E-Coating Ultrafiltration System Maintenance using Machine Learning Techniques

Abstract

Ultrafiltration process is one of the important processes in e-coating of metal parts. It is important to maintain and improve the performance of ultrafiltration membrane (UF membrane). This UF membrane should not be degraded under any situation, if this happens then it might lead to the flooding of coating fluid all over the metal parts. Hence it has to monitor properly. This will also lead to the wastage of coating fluid, wastage of materials and maintenance cost will also be high. So to avoid this, the workers have to monitor it on periodically basis. During e-coating in electrophoresis painting plant there may occur fault in filter and excess of fluid will flood in the material to be coated. The filter used in the ultrafiltration subsystem has to be monitored manually each time by the worker. Sometimes, if the worker did not monitor properly, it might lead coating to inappropriate places, it will cause wastage of material and fluid and also it takes time for clean-up and to change after fault occurs. Hence, the timely detection of faults is very important to prevent damage. In order to prevent these kinds of situations, a machine learning model is developed which predicts the flow meter readings beforehand. It uses an ensemble learning algorithm known as XGBoost which is one of the more powerful tools for prediction.