Wear particles recognition through teachable machine

Abstract

Purpose The purpose of this paper is to adapt teachable machine as a web-based tool for recognition of wear pattern and type of wear by training a convolutional neural network (CNN) model. This helps to monitor the health of the lubricated system as a part of condition monitoring. Design/methodology/approach Ferrography technique is used for analysis of wear particles. It helps monitor the condition of lubricated mechanical system. In present paper, CNN model is developed for identifying the type of wear particles coming out of Gearbox system using teachable machine. Findings From the experimentation, it has been observed that the wear severity index has been increased due to increase in wear particle concentration. CNN model has achieved an accuracy of 95.4% to recognize five categories of wear particles. Originality/value Teachable machine is generally used for the prediction of images, gestures and sound features. An attempt is made to apply this model for micro and nano wear particles to classify them based on their characteristics.