Framework for a resilience system in safety management: a simulation and visualization approach

Because workers’ behavior and attitudes toward safety are crucial factors affecting the likelihood of construction accidents, traditionally, surveys have been taken or field observations conducted at a site to measure the safety climate. Although most accidents are directly or indirectly related to persons’ unsafe acts, such traditional methods have practical limitations for periodic application at a site in terms of workers’ participation, the efficiency of observation, and the absence of an expert. Thus, safety management could be improved if workers’ behavior could be monitored and analyzed automatically and if the measured result could be used to predict future risk. This paper presents a conceptual framework for a resilience system that uses a simulation and visualization approach to manage safety behavior. A computer vision technique capable of motion detection is introduced to monitor workers’ behavior with video cameras and to automatically analyze their level of unsafe acts. The procedures for and pace of activities are assumed to change under progress pressure, so any change may represent the amount of pressure workers perceive that may lead them to engage in unsafe acts. By detecting workers’ motions related to back injury, their attitudes toward safety can also be measured, based on the assumption that their attitudes should be managed if they repeatedly use unsafe poses after receiving education about back injuries. A system dynamics model was established from the perspective of resilience engineering to represent the safety culture at a site, and as the results are input, the organization’s level of safety is simulated to examine potential threats and restore safety to an acceptable level. This framework is expected to illustrate the applicability of a computer vision technique for safety climate measurement and to demonstrate the usability of a system dynamics model for a resilience system.