Based on parameters of the paint job MagnaPaint predicts types of paint imperfections and informs the operator on which parameters have the strongest influence.
Our industrial partner Magna is continuously trying to improve its processes and products via innovative technologies and methods. One focus area is the paint finishing process, where vehicles are coated with protective lacquer. Due to external and internal influences, the coating may contain imperfections, which need to be manually removed, which is a costly process. By applying data science methods we analysed the data and identified a number of root causes for various types of imperfections, which help the operator to increase the overall quality. The data consists of a large number of parameters, ranging from chemical measurements to process information. Together with the domain experts of our industrial partner we developed a machine learning model, in order to forecast the expected quality of the processes. In cooperation with the Knowledge Visualisation Area we developed a tool allowing the operator visually interact with the learnt model. With this tool, the operator can experiment with different parameter sets and observe the predicted results, without the need to actually testing these parameters in the production environment. This again saves time and costs and also avoid potential disruptions in the production process.