Safety evaluation in the geotechnical site applications: A comprehensive analysis and forecasting model

Abstract

Geotechnical site works play a crucial role in the construction projects. In this study, data mining was used to evaluate the accidents occurred during geotechnical site works. OSHA Integrated Management Information System was used to collect 1010 cases in a time frame from 1984 to 2022. These cases were analyzed to evaluate the root causes of the accidents. Input variables of this study were the soil type and condition, project type and cost, end use of the work, cause and type of accident, type and degree of injury, unsafe acts, and occupation / union status of the victim. The evaluation showed that accidents not only had a high recurrence frequency but also a high severity level (57.2% fatalities). Moreover, an artificial estimating model based on a Multi-Layer Artificial Neural Network (MLANN) architecture was developed using MATLAB® platform in this study. In this regard, metaheuristic technique so-called Differential Evolution incorporated Virtual Mutant (FDEVM) was employed to optimize the weighted coefficients of the system. The FDEVM method is introduced recently and FDEVM in combination with MLANN is used for the first time in this study. As the model was trained for discrete set of data, the F1-Score test was performed to evaluate the performance of the models and based on the attained outcomes, it can predict the tasks with an accuracy of 78%. The root causes of accidents showed that project-specific countermeasures should be of high priority along with the implementation of vigorous strategies to develop safety measures.