Energy efficiency and a high degree of reliability are essential aspects in complex process plants. For this reason, process monitoring, fault detection and diagnosis are an important part. Due to a high degree of individualization regarding customer demands, the development of model-based methods involve high engineering effort. The high number of sensors and actuator used for process control, monitoring and quality control generate large amount of data motivating the application of data-based methods.
In cooperation with the Dürr Systems AG, approaches and methods for the data-based modeling, fault detection and diagnosis are developed. These methods are applied and validated for different processes in a paint shop, like pretreatment processes, cathodic electrocoating processes or air supply units.
Different machine learning algorithms and a large amount of data are used for the analysis and data-based modeling of the system. Based on the data-based model, process monitoring as well as fault detection and diagnosis is performed.