Analysis of GRAFCET specifications to detect design flaws (AGRAFE 2)

• Applicants: Professor Dr.-Ing. Alexander Fay; Professor Dr.-Ing. Stefan Kowalewski
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 445866207

GRAFCET is a graphical specification language for the specification of discrete control sequences, which can thus be specified independently of the type and brand of the programmable logic controller on which the control program will be executed later. GRAFCET is widely used in France and is also becoming increasingly well known in Germany, partly because GRAFCET has become a mandatory part of the vocational education in the field of mechatronics a few years ago. In the previous project AGRAFE - besides the completion of a meta-model and a code generator for IEC 61131-3 control code - methods in the area of model checking and static analysis were developed, which can be used for the detection of design errors in GRAFCET. The goal of the proposed research project is to further improve these methods. During the work on AGRAFE, it has become apparent that both approaches are suitable for identifying different types of errors, so the approaches are now to be combined with each other in order to exploit their respective strengths. In particular, in the context of model checking, it has become apparent that state-space explosion is also a significant problem in GRAFCET, which means that analysis of industrial-scale GRAFCET specifications is not possible with a reasonable amount of resources. However, initial results from AGRAFE have shown that incorporating more information from GRAFCET-specifications can counter this problem. Thus, a first research goal is to counteract the state space explosion by using information from the static analysis. Another goal is to include information from the underlying process in the analysis to reduce the state-space explosion and improve the analysis results. Another research goal concerns the automatic generation of queries to be evaluated during model checking as well as the evaluation of analysis results that are presented as so-called counterexamples. Both processes require expert knowledge in today's practice. The last research goal is the automatic identification of error causes and, based on this, correction suggestions in order to support the correction of design errors in an automated way.

• DFG Programme: Research Grants