As industries shift from the machine-driven focus of Industry 4.0 to the cooperative vision of Industry 5.0, a new story unfolds. While Industry 4.0 pushed for more automation and less human involvement, it soon became clear that machines, despite their precision, couldn't match human adaptability. This led to the birth of Industry 5.0, highlighting a balance between human insight and machine capability. To make this partnership work, both sides need to understand each other's functions. Machines, being programmable, have behaviors that are predictably mapped out. Humans, on the other hand, carry with them mental models—conceptual frameworks developed during their training—that influence their decision-making. Discrepancies in these mental models are frequently cited as primary contributors to human errors. If the essence of Industry 5.0 is to be realized, it is imperative to unravel and understand these mental models. Yet, a glaring gap emerges when we evaluate how well we understand these two collaborative entities. Traditional evaluations lean heavily on tangible benchmarks such as error frequency, response times, and task outcomes, along with insights from expert operators. These established methods, while valuable, tend to bypass the deeper cognitive processes underlying operator actions. With modern advancements, however, we are now equipped with state-of-the-art sensor technology, including eye-tracking technology. This thesis seeks to develop quantitative performance metrics using eye-tracking tailored to assess cognitive performance. Transitioning from the broader context of Industry 5.0, it becomes evident that the success of this human-centric approach rests on the bedrock of understanding operators' mental models.
Generally, there are two types of industrial operators: control room operators and field operators. Control room operators, often considered the nerve center of industrial operations, are stationed in front of panels. They sift through a deluge of data, making decisions that steer the larger process. To understand their cognitive dynamics, we employed screen-based eye tracking. This allowed us to delve deep into their attention patterns. Quantitative measures were developed using Fixation Transition Entropy (FTE), Correspondence Analysis (CA), and Hidden Markov Model (HMM) to objectively evaluate operator performance. The FTE provides quantitative information about the operator’s eye gaze transitions on the Human-Machine Interface (HMI). Metrics developed using Correspondence Analysis (CA) – association and salience metrics – went a step further, pinpointing the specific process variables that captivated their attention, especially during process abnormalities. With the intent of understanding the underlying cognitive frameworks opera tors rely on, the Hidden Markov Model (HMM) provided a mathematical blueprint of their evolving thought processes during operations. Two axioms of learning are derived from the proposed HMM-based mental models. These axioms help evaluate operator training in terms of understanding causal relationships and proactive monitoring strategies.
Yet, industries aren't solely steered by those behind the panels. Field operators, often the eyes and ears on the ground, interact directly with the processes. Their dynamic work environment, marked by constant movement, renders traditional screen-based eye tracking inadequate. Recognizing this, our study forayed into the use of eye-tracking glasses tailored for field settings. Using the bounding boxes indicative of their gaze, we harnessed the power of Self- Organizing Maps (SOM). This approach clustered similar gaze patterns, illuminating areas of interest and subsequently highlighting key attention zones in a shifting environment. This approach not only redefined our traditional perception of what constitutes an AOI but also pinpointed what truly captures the attention of operators. Furthermore, the method showcased remarkable time efficiency, significantly outperforming manual annotation methods. We also applied this methodology to identify AOIs for control room operators, achieving superior accuracy. By integrating SOM in both these scenarios, we've successfully built a bridge between two distinct operational worlds, giving a holistic perspective on cognitive behaviors across industrial landscapes. We conducted tests on both simulators and pilot plants, which validated the efficiency of these metrics for both operator types. In essence, this research offers a holistic view of operator performance, from panel- based decision-making to on-ground actions, ensuring that their choices are informed, precise, and prioritize safety. The quantitative nature of these metrics sidesteps the pitfalls of subjectivity, championing a consistent, repeatable, and neutral assessment methodology.
In summary, the crux of Industry 5.0 lies in a symbiotic partnership where both human intellect and machine precision work hand in hand. To realize this vision, it is vital to understand the cognitive frameworks, or mental models, that guide human decision-making in this partnership. Traditional evaluation methods often fall short, overlooking the deeper cognitive nuances. This thesis develops novel qunatitivate measures to assess mental models of industrial operators using eye tracking. By unveiling the intricacies of operator’s mental models, industries can tailor training programs, design more intuitive human-machine interfaces, and adapt machine learning algorithms to better align with human thought processes. This fine-tuning ensures a harmonized interaction between humans and machines, fostering a work environment that optimally leverages human intuition alongside machine efficiency, fully realizing the vision of Industry 5.0.