Increasing Dependability by Agent-Based Model-Checking During Run-Time

Agent-oriented software engineering (AOSE) is a paradigm for distributing intelligent control mechanisms (ICM) within an automated production system (aPS). Benefits resulting from AOSE have been surveyed in many applications as route-finding, plug-and-produce techniques and also in the control of Smart Grids. To ensure safe functionalities, i.e. dependability or uptime, of distributed technical systems for instance by conducting simulation, virtual commissioning, the execution of test cases and model-checking are commonly investigated in aPS during the design phase. In this paper we analyze an automatic diagnostic method to increase dependability by using model-checking during run-time, based on discretized models of the mechanical plant as well as models of the PLC software. Consequently the algorithm is incorporated into a software agent and logically coupled to a particular aPS module. Thus, the dependability for introducing novel product types, which have not been involved in the design process, could be increased. The evaluation of our approach is shown at a small lab-scale production system by searching for counter-examples of combinations with control actions and work piece (WP) types with modified mass, that may lead to a production halt.