Intelligent identification of machining damage in ceramic matrix composites based on deep learning

Abstract

This study proposed a method for identifying and quantitatively evaluating the machining damages of CMCs based on deep learning. Firstly, grinding tests of CMCs were conducted to create a dataset of machining damages. Then, six deep learning algorithms were trained using the dataset, and their comprehensive performance was compared. The results showed that YOLOv8 exhibited superior overall performance among the six algorithms. Besides, a professional software for identifying machining damage of CMCs was developed based on the optimal algorithm, and the influence of machining parameters on CMCs damages was investigated. Qualitative and quantitative evaluation results indicate that grinding speed is negatively correlated with the machining damage degree, and a higher grinding speed leads to less damages. In contrast, both feed rate and grinding depth are positively related to the machining damage. Furthermore, it is verified that the developed software is applicable to various conditions and has certain engineering application prospects.