Heterogeneous sensors-based feature optimisation and deep learning for tool wear prediction

During machining processes, accurate prediction of cutting tool wear is prominent to prevent ineffective utilisation and significant resource waste. Tool conditions progression involve complex physical mechanisms, a promising approach deploy heterogeneous sensors design deep learning algorithm conduct real-time monitoring precious prediction. To tackle the challenge algorithms in processing signals from sensors, this paper, systematic methodology designed combine signal de-noising, feature extraction, optimisation learning-based In more details, comprised following three steps: (i) de-noising carried out by Hampel filter-based method eradicate random spikes outliers for raw data quality enhancement; (ii) features extracted time frequency domains are optimised using recursive elimination cross-validation (RFECV)-based Isomap-based methods; (iii) convolutional neural networks (CNN) devised process implement case study showed that 80% were reduced originally 86% accuracy was achieved based on developed methodology. The presented benchmarked with several main-stream methodologies, superior performance over those comparative methodologies terms exhibited.