Development of an AI-supported Integrative platform for Green Building Material Selection

Abstract

Using Green Building Materials (GBMs) mitigates the negative impacts of the construction industry on the environment and occupants' health. However, the investigation results implied that the selection and use of GBMs in practical projects are still meager. The main reasons for this come from the lack of awareness, constraints, and transparent processes. These drawbacks lead to the hesitance of design teams and investors in evaluating and selecting GBMs. Many studies have been carried out to encourage the use of GBMs to assist the project team in guiding, evaluating, selecting, and managing GBMs. There are three main approaches, including (1) management processes solutions for using GBMs alternative and (2) studies proposing methods of evaluating the using GBMs alternative. Even proving the remarkable achievement of previous studies in evaluating and managing the use of GBMs in construction projects, both GBMs management solutions (1) and assessment methods (2) are still quite complicated for most construction project teams since the requirements of experience and information. Significantly, the lack of input information from GBMs makes the management and evaluation process unrealistic and badly affects the efficiency of construction projects. Thus, to provide the input information for the GBMs selection and material management during project time, various kinds of GBMs databases were developed. However, the resource dispersion in developing separate databases could never meet the GBMs's need for massive and constantly refreshed information since the rapidly increasing amount of GBMs data. Therefore, efforts to build centralized databases have been made for sharing resources. Prominent approaches to unifying these databases, such as standardization, API linkage, and increased degree of automation, have been implemented but have not yielded tangible results. In addition, the huge amount of information about GBMs, fast increasing, makes maintaining these databases extremely expensive. Therefore, this study proposed a new approach that integrates a centralized database and Natural language process (NLP) – machine learning (ML) classification to create a platform for providing GBMs information. To fulfill the mentioned gap, this study implemented (1) a supplementary framework GBMs selection task, (2) an analytic hierarchy process for evaluating GBMs alternatives, and (3) an integrative database that can automatically update and recognize GBMs's information using Natural Language Process (NLP) application and machine learning (ML) model was built to support information searching. Then, the results were integrated into a platform tested on a case project to evaluate their applicability. As a preliminary study for applying ML-NLP to GBMs information management, this study focused on the GBMs covered by the LEED DB+C Reference guideline. The study results showed that the combination of web-crawler and NLP-ML models could do well in automatically collecting and categorizing GBMs information with an accuracy of about 96%. For uncollected information, predictive models identified 85% of GBMs from descriptive information automatically collected by web crawlers. It means that the omission of GBMs information during searches across databases is greatly reduced. Given the limitations of model training data sources, this study did not support some GBMs information. Although the framework and FAHP models are developed for the entire project lifecycle, the GBMs information supported is primarily for the new construction phase of the building. By providing the ability to inform searching, assess, and manage the use of GBMs for all project team members, the platform proposed by this study contributes to the sustainability construction by promoting initiative and success in using GBMs. This work is also a pioneering study using NLP ML to improve the sharing of GBMs information.