Anomaly Detection for Bolt Tightening in Automotive Engine Assembly Based on Machine Learning

With the advancement of industrial automation, the quality requirements for automotive engine assembly bolt tightening have become increasingly stringent, as they are directly related to engine performance and vehicle safety. Traditional monitoring methods for bolt tightening rely on preset torque criteria, which often overlook anomalies such as material defects and tool wear. To address the issue of limited negative samples in industrial data, this study proposes an innovative positive sample boundary modelling strategy based on artificial intelligence, combined with a random forest algorithm for anomaly detection. By transforming time-series data into high-dimensional feature data, an efficient classifier is constructed through the integration of multiple decision trees. Experimental results from real engine assembly line data demonstrate that the proposed method can quickly and accurately identify abnormal conditions without increasing computational burden, significantly improving assembly quality and reducing failure rates.