Milling Cutter Condition Monitoring Using Machine Learning Approach
A D Patange, Jegadeeshwaran R and N C Dhobale from School of Mechanical and Building Sciences, VIT University, India assess the applicability of ML approach for milling cutter fault diagnosis for reducing power consumption of drive of machine tool.
Introduction
Cutting tool health monitoring originated from necessity of developing unmanned and intelligent machining systems. One of the most important monitoring requirements in an unmanned manufacturing system is in-process detection of tool breakages and executing prompt action. ‘Tool Condition Monitoring or Health Monitoring Of Tool’ evolved for years. The methodologies are usually classified into two broad types such as Offline/Direct methods and Online/Indirect methods’. The direct methods are for inspecting and analysing complex failures (hard faults), usually unpredictable and hence unsuitable for machine learning domain.
Indirect methods are appropriate for monitoring soft faults which evolve gradually with respect to time leading to continuous deterioration of the tool. The recent development in Machine Learning (ML) and its applicability for condition monitoring approach has drawn researchers attention. ML examines existing and past indications to predict conditions in future.
In this paper, the vibration signals acquisition of 4 insert milling cutter is carried out with healthy and various fault conditions. The Visual Basic (VB) code and script is used to extract statistical features and decision tree algorithm is used to select relevant features. The different conditions of milling cutter are classified using tree family classifiers i.e. J48, Logistic Model Tree and Random Forest algorithms.
