In plants with a continuous production process, abnormal behavior can lead to leaks, blockages or jammed valves. Due to the various possible fault conditions, classic anomaly detection is not sufficient - the abnormal behavior must also be able to be assigned to a fault class. Another complicating factor is that tests can only be carried out to a limited extent on plants with continuous processes.
Development of an AI-based system that detects abnormal behavior and assigns the current state to a defect class. The evaluation is based on streaming data from the plant.
First, an experimental design is created and executed. Then, an AI-based system is developed that achieves good classification results especially on small data sets. Finally, a web interface connects the system to the plant so that results of streaming data are presented via a web application.
Demo system at the customer's site.