Welding quality identification based on bionic pattern recognition and sound information

Aiming at the quality identification of melt inert-gas (MIG) welding, a novel algorithm based on bionic pattern recognition and sound features is proposed in this paper. Firstly, the features of Mel-frequency cepstral coefficients (MFCCs), zero-crossing rate (ZCR), root mean square (RMS) and spectral flatness measure (SFM) are extracted from the arc sound signals, and the feature matrix is constructed. Subsequently, the objective function is constructed in Clifford algebraic space to represent the distance between the two feature matrices. Finally, optimized by using the distance of two feature matrices as a criterion, the bionic pattern recognition theory is used to identify the welding quality. The experimental results indicate that the proposed algorithm can accurately identify the samples obtained from defective welding conditions with few reference samples, which has provided a new method and idea for welding quality identification of MIG welding. Furthermore, it also has profound theoretical research significance and extensive practical application value.