Decision Trees and IMU Sensors for Risk Prediction in Manual Material Handling

Publication Date

1-1-2025

Document Type

Conference Proceeding

Publication Title

Lecture Notes in Computer Science

Volume

15791 LNCS

DOI

10.1007/978-3-031-93502-2_24

First Page

353

Last Page

364

Abstract

Manual material handling (MMH) tasks, involving activities like lifting, carrying, and holding various loads, contribute significantly to lower back fatigue, pain, injuries, and musculoskeletal disorders in occupational settings. This study employs machine learning algorithms to assess ergonomic risks in repetitive lifting tasks, aiming to enhance injury prevention, refine training protocols, and inform workload design for improved worker health. Using data from eight Inertial Measurement Unit (IMU) sensors, we developed predictive models to classify lifting tasks based on their biomechanical strain. Ten participants performed repetitive lifts with two box weights, with acceleration and gyroscope data collected. Feature engineering included total acceleration, total angular velocity, and statistical measures computed over a one-second rolling window. Three decision tree-based models achieved classification accuracies of 79.35% (gradient boosting), 91.49% (random forest), and 95.45% (XGBoost). The superior performance of XGBoost highlights the potential of IMU-based machine learning for real-time ergonomic risk assessment. These findings support the development of wearable safety systems for instant feedback, enabling proactive interventions that promote safer work environments and reduce musculoskeletal disorders.

Keywords

ergonomic risk assessment, Inertial measurement unit (IMU), machine learning, Manual material handling, Occupational safety, wearable sensors

Department

Aviation and Technology

Link to Full Text

Share

COinS