Predictive Modeling of Work-Related Musculoskeletal Disorders among Shuttle Car Operators Using Artificial Neural Network

Background: Work-related Musculoskeletal Disorders (WRMSDs) are highly prevalent among equipment operators in the mining industry, particularly in underground coal mining. The multifaceted nature of risk factors contributing to WRMSDs necessitates a robust and accurate predictive model to assess the likelihood of these disorders among operators. Methodology: This study presents a novel Artificial Neural Network (ANN)-based predictive model developed to estimate the prevalence of WRMSDs among seated vehicle operators. Key input variables include demographic and ergonomic factors such as age, Body Mass Index (BMI), years of experience, posture scores (REBA), frequency-weighted Root Mean Square (RMS) acceleration, and Vibration Dose Values (VDV). Data on WRMSDs prevalence and severity were collected using the standard Nordic Musculoskeletal Questionnaire (NMQ). Results: The analysis demonstrated a strong positive correlation between the severity of WRMSDs and the identified risk factors. The optimal ANN configuration included hidden layers with sizes 5, 10, 15, learning rates 0.01, 0.1, 0.2, and class weights 1, 1, 0.5, 1.5, and 1, 2. The model's performance was validated using 12 unseen datasets, achieving excellent predictive accuracy (0.975 ± 0.014), precision (0.805 ± 0.083), recall (1.000 ± 0.000), F1 score (0.890 ± 0.053), and AUC-ROC (0.996 ± 0.005) (mean ± standard deviation). Conclusion: This ANN-based model offers a powerful tool for predicting WRMSDs across different body regions of seated vehicle operators. Its ability to incorporate diverse input factors will support the implementation of targeted occupational health and safety interventions, particularly for shuttle car operators and other mining professionals at high risk of WRMSDs. Major Findings: The study developed an ANN-based model to predict WRMSDs prevalence among seated vehicle operators in underground mining, incorporating demographic and ergonomic factors. The model achieved high predictive accuracy, with a strong correlation between WRMSDs severity and identified risk factors.