PREDICTION OF MILLING MACHINE FAILURES USING MACHINE LEARNING MODEL,

Abstract

This study developed a random forest machine learning model for predicting milling machine failure using five input parameters, which include air temperature, process temperature, rotational speed, torque, and tool wear. Deploying the model helps to know when failure is likely to occur and what measures can be taken before the machine fails, including pre-emptive investigation, maintenance schedule adjustments, and repairs. The performance evaluation of the developed random forest model revealed that it predicts milling machine failure with high precision and accuracy, as evidenced by the performance metrics obtained during the model testing via accuracy, precision, recall, and F1 values of 0.98533, 0.71287, 0.82758, and 0.76595, respectively. The confusion matrix analysis shows the model correctly predicted 2884 no machine failures as true positives (TP) and 15 no machine failures as false positives (FP) out of 2899 no machine failure targets. In addition, the model predicted 72 machine failure targets as true negatives (TN) out of 101 machine failure targets and 29 as false negatives (FN). The model design determines the condition of the machine and predict when maintenance is needed to avoid breaking down. This improves the efficiency and productivity of milling machine operations, enabling proactive maintenance and reducing unplanned downtime.