EMPIRICAL MODELING OF ULTRASONIC PHENOMENA

An exact treatment of ultrasonic phenomena is based on a quantitative description of measured displacement or velocity vector fields, but th e corresponding mathematical relation can be derived from the partial differential equations of elastodynamics only for idealized and rather simple examples. To avoid these limitations, we have investigated emp irical methods, which stem from either a non-parametric or a parametri c regression and they can be related to simulations of artificial neur al networks. The nonparametric approach is convenient for a general, n on-linear modeling while the parametric one is more suitable for linea r modeling. Here we summarize the associated procedures and describe t heir applicability with examples that include both active and passive ultrasonic phenomena.