Research Overview

As manufacturing transitions from Industry 4.0 to Industry 5.0, a renewed focus is placed on human-centric operations and the well-being of workers. According to the data analysis from the Bureau of Labor and Statistics by the National Safety Council (NSC), about 24% of all non-fatal injuries that led to days away from work were due to back and lumbar injuries in 2019. Further, about 30% of these injuries were recorded among workers involved in manual lifting and material handling tasks

Research team: Vignesh Selvaraj, Aditya Nagaraj

Publications

Why was this study conducted?

  • To improve worker safety in manufacturing industries.
  • To enable real-time assessment of lifting operations as 'safe' or 'risky'.
  • To perform elaborate feature importance study to better understand the factors contributing to lifting activity assessment.
  • To test and evaluate the wireless sensor systems developed here.

Human silhouette
Sensor system locations on the human body for worker monitoring.
Human silhouette
F1-scores for the ML models considered in this study.
Human silhouette
F1-scores across ML algorithms for different sensor types.

Code Repositories

Harnessing physics and data to better understand the problem

sensor system location
Most contributing sensor type categorized by sensor system location.
sensor system location
Most contributing among the extracted features categorized by sensor system location.

Architecture of real-time inference deployed on sensor system using tinyML

sensor system location
Architecture and State Machine (SM) for real-time inference.