the model uncertainty is typically assumed normal (1) or log-normal (2) distributed.
Model uncertainty can only be quantified either by comparison with other more involved models that exhibit a closer representation of nature, or by comparison with collected data from the field or the laboratory. Research is focused on developing a pragmatic way to incorporate the model uncertainty in the limit state functions, and to estimate proper values for the parameters , by means of comparison with test results. |
|||
|
|