Dimensionality reduce-based for remaining useful life prediction of machining tools with multisensor fusion

Abstract

The remaining useful life (RUL) prediction has received increasing research attention in recent years due to its essential role in improving industrial manufacturing systems' productivity and reliability. High dimensionality time-series data is collected during the system's operating time with the implementation of various sensors. To monitor the machining tool's RUL, a novel multisensor fusion method based on t-SNE-DBSCAN dimensionality reduction is proposed in this paper to aggregate the multiple sensor measurements into a healthy indication that represents the RUL. The robust MCD-Estimator is used to denoise multidimensional sensor data, flag the outliers, and enhance the robustness of the denoising process. The high dimensional statistical features are extracted, and the dimensionality is reduced with t-distributed Stochastic Neighbor Embedding (t-SNE) according to the RUL labels, and optimum features selected by Density-Based Spatial Clustering of Application with Noise algorithm (DBSCAN). Ultimately, Long Short-Term Memory (LSTM) is adopted for RUL estimation with multisensor fusion. A case study with a practical application of the proposed approach, which can be demonstrated, is compared with state-of-the-art methods by evaluating its performance and migration capability in different working conditions.