Enhancing automated defect detection through sequential clustering and classification: An industrial case study using the Sine-Cosine Algorithm, Possibilistic Fuzzy c-means, and Artificial Neural Network

Most existing inspection models solely classify defects as either good or bad, focusing primarily on separating flaws from perfect ones. The sequential clustering and classification technique (SCC) is used in this work to not only identify and categorize the defects but also investigate their root causes. Conventional clustering techniques like k-means, fuzzy c-means, and self-organizing map are employed in the first stage to find the defects in the finished products. Then, a novel clustering method, that combines a sine-cosine algorithm and possibilistic fuzzy c-means (SCA-PFCM), is proposed to classify the detected defects into multiple groups to identify the defect categories and analyze the root causes of failures. In the second stage, the ground truth labels taken from the clustering technique are used to construct an automated inspection system using back propagation neural networks (BPNN). The proposed approach is applicable for detecting and identifying the causes of errors in manufacturing industry. This study applies a case study in nipper manufacture. The SCA-PFCM algorithm can detect 97 % of defects and classify them into four types while BPNN shows a predicted accuracy of up to 96 %. Additionally, an automated inspection system is developed to reduce the time and cost of the inspection process.