An effective LSSVM-based approach for milling tool wear prediction

In order to realize real-time and precise monitoring of the tool wear in milling process, this paper presents a predictive model based on stacked multilayer denoising autoencoders (SMDAE) technique, particle swarm optimization with an adaptive learning strategy (PSO-ALS), least squares support vector machine (LSSVM). Cutting force vibration information are adopted as signals. Three steps make up unique feature extraction fusion method: multi-domain features extraction, principal component analysis (PCA)-based dimension reduction, SMDAE-based increment. As novel representation approach, SMDAE technique is utilized fuse PCA-based enrich effective by increasing dimension, thus helping polish performance proposed model. PSO-ALS used obtain optimal parameters for LSSVM, simplifying problem population diversity. Twelve sets experiments conducted demonstrate reliable The experimental results show that presented superior models such PSO-LSSVM performance, effectively improves prediction accuracy established findings offer theoretical guidelines real industrial situations.