Supporting Model-Based Diagnostics with Equation-Based Object Oriented Languages

The paper focuses on the application of equation-based object oriented languages to creating models for modelbased diagnosis. We discuss characteristics and language constructs essential for diagnostic purposes. In particular, we describe the main features of the declarative modeling language Rodelica, which is based on the well-known language Modelica but enhances it with additional features related to diagnosis. Rodelica is used in a commercial model-based diagnosis tool to build and exploit complex diagnostic models of industrial size. Developed models can be used in an interactive diagnostic process as well as for the generation of more compact forms of diagnostic knowledge like diagnostic rules or decision trees which are popular for on-board diagnostics or troubleshooting in the service bay. A case study concludes the paper, illustrating those applications and emphasizing their implications for the language itself.