Differential evolution-based feature selection and parameter optimisation for extreme learning machine in tool wear estimation

Cutting tool wear degrades the product quality in manufacturing processes. Hence, real-time online estimation of tool wear is important for suggesting a tool replacement before the wear limit is reached, in order to protect the workpiece and the CNC machine from damage and breakdown. In this study, using both statistical features and wavelet features extracted from sensor signals, an adaptive evolutionary extreme learning machine (ELM) learning paradigm is developed for tool wear estimation in high-speed milling process. In the proposed method, a discrete differential evolution (DE) algorithm is used to select input features for the ELM, and a continuous DE algorithm is used for parameter optimisation of the mixed kernel function for the ELM. The experimental results indicate that the proposed adaptive evolutionary ELM-based tool wear estimation model can effectively estimate the tool wear in high-speed milling process. Empirical comparisons show that the proposed model performs better than existing approaches in estimating the tool wear.