Prediction on Industrial Processes through Explainable Artificial Intelligence (pipeAI)
Call: R&D program "Information and Communication Technology"
Project management: Prof. Dr. Christian Janiesch
Project participants: ROBUR Automation GmbH, SKZ - KFE gGmbH, Julius-Maximilians-Universität Würzburg
Project start : 01.09.2020
Project end: 31.08.2023
Funding volume: 772,000 euros
Funding source: Bayerisches Staatsministerium für Wirtschaft, Landesentwicklung und Energie (StMWi)
Machine learning methods have been applied to industrial data for a long time, e.g. to predict machine failures. Their effectiveness depends on the reliability of the predictions and the acceptance by the user. To promote the widespread use of AI tools by AI-inexperienced employees, pipeAI addresses this problem. New AI approaches (e.g. GQNs) are tested in an industrial context, supported by improved methods for data cleaning and automated labeling. Furthermore, with Explainable AI and active participation of the user, e.g. in Transfer Learning, AI tools are made more comprehensible. As an example, the use case Predictive Maintenance is pursued. The resulting implementation is exemplary and generalizable.