Model management for computational system design

Abstract: Computational system design is as an area within engineering design that to a large extent is based on simulation models together with tools and methods for analysis and design. This area is becoming increasingly important for providing information to designers and decision makers early in the design process, and enabling qualitative decisions to be made.Engineering design in general is today carried out in a collaborative, distributed, and at the same time competitive fashion where complex products are to a large extent assemblies of subsystems developed by several suppliers. It is important that models and computational design methods can be managed efficiently in such design projects, where models can be exchanged and integrated, design data managed, and proprietary information protected.In this thesis, a framework is presented which facilitates management of simulation models, design data, and computational processes. The framework emphasizes a model-centric view, which means that models are represented on a high level of abstraction, without tool-specific information.By defining models using the extensible Markup Language, XML, it is a straightforward matter to apply generic operators to the model and transform the model to several tool-specific implementations.The framework also enables integration of distributed computational resources. By implementing standards for so-called web service technology, simulation modules are published as remote model services, enabling partners in a project to deploy and invoke simulation models as black boxes. By publishing an interface to the model in a standardized format, and keeping the model at its original location, the model can be accessed and integrated without revealing proprietary information about the model. This approach also facilitates integration of existing so-called legacy tools and models.A data management approach is also presented where design data and the computational processes are handled. Variability measures for the design data are included, enabling probabilistic methodologies and design optimization to be applied in a straight-forward way. The computationaldesign process is formalized by representing the process as an executable process description enabling direct implementation, modification, and exchange of processes.In the last part of the thesis, a number of evaluation examples are presented where some of the tools and concepts of this thesis are illustrated. These examples are mainly from the area of aircraft system design.