Knowledge mining for collaborative relationships in hydropower project safety management

• Data-driven collaborative relationships automatic generation and mapping. • A multilevel collaborative relationship knowledge mining approach. • Social network analysis is expanded to visualize and examine complex collaborative relationships. • Quantitative analyses incorporate both macro and micro factors across the project, department, and employee levels. • The quantified knowledge is utilized to facilitate practical safety management. Safety management requires frequent collaboration for various activities efficiently. The collaborative relationships among stakeholders contain valuable knowledge, including participant roles and information transmission, crucial for enhancing communication efficiency. This study proposes a knowledge mining approach for collaborative relationships in safety management, with three procedures: 1) collecting structured data to automatically generate collaborative networks, 2) employing social network analysis (SNA) to mine knowledge of key participants, information flows, and evolutionary characteristics of collaborative relationships, and 3) conducting analyses at project, department, and employee levels. Validated in a hydropower project, the approach extracts knowledge and provides specific safety management measures at different levels. The results demonstrate they enhance collaboration efficiency and shorten the time required for safety hazard rectification. In hydropower project safety management, collaboration relies on a few key participants who connect most others. Effectively managing these participants is crucial, as they play a pivotal role in overall collaboration. Centrality measures reveal differences in departmental roles, facilitating resource allocation and process optimization. Employee performance is quantified through the defined collaboration contribution rate and collaborative performance, enabling targeted incentives. This study provides both theoretical and practical insights into collaborative safety management, particularly in data acquisition, multilevel quantitative analysis, and knowledge mining.