Framework for Developing Trustworthy Industrial AI Applications

Artificial intelligence (AI) offers great potential to increase productivity in production engineering. Nevertheless, manufacturing companies fail to successfully establish industrial AI applications on the shopfloor. Reasons are the great complexity and intransparency of AI as well as high performance and reliability requirements of production systems. To overcome these challenges and to comply with increasing regulatory requirements (e.g., the EU AI Act), we propose a framework for guiding the development of trustworthy industrial AI applications. Based on the core requirements of the production engineering domain, we propose a risk-based approach and define four qualification dimensions: safety, robustness, transparency and adaptivity.