A Knowledge-Driven Approach to Automate Job Hazard Analysis Process

Abstract

Automating the job hazard analysis (JHA) process is an urgent requirement in the construction safety management field due to limitations of the conventional process. The manual nature of conducting the JHA and the dynamic environment of construction sites make it necessary to perform the analysis before commencing the job and to then regularly update it in accordance with changes in the construction plans. With this in mind, this research aims to develop an automated approach to support safety personnel during the JHA process. In seeking to automate the JHA process, the nature of construction accidents, hazards and risk assessment needs to be studied in light of the theoretical knowledge on accident causation. Thus, this research was designed according to the constructive research approach to develop a job hazard analysis knowledge graph (JHAKG) to automate the JHA process. The JHAKG incorporated an ontology (O-JHAKG) built according to the systematic ontology development method, METHONTOLOGY, which formalises both explicit and implicit knowledge inherent in the JHA process. The data were imported to the JHAKG from an incident database using rule-based natural language processing (NLP) which helped to extract implicit information not evident in the traditional JHA document. The validation of the JHAKG was conducted in two stages: the first stage validated the information extraction process by calculating performance metrics, while the second stage validated the data population process and the JHAKG's reasoning capability. The overall research resulted in a comprehensive JHAKG with advanced inferencing capabilities which can assist safety personnel in effectively executing the JHA process.

Keywords

• construction industry
• knowledge graph
• nlp
• ontology
• safety management