Download Presentation (231 KB)
Worker behaviors are complicated and under the influences of various factors when worker-vehicle collisions happen on construction job sites. The proposed research targets the safety challenges of construction management when industrial trucks are operating around workers. To solve the research question of how to identify the most influential safety hazards and patterns of the worker-vehicle coordination, this research first reviews and compares multiple data-mining algorithms for pattern analysis to select the Latent Dirichlet Allocation (LDA) approach and design the corresponding analysis system. Then it investigates the patterns of collision accident from the Occupational Safety and Health Administration (OSHA) database with the expectation to understand safety hazards and violations in worker-vehicle collisions based on the unstructured OSHA data. The intellectual meanings that occur in the collection of documents through the proposed LDA and statistical analysis of this research can support their future implementations of automated construction. This research also models the topics through text classification and suggests that the uneven ground and objects that are under-construction are the primary obstacles when workers and trucks move on the sites and should be managed for safety improvement.
Shi, Tianyuan, "Vehicle-Collision Warning System And Deep Learning Approach" (2021). Information Technology. 3.