Physics-Informed Machine Learning Model for In-process Estimation of Cutter Runout Parameters in End Milling

Cutter runout is inevitable in machining operations involving rotary cutting tools, such as end milling. The runout parameters are usually measured offline using a dial indicator or gap measurement sensor. Alternatively, measured cutting force data can also be used to determine these parameters using a non-linear optimization technique. It is an iterative approach sensitive to measured force data and could not determine parameters online. This work proposes utilizing a machine learning algorithm to extract cutter runout parameters from real-time cutting force data. Developing a machine learning model requires a large amount of experimental data over a wide range of conditions which is costly, time-consuming, and impractical. The present work generates synthetic datasets through a physics-based Mechanistic force model instead of performing milling experiments for the model training. The synthetic training datasets generated using the Mechanistic model minimize the effects of process disturbances and measurement outliers on the prediction abilities. A set of end milling experiments are carried out to compare the performance of the proposed approach with experimentally measured parameters and the non-linear optimization model presented in the literature. It has been shown that the proposed approach has robust prediction abilities and can be effectively employed for the online estimation of cutter runout parameters.