(537g) Emerging Cognitive Engineering Approaches to Enhance Control Room Operator Capabilities in Chemical Industries
AIChE Annual Meeting
Wednesday, November 18, 2020 - 9:30am to 9:45am
Control room operatorâs actions can have significant effect on the throughput, quality and safety of process plants (Baron, 1998). Typically, control room operators monitor the process via a Human Machine Interface (HMI) on which all the necessary information about the process state is presented. The complexity of operating modern chemical plants have increased significantly due to tight mass and energy integrations, deployment of sophisticated control automation strategies, and the need to operate the process in agile fashion in order to capture market and supply chain dynamics. However, there has been no concomitant increase in operations personnelâs cognitive ability. Any mismatch between the operatorsâ capability and the process control demands leads to high cognitive workload in human operators, often a precursor for poor performance which in-turn leads to sub-optimal operator actions or at worst accidents. This paper addresses the critical need to understand and enhance control room operatorâs cognitive capabilities.
Traditional operator performance assessment methods ignore cognitive aspects of performance. Operator's performance is assessed primarily in terms of number of successes and failures, response times, deviation of operator actions from predefined sequence (Lee et al., 2000), etc. However, in addition to overt performance (as done traditionally) it is imperative to consider covert aspects of performance i.e., cognitive workload for a holistic assessment of human performance. Various emerging cognitive engineering based sensors such as eye gaze tracking, Electroencephalography can reveal insights into the operatorsâ cognitive state that determines these covert aspects. Specifically, in this paper we propose an Electroencephalography (EEG) based methodology to assess the operatorsâ cognitive workload, especially during abnormal situation tasks where their active involvement is essential to safe process operations. The potential of EEG to understand cognitive workload for dynamic tasks which evolve with time has received some attention in aviation (Liang et al., 2018) and driving (Morales et al., (2017). However, there are hardly any studies relating to chemical control room operators.
We propose a single dry electrode EEG based methodology to dynamically assess the cognitive workload of operators while they perform various actions to control the process during abnormalities. The proposed methodology is illustrated on a simulated ethanol production plant testbed. A set of 10 different operators were considered in this work. Each operatorâs performance during the course of experiments involving monitoring and control of the process was studied. In addition to the process data and operator actions, we recorded the EEG data of the operators. Each experiment involved 6 tasks (comprising 5 different scenarios, out of which was repeated) of different abnormal situation. Each operator repeated the same experiment 7-8 times. Overall, 73 experiments were conducted by the 10 operators for a total of 438 tasks. Due to data quality issues, we evaluate results for 8 participants only.
We analyze the EEG signal in frequency domain by applying Fourier transform. Our results reveal that the power spectral density (PSD) of brain waves (4-30 Hz) is a biomarker that can identify the workload of operators. The PSD data of each operator was partitioned into two clustered using k-means clustering. Cluster 1 represents low power spectral density with low magnitude and Cluster 2 showed high power spectra density. We studied the workload variation with the help of cluster fraction over the course of the task as the operatorsâ engagement with the process increased as well as with repetition. To evaluate the cluster variation within each task, a task is divided in three phases. In Phase 0 the operator monitors the process and ensures all the process variables are within the desired limits. During an abnormal condition an alarm is triggered, which leads to Phase 1 where the operator uses his mental model of the process to decide the control action(s) in order to clear the alarm. The portion of the task corresponding to all the control actions taken by the operator is called as Phase2. We observe distinct differences in the cluster fractions across the 3 phases. For a typical operator who successfully completes a task, during the monitoring Phase 0 we found the fraction of Cluster 1 to dominate (about 60%) indicating a correspondence between low workload during monitoring. During abnormality when alarms are triggered, in phase 1 the Cluster 1 fraction decreases (to about 40%) indicating that the operator uses his mental resources to decide the control action, i.e., increased workload. Further in phase 2, as the operator takes various control actions the Cluster 1 fraction further drops significantly (to about 0.20%) until the process becomes normal at which juncture it increases back to the 60% observed in Phase 0. The above patterns of change in the cluster pattern was observed for 81% (243 out of 300) of the tasks conducted by the 8 participants. This brings forth the capacity of the high accuracy of the EEG based cognitive sensor and the proposed machine learning algorithm to accurately estimate the workload of the control operator during normal and abnormal conditions. These distinct patterns were also highly correlated various process metrics (extent of variable deviation beyond alarm limits, time taken to clear alarm, number of standing alarms etc). In this paper, we will describe the proposed experimental and computational methodology and reported detailed results from our human subject experiments.
Baron, RG. (1988). Human Factors in the Process Industries. Human Factors and Decision Making: Their Influence on Safety and Reliability, Symposium for the Safety and Reliability Society. ed. Sayers, BA. (pp.1-9)
Heard, J., Harriott, C. E., & Adams, J. A. (2018). A survey of workload assessment algorithms. IEEE Transactions on Human-Machine Systems, (99), 1-18.
Srinivasan, R., Srinivasan, B., Iqbal, M. U., Nemet, A., & Kravanja, Z. (2019). Recent developments towards enhancing process safety: Inherent safety and cognitive engineering. Computers & Chemical Engineering, 128, 364-383.
Neurosky. (2015). Greek Îlphabet Soup â Making Sense of EEG Bands (Last accessed: Nov 10, 2018) http://neurosky.com/2015/05/greek-alphabet-soup-making-sense-of-eeg-bands/
Pfurtscheller, G., Stancak Jr, A., & Neuper, C. (1996). Event-related synchronization (ERS) in the Î± bandâan electrophysiological correlate of cortical idling: a review. International journal of psychophysiology, 24(1-2), 39-46
Liang, Yongqiang, Wei Liang, Jue Qu, and Jie Yang. "Experimental Study on EEG with Different Cognitive Load." In 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pp. 4351-4356. IEEE, 2018.
Morales, J. M., Diaz-Piedra, C., Rieiro, H., Roca-GonzÃ¡lez, J., Romero, S., Catena, A., Fuentes, L.J. & Di Stasi, L. L. (2017). Monitoring driver fatigue using a single-channel electroencephalographic device: A validation study by gaze-based, driving performance, and subjective data. Accident Analysis & Prevention, 109, 62-69.
Sharma, C., Bhavsar, P., Srinivasan, B., & Srinivasan, R. (2016). Eye gaze movement studies of control room operators: A novel approach to improve process safety. Computers & Chemical Engineering, 85, 43-57.
Bhavsar, P., Srinivasan, B., & Srinivasan, R. (2017). Quantifying situation awareness of control room operators using eye-gaze behavior. Computers & Chemical Engineering, 106,191-201.
Iqbal, M. U., Srinivasan, B., & Srinivasan, R. (2019). Electroencephalogram based Biomarkers for Tracking the Cognitive Workload of Operators in Process Industries. In Computer Aided Chemical Engineering (Vol. 46). https://doi.org/10.1016/B978-0-12-818634-3.50233-2