Plantar Pressure Distribution During Stooping by Novice Roofers: Effects of Slope Angles Across Foot Zones and Machine Learning-Based Classification

Rujan Kayastha; Krishna Kisi; Anandi Dutta

Abstract:

Construction roofers face hazardous environments due to their activities at sloped and uneven surfaces, carrying heavy loads, and awkward postures. These postures increase musculoskeletal risks, particularly for novice roofers who struggle to balance their weight distribution on slopes. However, research on plantar pressure distribution among novice roofers remains limited. The study examines the effects of the roof slopes (0°, 15°, 30°) on plantar pressure distribution across foot zones (toe, metatarsal, midfoot, heel) during kneeling, standing, and stooping to identify high-risk areas and evaluate machine learning methods for pressure data classification. High-risk zones were identified based on statistically significant increases in peak pressure values (p < 0.05), indicating greater biomechanical load and potential musculoskeletal strain. Eight novice participants simulated shingle installation on a sloped wooden mock-up in a laboratory setting. The study used ANOVA to analyze plantar pressure distribution of peak pressure across various foot zones at different slopes by using XSensor pressure insoles. Three-way ANOVA was conducted to statistically analyze the effects of posture, slope angle, and foot zone on peak pressure. Further, four machine learning models - Random Forest, Decision Tree, Support Vector Machine, and K-nearest neighbors were applied for the classification. The results indicate posture significantly affects plantar pressure. Among machine learning models, Random Forest achieved the highest accuracy (training: 92%, test: 91%), demonstrating its strength in handling non-linear relationships and high-dimensional data. This study advances roofing ergonomics by quantifying plantar pressure variations on inclined surfaces, identifying high-risk foot zones for musculoskeletal disorders, and showcasing machine learning's potential for automated posture classification. These findings can inform safer work practices and ergonomic interventions for novice roofers.