Abstract and Benefits
Over the last decade, the development of many types of products has evolved into a systems integration problem. Large numbers of sub-systems are integrated to achieve unparalleled functionality and reliability at an ever decreasing cost. To continue this trend, there is a need for computer-based support that allows the knowledge and expertise of many domain experts to be shared, managed, and applied towards the thorough exploration of different system architectures while considering complex trade-offs between multiple objectives under uncertainty. Supporting this complex endeavor is the focus of Model-Based Systems Development.
In system design, one often formulates design problems as a sequence of decisions about increasingly detailed characteristics of the sub-systems and components. In each decision, information is needed about the sub-system characteristics (e.g., weight, power consumption, or cost). However, these characteristics depend on the choices made in subsequent decisions and are thus currently unknown or uncertain, hence predictive models are needed that can fortell the characteristics likely to be achieved by the subsystem. Specifically, the speakers addressed these questions:
- How to model and support system-level design decisions?
- How is model fidelity and refinement managed with respect to uncertainties?
- How to capture and re-use system-level knowledge?
- How to predict the performance, cost, reliability of systems effectively?
- What are the limits of predictive modeling?
- How can system complexity be defined in this context?
Presentations
- M.Sc. Rich Malak (Georgia Institue of Technology, USA). Predictive Tradeoff Modeling for System-Level Decision Making.
- Dr. Petter Krus (Linköping University, Sweden). Performance Estimation Using Similarity Models and Design Information Entropy.
- Dr. Stephane Caro (Ecole Central de Nantes, France), The Role of Entropy in Design Theory and Methodology
- Dr. Harrison M Kim (University of Illinois at Urbana-Champaign, USA). Predictive System Design by Data Mining and Analytical Target Cascading.
In between, and after the talks there was lively discussion among the 24 participants who had come to participate on this Sunday morning. It was noted that this is a timely subject, and that a proper definition should be made with a coherent nomenclature. There was no time to go in depth with this discussion, but one candidate name could be engineering design predictive models.