Model selection via worst-case criterion for nonlinear bounded-error estimation

In this paper the problem of model selection for measurement purposes is studied. A new selection procedure in a deterministic framework is proposed. The problem of nonlinear bounded-error estimation is viewed as a set inversion procedure. As each candidate model structure leads to a specific set of admissible values of the measurement vector the worst-case criterion is used to select the optimal model. The selection procedure is applied to a real measurement problem: grooves dimensioning using remote field eddy current inspection.