A systematic approach to identify the hierarchical structure of accident factors with grey relations

• A quantitative approach to identify the hierarchy of accident factors.
• Automatic conversion from factors behavior sequences to systematic structure.
• Helps to effectively trace backward and forward the related accident factors.
• Theoretical foundation to conduct comprehensive treatments on accident factors.
• A way to learn from the law-enforcement records before accident happens.

Work safety represents a complex systems-phenomenon and should be analyzed with systematic approaches. Hierarchies are fundamental in the study of complex systems. By integrating Grey Relational Analysis and Interpretive Structural Modeling, this paper proposes a quantitative approach which permits an automatic development of the graphic hierarchy of accident factors on the basis of their behavior sequences. The accident factors are identified from the law-enforcement checklists. With Grey Relational Analysis, the relations among the accident factors are calculated based on their time series, which are generated from law-enforcement checklist records. By matrix operation, the grey reachability matrix of these factors is derived from the grey adjacency matrix which is composed of the grey relations. Following the steps of Interpretive Structural Modeling, the reachability matrix is partitioned into different levels by algebraic manipulations, and the factors are arranged in a hierarchical structure. The resulting hierarchy provides a holistic scenario of accident factors, which helps to effectively trace backward and forward the related accident factors in spot check, and serves as a theoretical foundation to conduct comprehensive treatments to eliminate accident factors. An example with the data from Xicheng District Administration of Work Safety is illustrated to show how the proposed approach works.

Identification of the systematic hierarchy of accident factors by its behavior sequences.Figure optionsDownload as PowerPoint slide