Aplicação de inteligência artificial na estimativa da espessura da camada de tinta depositada sobre chapas de aço galvanizado em linha de pintura industrial

Abstract

This work consists of the study and development of an artificial intelligence integrated with a PLC through an industrial network, capable of estimating the thickness of the paint layer applied to galvanized steel sheets on an industrial production line. Currently, the adjustment of process variables depends on a rather inefficient experimental process, which demands a long downtime of the production line to perform, in a laboratory and through a sampling process, the measurement of the paint layer applied to the sheet in production. In this context, the proposed simulation system, based on the training of an artificial intelligence using a real database, should provide an estimate of the thickness of the paint layer that will be applied to the sheet, according to the parameters collected during the line setup or through manual data entry by the operator. The proposed simulator has the potential to significantly reduce production costs caused by line downtime, avoid raw material waste, monitor the thickness behavior according to the data provided by the controller in real time, and enable the evaluation of new parameter configurations, thus reducing the operator’s workload and removing the responsibility of becoming an expert in the experimental process of measuring the thickness of the applied paint layer. The performance of the proposed system was evaluated by comparing the results provided by the MLP and RF artificial intelligence models with the actual values sampled from a production line. According to the confusion matrix metrics presented in Sections 5.4 and 5.5, the MLP achieved an accuracy of approximately 70%, whereas the RF reached over 80% precision among the classes, with an overall accuracy of 84%.