Abstract This paper studies the application of kernel principal component analysis (KPCA) and support vector machines (SVM) for source location of the acoustic emission signals of cracks. The results show that the accuracy of location using the feature parameters using KPCA technique is improved compared with raw parameters. That is, the recognition rate of crack region is 100 percent and the maximum error of support vector regression for distance from source of cracks to welding seam is 20cm when the number of the input feature parameters is nine. As a result, it is a good method for source location of complex big-size structures to combine KPCA with SVM. It decreases the dimension of input signals and improves the accuracy of location as well.