Bi-directional Inference in Thermal Design

This paper demonstrates a computational bi-directional energy modeling approach for building design development. Conventional simulation tools may be labeled as mono-directional in that they require a more or Iess complete design definition in order to derive performance indicators. However, in certain circumstances, it may be desirable to reverse this process: a bi-directional (or "open") inference mechanism would allow for the identification of those changes in the design variables that would accommodate a desired change in a performance indicator. The performance-to-design mapping process is an ambiguous one: the same performance (e.g. energy use of a building, temperature variations in a space) may be achieved by different design configurations (various wall and window dimensions/properties, building orientation/massing, etc.). As a result, the actual implementation of a bi-directional inference tool is a rather difficult task. The development described in this paper utilizes a preference-based approach that involves the formalization of various external or internal constraints and preferences (such as code and standard requirements, results of post-occupancy studies, individual priorities of designers and their clients, etc.) in terms of normalized numeric scales. After a brief review of the underlying technology for the implementation of the inference engine, the paper demonstrates an actual design session using a bi-directional thermal simulation tool. Specifically, a use-scenario is described in which the designer explores the tradeoffs between various design variables (glazing area, glazing type, and floor mass) in view of the resulting energy performance of a typical residential building. The paper concludes with a discussion of the potential and limitations of the bi-directional approach toward active convergence support for performance-oriented design development.