Utilising Machine Learning for Tool Condition Monitoring of Diamond-Coated Burrs with Acoustic Emission

Abstract

Within manufacturing there is a growing need for autonomous in-line Tool Condition Monitoring (TCM) systems with the ability to predict tool wear and failure. This need is only increased when using specialised tools such as Diamond-Coated Burrs (DCBs), in which the random nature of the tool and inconsistent manufacturing methods, create large variance in tool life. This unpredictable nature leads to a significant fraction of a DCB tools life being under-utilised. Acoustic Emission (AE) presents a possible inline monitoring technique which in conjunction with Machine Learning (ML) models could be used as a prediction method for DCB wear. Four wear life tests were conducted with a ø1.3 mm #1000 DCB, in which AE was continuously acquired during grinding passes, followed by surface measurements of the DCB. Three ML model architectures were trained on AE features to predict DCB mean radius, an indicator of overall tool wear. All architectures showed potential of learning from the dataset, with Long Short-Term Memory (LSTM) models performing the best, resulting in prediction error of MSE = 0.559 μm2. Additionally, links between AE kurtosis and the tool’s runout/form error were identified during an initial review of the data, showing potential for future work to focus on grinding effectiveness as well as overall wear. This paper has shown that AE contains sufficient information to enable inline monitoring of DCBs during the grinding process. ML models have been shown to be sufficiently precise in predicting overall DCB wear and have the potential of interpreting grinding condition.