Model-based testing methodology using system entity structures for MATLAB/Simulink models

Abstract

The increasing complexity of systems entails an increasing complexity of simulation models. Likewise, heterogeneity in system components corresponds to heterogeneous simulation models. Cyber physical systems (CPS) represent an emerging class of technical systems characterized by their complexity and heterogeneity. Developing simulation models for CPS brings various challenges, one of which is determining the simulation fidelity. Fidelity evaluation can be introduced as the degree to which a simulation model matches the characteristics of the system it represents. Due to the growth of system complexity in CPS, the number of test cases required to reach admissible coverage to assure adequate simulation fidelity is very high. Along with that, heterogeneity in system components comes on top as another challenge. Therefore, adaptability, flexibility and automation can be identified as the key characteristics of a fidelity evaluation approach that determines its success. Model-based testing (MBT) advocates the use of models for the specification of test cases and proposes workflows for automatic test case generation. This paper presents an MBT approach for objective fidelity evaluation of complex, modular simulation models. The methodology implies that appropriate data for fidelity evaluation are available. Each test model is represented according to the formal structure of experimental frame (EF). For generating an executable EF for a model under test (MUT), configurable basic models are provided by a model base (MB). In the same manner, configurable basic models for composing various MUTs are stored in the MB. The system entity structure (SES) ontology is used for the specification of a family of MUT and test model designs on an abstract level. This means that the SES describes a set of various MUT and test model structures and parameter settings. Using the SES and MB, a specific executable model consisting of an MUT and a test model can be generated. Based on these ideas an infrastructure implementation for automated fidelity evaluation of complex, modular simulation models within MATLAB/Simulink is proposed in this paper.