Modelling performance during repetitive precision tasks using wearable sensors: a data-driven approach.

Area of Science:

• Industrial Engineering
• Human Factors Engineering
• Biomedical Engineering

Background:

• Repetitive precision tasks in manufacturing assembly lines can negatively impact worker well-being and system efficiency.
• Human dexterity and flexibility are crucial in modern manufacturing, but their performance in repetitive tasks needs objective evaluation.

Purpose of the Study:

• To develop a data-driven approach for evaluating worker task performance in assembly lines using wearable sensor data.
• To model and predict task performance metrics such as errors and speed.

Main Methods:

• Collected data from 18 participants performing maze tracking and assembly/disassembly tasks using wearable sensors (kinematics, electromyography, heart rate).
• Employed various classification algorithms, including Linear Discriminant Analysis (LDA), k-nearest neighbors (kNN), and random forests (RandFor), to model task performance.
• Evaluated model accuracy using different combinations of sensor data types.

Main Results:

• Linear Discriminant Analysis (LDA) combined with kinematic data showed promising classification performance.
• The highest accuracy was achieved using LDA with all data types, predicting maze errors (62.4%), maze speed (88.6%), assembly/disassembly errors (85.8%), and assembly/disassembly speed (94.1%).

Conclusions:

• The proposed data-driven approach using wearable sensors and LDA enables comprehensive, real-time monitoring of worker performance in assembly lines.
• This technology can help identify potential issues related to worker fatigue or task execution, optimizing system performance and worker well-being.