High power laser welding state recognition based on feature fusion

Under high power hybrid laser arc welding and high power CO2 laser welding, welding state recognition experiments based on pool image features, based on sound signal features and based on feature fusion of pool image features and sound signal features were carried out respectively, and recognition rates were compared. In the process of feature fusion experiment, principal component analysis (PCA) was used to reduce dimensionality, and cumulative contribution rate was set to determine the number of principal components. Recognition experiments were realized by probabilistic neural network (PNN). Experimental results show that welding state recognition based on feature fusion has higher rate, which is more reliable.