Alexandra Mehlhase

A Python framework to create and simulate models with variable structure in common simulation environments

In modelling and simulation, model complexity increases with the complexity of real systems, often resulting in unfeasibly long simulation times. Variable-structure models, which can change their equation set during run-time, offer a solution. This article introduces an object-oriented approach that describes such models independent of a simulation environment. The problem of changing equations during run-time is addressed using a Python framework. Three tools are already integrated, namely, Dymola, OpenModelica and Matlab/Simulink. With this framework, existing models can be easily reused, and the advantages of different simulation tools can be leveraged. The framework is illustrated using a simple satellite launch example, along with instructions for how to use the framework. The main aim of the framework is to simulate complex models with a few mode changes and thus save simulation time. With two real-world examples and a scalability analysis, it is shown that the framework fulfils these requirements.

A Library and Scripting Language for Tool Independent Simulation Descriptions

In modeling and simulation it is often necessary to simulate a model with a variety of settings and evaluate the simulation results with measured data or previously acquired results. As doing this manually is error-prone and ineffective, scripting languages are often used to automate this process. In general a simulation description is tool and model dependent. Therefore, simulating the same model with the same simulation description in different simulation tools or comparing two different models with the same settings is often not easily achieved. We propose an object-oriented, tool-independent, easy-to-use, domain-specific scripting language to describe simulations in an exchangeable and uniform manner. Through this simulation description the simulation settings and the simulation environment can easily be changed while syntax and sequence of commands remain the same. The language is Python based and is designed to be simple, well-readable and intuitive even with marginal programming experience while maintaining Pythons’ strength. The language uses an in-house Python library which provides interfaces to different simulation environments (so far Dymola, OpenModelica, Simulink). This library can also be used directly in Python, enabling experienced Python users to keep describing their simulations in Python but benefiting from our efforts to achieve tool-independence.

Challenges when Creating Variable-structure Models

Variable-structure models can switch their system of equations during a simulation, allowing for a change of level of detail or behavior. The need for this kind of models has been well-established, and there are simulation environments that can handle them. While most research papers on this topic focus on language and tool issues regarding variable-structure models, in this paper, we will shed some light on the pragmatics of actually creating such a model in a reusable way. During the construction of a variable-structure model, the modeler will face several challenges, such as the initialization of new modes during mode switches. We will collect and discuss the most important challenges and, if possible, provide rules of thumb on how to handle these challenges appropriately.