Human learning curve forecasting & optimization framework for manual assembly operations

Abstract

The manufacturing industry is undergoing a significant transformation, led by the widespread adoption of Industry 4.0 technologies and data-driven production management systems. While monitoring and optimization have become common for machine operations, manual operations are mostly disconnected from these advancements, due to persistent challenges in data acquisition and the intrusiveness of monitoring methods. More importantly, low-cost countries keep manual assembly a core part of their operations, based on costs and flexibility. This second element, however, presents a challenge to companies, due to human behavior not being as perfectly repetitive as machines, leading to differences between planned production time and actual production time. One factor not currently considered in planning cycle times and production capacity is the learning effect, represented by prolonged cycle times in the first production units, but improving over time. Traditional approaches to track the learning effect have seen little application on processes in recent times, resulting in missed opportunities for productivity forecasting and optimization. The primary objective of this thesis is to present a comprehensive framework for the collection, operational forecasting, and productivity enhancement of production cycle times in manual operations by leveraging data paired with a simple data collection method. This work proposes a novel human learning curve measurement and optimization solution that mirrors the sophistication of machine monitoring applied to humans. It also considers a data problem commonly found in the manufacturing industry, which is excessive data collection, making predictions and fitting curves computationally expensive, by considering a simplification method. Key contributions of this Ph.D. thesis include a state-of-the-art review on learning curves, learning curve parameterization methods, and data simplification techniques, which led to the development of a “Human Learning Curve Forecasting and Optimization Framework”. The Ph.D. thesis also presents both controlled and industrial experimentation for the validation of the framework. The Ph.D. thesis results present the benefits of analyzing the human learning effect in productivity, presenting the industry with the opportunity to take immediate action to improve and increase efficiency in the short- and long-term, ultimately integrating the human factor in the decision-making for performance improvements. The Ph.D. thesis presented calls for a change in the way manual operations are being analyzed, by considering the learning curve effect, analyzing it in the short- and long-term, and presenting an alternative way to plan production in line.