Decision makers in the housing industry need to carefully analyze housing design and construction decisions to improve housing environmental and economic performances. Available energy optimization models are able to find minimum-cost housing design and construction decisions at different target energy-saving levels. The application of these models, however, is limited due to their time-intensive and often impractical computational requirements. This paper presents a scalable and expandable parallel computing framework to reduce the computational time that is required to optimize the trade-offs between the environmental performance of housing units and their initial cost. The framework is designed as a global parallel optimization algorithm to provide an efficient distribution of the multiobjective genetic algorithm computations over a number of parallel processors. The optimization algorithm is also coupled with an external building energy simulation engine to enable an accurate modeling of housing energy performance. The performance of the parallel computing framework was tested using seven experiments that utilized 2, 4, 6, 8, 10, and 12 worker processors. The results of this analysis illustrated that the present framework is able to reduce the computational elapsed time from 12 days to 1.7 days by utilizing eight parallel worker processors that are commonly found on many PCs with quad core and eight threads. This significant reduction in the elapsed time proves that the developed parallel computing framework is capable of transforming the optimization of housing units from a time-consuming and often impractical problem to a feasible and practical one.
|Original language||English (US)|
|Journal||Journal of Computing in Civil Engineering|
|State||Published - Mar 1 2016|
ASJC Scopus subject areas
- Civil and Structural Engineering
- Computer Science Applications