Data-Driven Flank Wear Estimation for Single Point Cutting Tools using Regression Analysis

Abstract

Industrial revolution 4.0 is a phenomenon that has been characterized by massive machine-to-machine connections which have yielded the benefit of being able to collect operational data at unprecedented scales. In this regard, the authors propose a data-driven framework using which an approximate, yet reliable measure of the flank wear of a single point cutting tool can be made using a body of machine learning techniques broadly termed as Regression Analysis. For this study, the authors received ground truth operational data from a consulting industry which specializes in machining operations, which consists of flank wear being the dependent attribute, being a function of 3 independent attributes, namely - speed, depth of cut and feed rate respectively. Various regression models, including multiple linear regression, polynomial regression, Ridge regression, Lasso regression, decision tree regression, K-nearest neighbours regression and support vector regression, were trained on the dataset to approximate the target function associated with the problem. After experimentation, the authors were able to isolate the most appropriate shallow learning algorithm which fits the data optimally. It was found that the K-nearest neighbours regression strategy was the best-fitting out of all the algorithms which were used in the experimentation process. The work done by the authors is aimed at advancing research on real-time Tool Condition Monitoring (TCM) which is an important area in the broader field of study termed as Predictive Maintenance (PDM).