Application of artificial intelligence in model-based systems engineering of automated production systems

Abstract

Despite the incontestable appeal, the application of artificial intelligence (AI) in engineering processes is still limited to isolated applications and, in some fields, enthusiasm has given way to disillusionment. This paper aims to contribute to a concept of a framework that allows the application of AI in model-based systems engineering (MBSE) processes of automated production systems; the main focus is hereby on the MBSE processes. The aim of the complete framework is to realize an AI-based, self-learning digital twin that automatically adapts to the real system behavior and represents an optimal image of a product and its production process at all times. An expressive, semantic overall model serves as the basis for new approaches to artificial intelligence. In the complete framework, knowledge gained using AI methods is integrated into the overall model and thus brought into an overall context. Such an overall model improves the interpretability and explainability of the AI models and enables complex analyses, simulations and forecasts. The core element of the approach is a novel, AI-based, self-learning engineering model consisting of a product and production model that maps function, behavior and product geometry. Graph-based design languages are used for forming a central data model and functional mock-up units are applied for continuous co-simulation. The approach is underlined by means of an application to the design of automated assembly systems.