Quality estimation of resistance spot welding based on kernel fisher discriminant analysis

In this research,the displacement signal of the resistance spot welding process is monitored and mapped into a 15×25 element bipolarized matrix by fuzzy theory method.Some welded spots with different welding currents,are classified into five classes according to the prototype pattern matrices.These pattern matrices,which represent different weld quality respectively,are used to construct the input data space.By using Kernel Fisher Discriminant Analysis(KFDA) technology,a nonlinear classification problem of the original data space can be converted to a linear one in feature space of high dimension.An effective pattern recognition system is developed,which realizes quality estimation of the welded spots.The results of cross-validation test show that the method utilizing pattern matrix of the electrode displacement signal to characterize formation process of nugget is effective.At the same time,the Fisher algorithm introduced kernel technology improves the recognition accuracy of the welded spot pattern under the small sample circumstance.