Nonlinear Feature Fusion Scheme Based on Kernel PCA for Machine Condition Monitoring

Feature fusion is an important approach in the field of fault diagnosis due to its ability to synthesize complementary information of different feature variables from multi-signal sources. However, the heavy computational burden induced by the tremendous size of the feature space is a tiresome problem. As most running statuses of machines are nonlinear and non-stationary, a nonlinear feature fusion scheme based on kernel principal component analysis (kernel PCA) is proposed to recognize the different fault patterns in running machines. Kernel PCA is applied to extract and fuse nonlinear features from acoustic signals and vibration signals. The computational problem is also effectively settled by using a kernel function in the input space without explicit computation of the mapping in feature space. The results show that the proposed scheme can greatly improve the robustness of feature extractor for mechanical faults.