Keywords: product modelling, process modelling, decision support.

Previous research has shown the need to improve the efficiency of building design through the adoption of integrated product models such as the Industry Foundation Classes (IFCs) and CIMSteel Integration Standards (CIS/2). The focus for the development of these standards has been upon the formal description of each logical item that may exist, for design, analysis, manufacturing or organisational purposes, at each stage during the buildings design process. However, the information and relationships specified in this way are sometimes inappropriate for describing the intentions of participants in the buildings design.

A building design is initiated with the clients brief that sets forth requirements relating to the buildings intended use and quality, upon which the architect must make decisions regarding the buildings layout and design which produces a subset of constraints that the engineer must satisfy. The design tasks carried out by the engineer consist of high-level tasks that have a significant influence on the buildings design and which produce a further subset of requirements that must be satisfied by low-level tasks, which individually have only local impact on the building. As many of these low-level tasks are highly deterministic it is desirable to automate as many of them as possible. However, such automation is not possible without formalising the design constraints and intentions from the engineer's high-level design. Software agents could then carry out these delegated tasks, saving time and reducing human error.

The idea behind storing design intent in the buildings model is to capture key decisions made by the engineer significant to the buildings realisation. The aspects that the engineer deems significant to the design of the building should be stored in a way that allows less significant design issues to be postponed until a later stage in the design process, or to be specified in such a way as to provide clear boundaries and requirements so that the remaining decisions can be carried out by intelligent agents. For example, when initially planning the design of a multi-storey building, it is necessary at some stage to fix upon a floor plan detailing the column positions with primary beam locations and flooring span direction. It should not be necessary at this stage to fix upon the column or primary beam dimensions, or secondary beam spacing and dimensions. Instead, by specifying the design requirements of the floor, the designer may delegate the task of detailing the flooring configuration to an appropriately configured automated agent capable of determining the secondary beam spacing and dimensions as a function of the floor type specified by the designer, and the loads imposed on the floor above. In this way, much repetitive work appropriate for calculation by the computer may be removed from the designer so that he/she can concentrate on the more significant design issues, and test more design options at the initial stages of design.

This fundamental change to the design process will also help alleviate some of the problems associated with making late changes in the buildings design, as relatively significant procedures such as removing a column from the centre of the building could be largely accommodated by the intelligent agents making necessary alterations to individual element dimensions. This would then require the engineer to make only the important alterations to the buildings design necessary to provide an economically feasible solution.

The emphasis of the CIS/2 model is upon the design and fabrication of the buildings steelwork, while the IFCs place a greater emphasis on architectural design issues. The Virtual Building Product Model (VBPM) is based on the CIS/2 model due to its comprehensive inclusion of structural analysis support, and enhances the model with a 'design intent' layer that allows for the specification of the design intentions of the engineer. Without a product model suitably designed to take into account the engineers design intent, it is not possible to improve the building design process by making use of the many developing A.I. systems that would be useful in the mainstream building design process. The provision of a formal definition for this type of information in the product model is important to facilitate an improved design process in which design participants from multiple disciplines are able to express decisions at the conceptual stage of design. The advantages of such a system become increasingly apparent as more design participants begin working directly with a single product model, and require a more natural way of specifying requirements and design decisions.

The development of automated systems in the design process raises several important issues relating to the allocation of responsibility among design participants and software agents. There is currently little evidence of systems implementing rigorous systems for information management within a comprehensive product model, and the implementation of intelligent agents will add an extra level of complexity in which it may be appropriate for these agents to be considered as true design participants themselves. In this case it will be necessary for one or more parties to be responsible for the decisions made by the agent, and this could require a fundamental change in the way that inter-disciplinary alliances are formed.

These issues are discussed further in the paper, along with a description of the product and process model and it's implementation in the prototype `3D Virtual Building Software'.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £82 +P&P)