(371m) Analysis of Decision Support System for Control Room Operators Using Cognitive Measures.

Modern chemical process industries rely heavily on Distributed Control Systems (DCS) for efficient control and supervision. The information provided by the DCS is utilized by control room operators to make necessary adjustments in the plant, ensuring steady and safe operations. These adjustments rely on the operator's decision-making ability, involving high-level cognitive tasks. This cognitive task comprises the orientation of the operator's focus on relevant information, interpreting this information to find the root cause and finally, execution, which refers to the corrective actions taken by the operator to restore the process to its normal state(Adhitya et al., 2014). To ensure the accurate execution of these cognitive steps, operators are provided with decision-support tools. Alarms are employed to facilitate precise orientation, while values displayed in the graphical user interface (GUI) and trend panels serve as decision support aids during diagnosis and execution. Failure to comprehend the situation in any of these tasks can result in equipment outages, plant shutdowns, and production accidents. The DCS provides operators with information in the event of a system fault, but the critical aspect lies in utilizing this information effectively for fault diagnosis, which involves skillfully taking into account a subset of information and applying technical expertise to draw conclusions. This decision-making process can be enhanced if the operator knows the causality between different process variables; thus, in this study, we employed causal maps to assess their efficacy in aiding operators during the decision-making process.

The hierarchical method proposed by Suresh et al. (2019) was used to generate the causal maps, depicted as a series of directed graphs, illustrating the connections between strongly and weakly connected process variables. Many studies have been conducted using different cognitive measures to calculate the effectiveness of a proposed approach. Madhu et al. (2016) demonstrated that eye tracking is a reliable sensor of various cognitive tasks undertaken by operators and identifies unique fixation patterns reflecting the operator's abilities to orient, diagnose, and execute recovery actions in abnormal situations. Das et al. (2017) demonstrated that analyzing available input (from eye tracking) and output (operator actions) data in relation to the process state can assist in deducing the operator's mental model at any given moment. Similarly, Shahab et al. (2021) utilized the eye gaze data and proposed two quantitative metrics - association metric and salience metric- demonstrating that these metrics effectively quantify the operator learning process. The current study examines the impact of overlaying causal maps onto DCS, and data from eye tracking is analyzed to infer the causal maps treatment and learning effects.

Consider a typical DCS interface depicting a chemical process designed to simulate various tasks to examine the operator's ability to diagnose the root cause of the state variation. This evaluation is conducted initially using a traditional or standard interface and subsequently with the integration of causal maps, following an A-B-A-B single-subject design. Eye gaze data, including fixation and saccade information, is collected during a single process run. The Human-Machine Interface (HMI) continuously provides φ(t) as real-time information. Operators utilize this information to evaluate the plant's current state. In the event of abnormalities, operators rely on their mental model, which is informed by their understanding of the plant's dynamics or current information, to assess the root cause of the system's abnormality. The integration of causal maps affects the operator's mental model and ability to identify the root cause. In subsequent process runs, treatment and learning effects influence the operator's mental model, with pronounced impacts observed over time. Eye gaze data collected throughout the case study is analyzed to depict gaze transitions across the Human-Machine Interface (HMI), represented as a digraph linking different Areas of Interest (AOIs). The difference in the structure of this digraph pattern and causal maps will give the treatment effect, and the frequent transitions between AOIs indicate the learning effect. The effectiveness of the causal maps can be calculated by comparing the enhanced information provided through it and observing the resulting shift in the operator's mental model.

For the evaluation of the proposed methodology, it is implemented on the revised version of the Tennessee Eastman case study consisting of a typical DCS interface(Naef et al.,2022). Human factor studies were carried out with participants interacting with the interface both with and without the integration of causal maps. During the various simulated tasks, we recorded process data, alarm information, and participants' eye gaze patterns. Fixations on the HMI, weighted by their durations, illustrated the amount of attention allocated to different areas of the interface. Additionally, gaze transitions (saccades) indicated a more coherent navigation between related sections of the interface. The results indicate that in the case of complex systems, this layered framework could enhance the operator's understanding of couplings and the dynamic effects of process operations. Moreover, it could serve as a basis for enhancing HMI design and implementing causal maps as real-time decision support for operators.

References

Adhitya, A., Cheng, S.F., Lee, Z., Srinivasan, R., 2014. Quantifying the effectiveness of an alarm management system through human factors studies, Computers & Chemical Engineering, 67, 1-12, https://doi.org/10.1016/j.compchemeng.2014.03.013.

Suresh, R., Sivaram, A., Venkatasubramanian, V., 2019. A hierarchial approach for causal modeling of process systems, Computers & Chemical Engineering, 123, 170-183, https://doi.org/10.1016/j.compchemeng.2018.12.017.

Madhu, K., Srinivasan, B., Srinivasan, R., 2016. Towards predicting human error: Eye gaze analysis for identification of cognitive steps performed by control room operators, Journal of Loss Prevention in the Process Industries, 42, 35-46, https://doi.org/10.1016/j.jlp.2015.07.001

Das, L., Srinivasan, B., Srinivasan, R., 2017. Cognitive Behavior Based Framework for Operator Learning: Knowledge and Capability Assessment through Eye Tracking, Computer Aided Chemical Engineering, 40, 2977-2982, https://doi.org/10.1016/B978-0-444-63965-3.50498-0

Shahab, M.A., Iqbal, M.U., Srinivasan, B., Srinivasan, R., 2021. Metrics for objectively assessing operator training using eye gaze patterns, Process Safety and Environmental Protection, 156, 508-520, https://doi.org/10.1016/j.psep.2021.10.043

Naef, M., Chadha, K., Lefsrud, L., 2022. Decision support for process operators: Task loading in the days of big data, Journal of Loss Prevention in the Process Industries, 75, 104713, https://doi.org/10.1016/j.jlp.2021.104713