Experimental Designs for an Experimental Modelling of Material or Structural Tests
Résumé
The experimental study of material or structural behaviours often consists of a multiparametric analysis. To describe the physical phenomenon in a limited region, it is valuable to identify the coefficients of a multilinear model with a regression. But the precision on these coefficients depends on the arrangement of the test matrix which represents the parameter values taken for each test. The optimal matrix form is reached by an orthogonal test matrix, but these orthogonal designs are restricting and sometimes impossible to realise because of physical constraints. Nevertheless, according to a criterion, optimal designs can be defined by respecting certain constraints: D-optimal designs are presented.