Ai-based framework for risk estimation in workplace

Abstract A significant problem for industrial management has always been the frequency of workplace fatalities and injuries. From production plants to offices to construction sites and beyond, every workplace has safety threats and hazards. To recognize certain objects, circumstances, practices, etc., that can cause harm, especially to individuals, a detailed look at the workplace becomes crucial. It is important to evaluate and determine how severe and likely the danger is after identification is made. The range of cumulative challenges, partially linked to technology developments with rising expectations, has to be broadly considered. Currently, it needs systematic risk estimation, enhancement of past lessons in learning, and the concept of appropriate data processing techniques to be combined with adequate capacity to cope with unforeseen events and provide the right support to enable risk management. Therefore, this paper suggests a risk management methodology focused on artificial intelligence (AI). A Deep Neural Network for Workplace Risk Estimation (DNN-WRE) model is explicitly generated and trained to predict the working environment's risks. The DNN-WRE is evaluated by a common risk prediction seen in the workplace, i.e., Musculoskeletal Disorders (MDs). Furthermore, this paper re-tuned the transfer learning network ResNet 18 to predict MDs. The training and validation of the proposed DNN-WRE have been observed with the highest prediction accuracy of 93.86% compared with the pre-trained ResNet 18 model.