Diagnosis on Welding Defect Signals Using Sample Entropy

Abstract

Abstract: A new diagnostic method to identify alternating current field measurement (ACFM) signal based on sample entropy combined with wavelet packet feature is put forward in order to accurately evaluate the damage degree of welding defects. A butt welded tubular specimen with three kinds of different welding qualities was inspected and the corresponding ACFM signals were recorded by a commercial instrument. Subsequently sample entropies of the original signals and their wavelet package coefficients were computed respectively and compared via the bar graphs. The results show that the sample entropy successfully discriminates between different welding defects, and moreover it can be utilized to detect early or slight damage, demonstrating that the proposed approach is a promising and effective tool in characterizing the ACFM signals of welding defects.