Automated Assembly Machine Stopping System Using Machine Learning for Electrical Failure Prediction in Hard Disk Drive Manufacturing

Abstract

The assembly of diode lasers and sliders is a critical process in hard disk drive (HDD) manufacturing, directly affecting product electrical performance. Delays in detecting problematic assembly machines increase scrap rates, reduce yield, and raise production costs. This research presents the development of supervised machine learning models to predict electrical testing results using key parameters from assembly machines, for example, diode laser alignment position after assembly and laser intensity measurement during assembly. Dataset was preprocessed through outlier removal and feature selection using correlation analysis and permutation importance before applying SMOTE technique to balance the minority class in the training set, followed by normalization before modeling and evaluating performance. Multiple models were developed and evaluated to select the best performing model. Evaluation results showed that the AdaBoost model combined with Random Forest achieved the highest performance, with an AUC of 92.36 %. The selected model was integrated into monitoring, alerting and machine auto stopping system when the predicted failure rate exceeded a predefined threshold. The implemented system reduces the detection time for problematic assembly machines, minimizes defective parts, and improves overall manufacturing efficiency.