A1305
Title: Optimal transport-based domain adaptation for sensor data with application in smart manufacturing
Authors: Rui Xie - Universify of Central Florida (United States) [presenting]
Dazhong Wu - University of Central Florida (United States)
Abstract: Tool wear prediction plays a crucial role in smart manufacturing, where advanced technologies and data-driven insights are employed to optimize production processes and minimize downtime caused by tool failures. In recent years, numerous machine learning-based predictive modeling approaches have emerged for tool wear prediction. However, accurately predicting tool wear under varying operating conditions, such as depth of cut, feed rate, and workpiece material, remains a daunting task due to the intricate nature of tool wear mechanisms. To tackle this challenge head-on, an algorithm leveraging optimal transport (OT)-based domain adaptation has been developed. This innovative approach enables the transfer of knowledge on tool wear from one operating condition to another. The effectiveness of the OT-based transfer learning model has been verified using a limited dataset encompassing diverse operating conditions. Through rigorous experimentation, the results have demonstrated a substantial improvement in tool wear prediction accuracy achieved by the OT-based transfer learning method.