Welding defect signal extraction technology based on GAMP-SBL algorithm

Abstract

Flip chip technology has been rapidly developed and widely used in the field of microelectronic packaging, and defect detection has also received more and more attention. Aiming at the problem that noise affects the location and extraction of signal defects in ultrasonic testing, the sparse Bayesian learning based on generalized approximate message passing (GAMP-SBL) algorithm is used to extract signal defects, and the over-complete Gabor dictionary is used to reconstruct signal defects to effectively improve Sparse decomposition algorithm. The precision experiment tested the defect simulation signal and the actual ultrasonic signal respectively, and compared with the greedy algorithm. The experimental results show that the GAMP-SBL algorithm can more effectively extract the defect signal under the noise background.