Machine learning application in predictive maintenance on an automation line

Abstract

The purpose of this study is to explore application of Machine Learning algorithm in the Predictive Maintenance on an Automation line. Screw height, torque and height data from Auto Gang Drive were used to train machine-learning model. Proper control of the driving process is critical for screw torque process that applied the clamp force is equally distribution. Auto Gang Driver module cycle time is 4.5 seconds, and rapid process control is required to ensure successful process. A supervised machine learning approach is applied for this study. The data were pre-processed and classified into two types of classifications, which are “passed” and “failed”. The ground truth was performed by visual inspection of the workpiece, which is a Hard Disk Drive disk clamp screw driving assembly. Two models of machine learning, Support Vector Model and Decision Tree models, were explored to compare the accuracy of the model. The result showed that Decision Tree has 100% accuracy in predicting the detection of the failure. The Decision Tree model was then deployed on the Auto Gang Driver module to monitor the screw driving process. A framework for machine learning implementation was drawn to replicate the implementation to other automation module. Future work such as monitoring of the health of the machine using data such as incoming compressed air, pressure and flow by applying machine learning can deploy predictive maintenance on the machine.