Using blockmodeling for capturing knowledge: The case of energy analysis in the construction phase of oil and gas facilities. (January 2019)

• Record Type: Journal Article
• Authors: Aragao, Rodrigo Rodrigues; El-Diraby, Tamer E.
• Is Part Of: Advanced engineering informatics. Volume 39(2019)
• Page Start: 214
• Page End: 226
• Publication Date: 2019-01
• Subjects: Knowledge retrieval and representation -- Network analysis -- Blockmodeling -- Construction -- Energy assessment -- Oil and gas; Computer-aided engineering -- Periodicals; Engineering -- Data processing -- Periodicals; 620.00285
• DOI: 10.1016/j.aei.2019.01.003
• Languages: English
• ISSNs: 1474-0346

Abstract

Highlights: Blockmodeling is used to obtain a blockmodel of concepts related to the energy consumption of construction projects. This baseline is compared with three case-based networks of real oil and gas projects. The results show that the knowledge constructs of the cases are unique and context-sensitive. The approach represents a meaningful way to retrieve the knowledge in textual cases. The method can be used to guide the planning phase of similar projects in the future. Abstract: In this paper, blockmodeling, a network analysis clustering approach, is used to study and capture clusters of concepts. The semantics networks are a former stage of the extracted, formalized knowledge from unstructured data (text). As a sample domain of the application of the proposed approach, the networks are focused on concepts related to planning for energy management during the construction phase. Text describing the status or lessons learned from projects is transferred to semantic networks. A benchmark concept network was generated based on surveying experts. Blockmodeling algorithms were used to create a set of concept blocks: clusters of related concepts that capture some of the knowledge of a generic scenario. Through interviews with staff from three projects, we developed 3 case studies (text) to capture the conditions and knowledge gained in these projects. A concept network of main concepts was extracted for each case. Also, blockmodels from these networks were also extracted. To facilitate the comparison, an average block analogy index k ¯ was introduced. The smaller k ¯ is, the more dissimilar the studied blocks are, and the more unique and unusual are the characteristics of the project at hand. By contrasting the blocks of the case projects against each other and against the benchmark network, we identified unique knowledge constructs (concept clusters) in the three projects. This can be beneficial in capturing project-specific knowledge; contrasting project conditions and knowledge concepts; and supporting a frequent upgrade of the benchmark concept network or a formal ontology.