Abstract & Overview
With the current rapid development of China's national economy, the detection of ceramic defect materials and product quality risk assessment of ceramic material nano-casting products have gradually become an important link in the process of industrial quality production in China.In this paper, starting from the perspective of ceramic machine engineering learning, by designing and constructing a ceramic defect material polymerization classification recognizer, it can realize the automatic clustering of standard ceramic product material defects.The main content of the research and the focus of technological innovation are mainly divided into the following several main aspects: First, the production of a standard ceramic defect detection atlas.Second, feature extraction and selection.Finally, this paper studies and constructs an industrial defect sample cluster classification recognizer based on the ultra-micro machine deep learning classification clustering analysis algorithm, and performs classification clustering algorithm identification based on ceramic material sample classification defect samples.First, we analyzed the actual effect of the four types of defect characteristics on the actual effect of the identification of industrial defects based on the formation of ceramic material samples, and then started the analysis training verification experiment based on the classification and identification of industrial defects of ceramic material samples.From the ceramic standard sample defect recognition map, 180 ceramic standard sample defect recognition samples were selected and extracted, and the defect training set and analysis verification set were obtained by using the "leave out method" and the new layered contrast sampling method, and then through comparative analysis PCA clustering algorithm and analysis ReliefF algorithm.The ReliefF clustering algorithm in this paper shows that the analysis accuracy of the ceramic material sample classification defect sample cluster classification recognizer has reached 86.5%, and a good sample classification defect recognition experimental effect has been achieved.