Estimation of multiple unknown constructive internal parameters from broadband 'black box' models for matched contact piezoelectric probes

A deep knowledge of the piezoelectric probe parameters is mandatory for optimizing the responses in typical ultrasonic transceivers used for Non-Destructive Evaluation and Medical Imaging. But, usually, practical construction details are not known for the users of commercial probes. A prediction of these data is necessary during the simulation tasks intended to improve the final performances of these broadband transceivers. Here, a Genetic Algorithmic (GA) based option is aimed to make “a posteriori” estimations of five internal parameters, in broadband piezoelectric probes acoustically matched to radiation media, under realistic spike excitation. An original GA procedure is described for “inverse” prediction of internal data, using a matrix model for the matched ultrasonic probes. Up to five probe characteristics could be found with this approach. Agreement in the probe input admittance |Yin| curve was used as an index of good estimation. In spite of noisy transfer function measurements and distortions in experiments, good results were obtained in the parameters estimations, and in the experimental vs estimated comparisons. This proves that our GA procedure is a valuable tool to discover hidden design data in matched ultrasonic probes.