Cutting force estimation from machine learning and physics-inspired data-driven models utilizing accelerometer measurements
Greg Vogl, Yongzhi Qu, Reese Eischens, Gregory Corson, Tony L. Schmitz, Andrew Honeycutt, Jaydeep Karandikar, Scott Smith
Abstract
Monitoring cutting forces for process control may be challenging because force measurements typically require invasive instrumentation. To remedy this situation, two new methods were recently developed to estimate cutting forces in real time based on the use of on-machine accelerometer measurements. One method uses machine learning, while another uses a physics-inspired data-driven approach, to generate a model that estimates cutting forces from on-machine accelerations. The estimated forces from both approaches were compared against cutting force data collected during various milling operations on several machine tools. The results reveal the advantages and disadvantages of each model to estimate real-time cutting forces.