Data-Driven Tool Remaining Useful Life Prediction in Ceramic Grinding: A Multisensor Machine Learning Approach

Diamond grinding tools are suitable for machining alumina ceramics, which are hard and brittle, because these tools have high hardness and wear resistance. Accurately predicting tool life is essential for enhancing production efficiency and reducing maintenance costs. Traditional methods for estimating tool wear often involve evaluating the surface properties of the grinding tool, such as the presence of cracks, surface roughness, and loss of abrasive particles. Alternatively, tool life can be assessed using process-related signals, such as acoustic emissions, tool temperature, spindle speed, feed rate, and grinding depth. This study analyzed grinding tool wear by collecting three-axis accelerometer data in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i>-direction and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">y</i>-direction and three-phase spindle current signals during the grinding of alumina ceramics. With continual grinding, the tool surface experiences increasing wear and cracking, which leads to greater vibration instability and fluctuations in frequency and acceleration. These dynamic variations serve as valuable indicators for predicting tool wear and estimating tool life. This study integrated machine learning and neural network models to analyze sensor data collected on tool wear. This analysis revealed complex variable relationships, enabling the development of a robust model for tool wear prediction.