A design and analysis of computer experiments approach for green building

Abstract

The coming shortage of energy sources and critical environmental impacts are two major factors that have forced a change in product design processes. The shortage of energy sources is related to the limitation on non-renewable energy sources on earth and requires the development of new concepts with lower energy consumptions. Environmental impact, on the other hand, is concerned about the negative effects of products on the natural environment. In recent years, research on designing more environmentally-friendly products that consume lower of energy with lower environmental impact has been initiated to address these issues [1-3]. Designing new powertrains for vehicles [4] and conducting research on developing airplanes with new sources of energy [5] are some of the examples. Building structures are of great interest, since the building have a significant impact on environment and energy consumption [6]. Buildings can be designed so that their energy consumption is reduced, by using new materials with higher thermal resistance, or implementing new design strategies to reduce the heat extraction from the building. In addition, there is a certain life cycle for any structure, which includes the time span between the manufacturing of the materials to the decomposition of these materials, and is called “cradle-to-grave” [6]. This cycle is usually used as a criterion for the environmental effect, and minimizing this effect is of great interest. While it is desirable to simultaneously minimize both energy consumption and environmental impacts, it is not straightforward to achieve because these two objectives depend on variety of factors. Therefore, it would be helpful to implement a multiple-objective optimization approach to design a building that satisfies both objectives. Buildings are complicated structures that include different subsystems, making it a multivariate, multi-response case study. In this study, two computer experiments are designed to evaluate the performance of a building with the focus on the energy consumption and the environmental impact. Since building variables include both categorical and continuous variables, two different design of experiments approaches are used to combine them together. The computer simulation of the energy consumption is performed in eQUEST [7], while the environmental impact is calculated in ATHENA impact estimator software [8]. The goal of the current work is to compare different experimental designs and different statistical modeling methods to help inform our approach for a multivariate, multi-response framework. For this purpose, a residential building is considered as the case study, and different design factors that affect energy consumption and the environmental impacts of the building are identified. A design of experiment is implemented to realize the simulations that can provide the data to study both performance objectives. Finally, the results of the experiments are studied using treed regression and multivariate adaptive regression splines approaches to identify important factors.