Design support for viaducts

The paper deals with the development of a system for supporting the design of highway bridges. The adopted approach is considered to innovate in relation to previous approaches of design systems for viaducts, for making a pragmatic combined use of a number of AI techniques such as CBR, neural networks and notions of agency. An interesting feature of the proposal is related to the fact that, instead endorsing a complex, and usually rigid design model for the adaptation phase of CBR-based design, the authors provide a conceptual (and partly implemented) framework for designers to interactively adapt the retrieved solutions. This adaptation takes place by using a constrained graphical/drafting environment, i.e. a CAD environment where design entities may carry physical and functional constraints relating to them in isolation, or to themselves in relation to the surrounding ones. Such constraints may be posted prior to the adaptation phase or throughout that process.