Constructability knowledge acquisition: A machine learning approach
Abstract
Project constructability can be devaluated significantly because of poor structural design decisions. However, the aspects of structural design decisions in constructability have not been thoroughly emphasized in the constructability concepts currently applied in the industry. This research proposes a methodology to acquire constructability knowledge according to structural design decisions made during conceptual phase. Constructability is understood as "an important feature of a structural design and construction project site conditions which determines the level of complexity of executing the associated structural assembly task". Constructability knowledge is acquired from structural design data of building structures, proposed construction methods, and resource availability conditions. Determining constructability of a project requires experience and expertise, which may not be available. A inductive learning system is proposed as an alternative knowledge acquisition tool. The system is capable of knowledge acquisition and generating desired concepts from classified constructability examples. Three methods for; (1) the preparation of constructability examples; (2) the constructability knowledge acquisition; and (3) the verification and validation of acquired knowledge, were proposed to develop such a learning system for constructability knowledge acquisition. Constructability knowledge is acquired in form of decision rules, and can be updated by implementing multistage knowledge acquisition process. Direct data extraction is proposed to extract structural design data from design drawings in CAD. Additional information necessary to the knowledge acquisition can be obtained from preliminary project plan and proposal. Acquired constructability knowledge can be used for future applications in the constructability domain, e.g. identifying potential structural design problems to improve overall project's constructability.
Degree
Ph.D.
Advisors
Skibniewski, Purdue University.
Subject Area
Civil engineering|Systems design|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.