Multi-Sensor Analysis of Tool Wear with Integration of Direct and Indirect Sensors

Accurate analysis of tool wear is essential for the efficient tool utilization, enhancing productivity and minimizing machine downtime. Direct sensors, such as optical vision systems, provide precise measurements but are sensitive to alignment and environmental conditions. Indirect sensors, including cutting force and acoustic emission, offer complementary process information and can be integrated into machining environments for real-time monitoring. However, their signals are often unstable and strongly affected by vibration noise, limiting the performance of conventional signal-processing techniques. This study introduces a multi-sensor framework that combines direct and indirect measurements in an image-based workflow for tool wear characterization. Images of the cutting edge, captured by a camera, are processed with a U-Net segmentation model to analyse the wear. In parallel, cutting force and acoustic emission signals are transformed into image-based representations using Gramian Angular Field (GAF) mapping. Together, these data sources enable a unified approach to tool wear analysis. Deep convolutional neural network method is employed to extract discriminative features and improve robustness against measurement uncertainties. ... mehrWhile optical measurements provide precise wear characterization, they are difficult to integrate for continuous on-machine monitoring due to alignment sensitivity and time requirements. In contrast, indirect sensors can be seamlessly integrated into the machining environment. Experiments on milling processes show that GAF-based signal representations allow indirect sensors to operate more independently of optical monitoring while enabling in-process tool wear analysis in the presence of vibration disturbances. The proposed method supports real-time, on-machine condition monitoring and contributes to the development of intelligent manufacturing.