Proposed Machine Learning Model Using Levenberg-Marquardt Algorithm to Predict the Remaining Useful Life of cutting tools by monitoring sound or temperature measurements

Abstract

The story of industrial optimization centers around advances in reducing waste generation and lowering costs. Any industrial institution needs to constantly optimize their production processes in order to gain competitive advantage over its competitors in the industry. One of the challenges manufacturing is to monitor and minimize the gradual failure of cutting tools. The remaining useful life (RUL) of a cutting tool must be used carefully to ensure precision of surface finish, since tool wear can cause damage to cutting tool and scraping machined.

This paper presents a research project to monitor and optimize the life of the cutting tool during turning process. Machine tool data was collected from sets of experiments to estimate parameters of the modified Taylor’s equation using the Levenberg-Marquardt (LM) non-linear least squares algorithm. The LM nonlinear results are used as data structure for a proposed machine learning model to alarm the factory to replace tool before reaching the end of life. In this study, the LM nonlinear estimation results were compared to linear least squares solutions of the linearized form of the extended Taylor equation. The LM nonlinear least squares model showed better fitting results. In this study, the cutting tool temperature is also recommended as other techniques to teach machine to monitor the RUL cutting tools.