(39d) Improving Control Room Operator Performance during Training Using a Cognitive Digital Twin

Process industries employ control operators to oversee plant operations. With the presence of advanced state of the art automation, the role of an operator in a process industry has shifted towards monitoring the plant and making decisions during abnormalities (Kluge et al., 2015). As operators are increasingly being expected to deal with abnormal situations, operator training and learning becomes significant. It is a common practice in the industry to use Operator Training Simulators (OTS) for training operators. Modern training simulators log the actions taken by an operator in response to situations to give feedback and understand operator’s learning of the process (Manca et al., 2014). Traditionally, operator performance in these simulators is studied by analysing the operator actions such as time to respond, number of actions taken, accuracy of the actions etc. (Adhitya et al., 2014). However, these methods do not account for the cognitive behaviour of operators. Cognitive aspects such as cognitive workload are important to assess operator performance during training (Huey and Wickens, 1993); increase in cognitive workload results in degradation of performance. Our previous research using eye-tracking reveals exciting insights into the cognitive behavior of control room operators. For instance, we found that there is a difference in gaze pattern of operators who successfully controlled the plant during abnormality against operators who failed to complete the task. Eye tracking is based on the “eye-mind” philosophy (Just and Carpenter, 1976); eye-gaze provides a trace of a person’s attention allocation and thus provides insights into their information acquisition pattern. In this work, we seek to utilize the difference in eye gaze pattern of experts and novices for operator training. Specifically, we propose a model-based approach using ACT-R (Adaptive Control of Thought – Rational) to obtain the eye gaze pattern of experts which is used to train the novice control room operators.

ACT-R is a theory about the working of human cognition and also provides a computational framework to model any cognitive tasks based on psychology experiments and facts. ACT-R has a symbolic component (responsible for procedural learning, decision-making) and a sub-symbolic learning (responsible for cognitive processes like thinking, learning). ACT-R also provides the locations of different modules responsible for different tasks (like decision-making, thinking, long-term memory, short-term memory, vision) inside the brain using fMRI studies (Anderson, 2009). ACT-R has been used for modelling behaviour in various safety critical domains such as aviation (Chen et al., 2021), and driving (Deng et al., 2019). In this work, we developed an ACT-R model to capture the eye gaze and actions of control room operators during plant abnormalities. The proposed ACT-R model has a total 48 production rules that modify the different modules so that it can interact with an HMI, take control actions, and learn about the process. For instance, changes in the production rule in visual module is utilized to search for various control valves in the HMI. The vision module of ACT-R also provides information about the eye gaze during the task. The eye gaze extracted from the proposed model is used to train the novice control room operators.

For development of the proposed ACT-R model, we created an ethanol production plant simulator and conducted experiments requiring a participant to control disturbances in the process. Six different scenarios were designed to test the understanding of the process. Participants would interact with the simulation using a custom designed HMI to control disturbance(s) corresponding to each scenario. Overall, ten participants performed the experiments consisting of six to ten trials for each participant with repetitive tasks. During the experiment eye gaze data, EEG, process data, and user actions data, and were recorded. We observed an improvement in the performance of participants across trials (due to learning from repetitive tasks). This is reflected in a reduced process error (measured as integral absolute abnormality), completion time, and cognitive workload across trials (Iqbal et al., 2021). This data was used to tune the production rules and parameters of the developed ACT-R model. The model learns the ethanol production process through its interaction with the HMI. It creates a mental model of the process and retrieves it to solve errors in subsequent trials. The model also replicates human subject results in a reduced process error (measured in integral absolute abnormality), completion time, and cognitive workload across trials. In this work, we will demonstrate the use of the eye gaze obtained from the proposed ACT-R model to train the control room operators during plan abnormalities.

Key words: Operator training, eye gaze, ACT-R, digital twin, Cognitive model

Acknowledgement: This work is partially funded by American Express Lab for Data Analytics, Risk and Technology, Indian Institute of Technology Madras

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

Anderson, J., 2009. How can the human mind occur in the physical universe?

Chen, H., Liu, S., Pang, L., Wanyan, X., Fang, Y., 2021. Developing an Improved ACT-R Model for Pilot Situation Awareness Measurement. IEEE Access 9, 122113–122124. https://doi.org/10.1109/ACCESS.2021.3108438

Deng, C., Cao, S., Wu, C., Lyu, N., 2019. Modeling Driver Take-Over Reaction Time and Emergency Response Time using an Integrated Cognitive Architecture: https://doi.org/10.1177/0361198119842114 2673, 380–390. https://doi.org/10.1177/0361198119842114

Huey, B., Wickens, C., 1993. Workload transition: Implications for individual and team performance.

Just, M.A., Carpenter, P.A., 1976. Eye fixations and cognitive processes. Cognitive Psychology 8, 441–480. https://doi.org/10.1016/0010-0285(76)90015-3

Kluge, A., Nazir, S., Manca, D., 2015. Advanced Applications in Process Control and Training Needs of Field and Control Room Operators. https://doi.org/10.1080/21577323.2014.920437 2, 121–136. https://doi.org/10.1080/21577323.2014.920437

Manca, D., Nazir, S., Colombo, S., Kluge, A., 2014. Procedure for automated assessment of industrial operators. Chemical Engineering Transactions 36, 391–396. https://doi.org/10.3303/CET1436066