A modeling method for monitoring tool wear condition based on adaptive dynamic non-bias least square support vector machine

Online monitoring of tool wear condition is of great significance to manufacturing system. In order to adaptively adjust the length of sliding time window, dynamically update the prediction model, and ultimately improve the monitoring accuracy under limited number of available training samples, an adaptive dynamic non-bias least squares support vector machine (ADNLSSVM) is adopted to build tool condition prediction model with dynamic accelerating coefficients particle swarm optimization (DACPSO) algorithm applied to optimize parameters. The proposed method is tested by the open database dataset of milling process, and the experimental results validate the effectiveness of the designed model.