Milling tool wear prediction using an integrated wireless multi-sensor tool holder and convolutional neural networks

In precision machining, it is necessary to predict tool wear and estimate the remaining useful life to produce high-quality products at a low cost. To meet this need, it is necessary to monitor the machining process by developing a multi-sensor dynamometer for data acquisition and a system for signal processing and tool wear estimation. The cutting force and vibration of the tool are two important parameters in tool condition monitoring. In this work, a new multi-sensor tool holder is designed, fabricated, and tested to simultaneously and wirelessly measure the torque, tri-axial vibration, and angular position of the tool. It has high rigidity, the ability to change the sensitivity, and the least interference for different working environments and provides data instantly in real-time. All sensors and data transmission systems are placed in the tool holder. In the present work, a new approach in the application of Convolutional Neural Networks is used for tool wear prediction. First, the Variational Mode Decomposition (VMD) parameters are optimized by the Sparrow Search Algorithm. VMD is a new technique for decomposing signals into intrinsic mode functions. Tool signals are then converted into images by using the Hilbert–Huang spectrum through optimized VMD. The sharp and worn cuttings tool shows different characteristics in the images. Finally, Dense Net is used for image classification and tool wear prediction. The results of the experiments showed that the designed system can quickly transmit sensor data and detect tool wear in a very short time with 93% prediction accuracy, which shows a suitable performance of the system.