A Probabilistic Accident Prediction Model for Construction Sites

Abstract

Despite the efforts in recent years to improve safety on construction sites, construction still has one of the highest fatalities rates among Canadian industries. The objective of this study is to improve the safety on construction sites by developing an Accident Warning System that is capable of predicting potential accidents, and prevent them by alerting the involved parties. Recent studies have shown that close monitoring the real-time status of moving objects and providing immediate feedback to the workers can improve the effectiveness of any safety strategies. The proposed system consists of an Ultra Wideband Real-Time Locating System and a probabilistic Accident Prediction Model. Using the information provided by the UWB RTLS, the model predicts the future locations of the monitored objects. These are analyzed, and a warning is issued when the potential occurrence of an accident is identified. As such, rather than detecting accidents shortly before they occur, the model is able to predict them and alert the involved parties in advance, providing ample time for reaction. The decision of issuing a warning is taken based on the expected consequences of the probable accident. The proposed system can improve the current status of safety on construction sites through predicting and preventing accidents. It can be used to monitor the operations on construction sites in order to reduce the number of injuries and fatalities.