Fault diagnosis method based on incremental enhanced supervised locally linear embedding and adaptive nearest neighbor classifier

A novel fault diagnosis method based on incremental enhanced supervised locally linear embedding (I-ESLLE) and adaptive nearest neighbor classifier (ANNC) is proposed to improve the accuracy of machinery fault diagnosis. Firstly, I-ESLLE is proposed for the non-linear dimensionality reduction of high-dimensional fault samples obtained from vibration signals. I-ESLLE can not only acquire the low-dimensional intrinsic manifold structure embedded in the high-dimensional input space, but also can deal with new fault samples in an iterative and batch model. Then, the low-dimensional fault samples are fed into the proposed ANNC for fault type identification. ANNC exploits “representation-based distance” to select the nearest training samples of new fault sample and identifies fault type in a weighting strategy. Moreover, the number of nearest training samples of each new fault sample is adaptively determined according to the density of the local distribution of the new fault sample. To verify the validity of the proposed fault diagnosis method, a fault diagnosis experiment of gearbox is performed, and the results indicate that the proposed fault diagnosis method outperforms the traditional methods and achieves higher diagnostic accuracy.