Removal of Dominant Reverberations from Ultrasonic Time-Records of Layered Material by Digital Predictive Deconvolution

Abstract

The interest in non-destructive evaluation of the space shuttle Solid Rocket Motor (SRM) calls for inspection of the integrity of the structure, mostly from the external side only. Difficulties however are encountered as the acoustic impedances combination of the various layers are such that the external steel layer generates strong multiple reflections that virtually mask everything that happens below. A method is presented to separate the steel reflection component from the deeper layers. Based on the fact that the reverberations are a predicted event, directly correlated to the first reflection, while variations deep in the material are pseudo-random, the predictive deconvolution estimates the next signal value based on previous signal values, over the prediction distance. The optimized operator is used to remove the predicted reverberations so that the residue that represents events such as defects in the deeper layers can be extracted and investigated.