Discriminant Neighborhood Preserving Embedding with Autoencoder for fault diagnosis

With the development of modern industry, fault diagnosis has become a hot research area. The Discriminant Neighborhood Preserving Embedding (DNPE) algorithm has demonstrated excellent performance in handling complex industrial data. However, DNPE only preserves the local structure and discriminative information after dimensionality reduction, without the ability to recover the low-dimensional embedded data back to its original form. This lack of recovery capability may lead to the loss of important information in the data. Inspired by the reconstruction mechanism of autoencoders, a new NPE algorithm, DNPE-AE (Discriminant Neighborhood Preserving Embedding with Autoencoder), is proposed. Experiments on the bearing fault dataset provided by Case Western Reserve University (CWRU), show that DNPE-AE outperforms other methods in fault recognition and classification.