Title: Thermal signal-enhanced unscented Kalman filter for tool wear prediction

Authors: Xianzhe Fu; Zhaoyan Fan; Burak Sencer; Karl Haapala

Addresses: School of Mechanical Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA ' School of Mechanical Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA ' School of Mechanical Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA ' School of Mechanical Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA

Abstract: Tool wear is a gradual failure of cutting components generally caused by complex cutting conditions during the machining processes. This paper presented a thermal signal-enhanced tool wear prediction method to predict tool wear development during the cutting process. The smart tool samples the multi-modal sensor signals, including tool temperature, vibration, strain, and acoustic emission, which were processed through a neural network model and an unscented Kalman filter (UKF) to predict the tool wear size. Specifically, the measured tool temperature was considered as a tool wear status indicator to update the state transition function in the UKF and enhance the prediction accuracy with the presence of noises in the sensor signals. This method was validated on a commercial CNC lathe. The experimental results showed that the new method achieved an improvement of 8% in tool wear prediction accuracy compared to the conventional UKF.

Keywords: tool wear prediction; UKF; unscented Kalman filter; tool health; multi-sensor system; thermal signals; deep learning; feature extraction.

DOI: 10.1504/IJMMS.2024.143034

International Journal of Mechatronics and Manufacturing Systems, 2024 Vol.17 No.2, pp.117 - 131

Received: 31 Dec 2023
Accepted: 23 Mar 2024
Published online: 02 Dec 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article