Blind multiridge detection for automatic nondestructive testing using ultrasonic signals

摘要

Ultrasonic imaging has been a significant means for nondestructive testing (NDT). Recently the NDT techniques via the ultrasonic instrumentation have shown the striking capability of the quality control for the material fabrication industry. To the best of our knowledge, all existing signal processing methods require either the a priori information of the ultrasonic signature signals or the manual segmentation operation to achieve the reliable parameters that characterize the corresponding mechanical properties. In this paper, we first provide a general mathematical model for the ultrasonic signals collected by the pulse-echo sensors, then design a totally blind novel signal processing NDT technique relying on neither a prior/ signal information nor any manual effort. Based on the automatic selection of optimal frame sizes using a proposed new criterion in our scheme, the signature signal can be blindly extracted for further robust multiridge detection. The detected ridge information can be used to estimate the transmission and attenuation coefficients associated with any arbitrary material sample for the fabrication quality control