434972 Event Driven Multivariate Analysis of Eye Gaze Data for Behavior Analysis in Process Operations

Monday, November 9, 2015
Exhibit Hall 1 (Salt Palace Convention Center)
Madhu Kodappully, ndian Institute of Technology Gandhinagar, Ahmedabad, India, Babji Srinivasan, Department Chemical Engineering, Indian Institute of Technology Gandhinagar, Ahmedabad, India and Rajagopalan Srinivasan, Department of Chemical Engineering, Indian Institute of Technology Gandhinagar, Gujarat, India


Accidents at onshore and offshore process facilities often lead to heavy, irrecoverable losses. Investigation reports aftermath such occurrences frequently cite human errors as a primary cause for more than 70% of accidents in the recent past (Leveson, 2004). Even at smaller scale, equipment outage, plant shutdown and various production accidents in process plants are often considered as a result of operator errors viz. slips, lapses, mistakes, violations (Kidam et al., 2010). Despite numerous interventions in the last twenty years by the governments and industry around the world to minimize the impact of human error and to improve process safety,  25% of the accidents that led to largest losses in the hydrocarbon industry over a period of 40 years have happened in the last 5 years from 2009 (Marsh,2014). Analysis suggest that some of the primary causes for human errors in process industries include increase in complexity of the processes, tight energy integration, recycles, presence of advanced control strategies, reduction in number of staffs without an effort to increase their cognitive skills in handling abnormal situations. Especially, the cognitive skills in handling plant abnormal situations play a key role. While the traditional approach has been to look at human error as the initiating event using a set of Human Error Probabilities (Munger et al., 1962), modified approaches that came later extended to cover task/environmental variables along with human-engineering-design characteristics (Swain, 1990). Further ahead, Embrey and Zaed (2010) described a set of techniques, supported by computer based tools, for predicting and preventing human errors in gas plant operations which primarily focused on Hierarchical Task Analysis and Predictive Human Error Analysis. However, none of the approaches to understand human error, till date, do not explicitly account for the cognitive behaviour of the operators under abnormal situations, an important phenomena that influences human error. To address this gap, we seek to understand the cognitive ability of a control room operator during an abnormal plant condition.

Eye movements are considered to offer a real-time window into cognition based on the “eye mind” hypothesis and therefore provide dynamic information about human’s cognitive behavior (Just and Carpenter, 1976; Cooke, 2005). Therefore, researchers in various fields, including safety critical systems viz. aviation, health care, have utilized eye tracking measures to analyze different aspects pertaining to cognitive processing. Matos (2010) suggested that the eye tracker is appropriate to monitor the cognitive state of plant operators since it is non-invasive and relatively simple to install and use. In our work, we employed a Tobii TX 300 eye tracker to track and record the eye movement of participants while interacting with the HMI of a plant simulator. As the participant has to look at several dynamic components on the HMI screen (Area of Interest-AOI) while resolving an abnormality, we used AOI dwell duration, AOI visit count, AOI transitions as a measure to understand the cognitive state of the participant during the abnormality. Additionally, researchers have categorized the cognitive tasks during an operation into orientation, diagnosis and execution steps (Chu et al., 1994) and employed eye tracking measures to study them.

 In this work, we demonstrate that there are distinct patterns even in the cognitive behaviors of plant operators.

This work has focused on determining the eye gaze patterns that manifest themselves while a control room operator is using a HMI to control a process abnormality. We conducted experiments where graduate students played the role of control room. From their eye gaze data, a strong correlation is evident between AOI-based measures and the orientation, diagnosis, and execution steps while rejecting a process disturbance. It was observed that while the trend pane in the HMI was the most dominant AOI during the entire course of the task, the dominance and relevance of the trend pane does vary among phases. During the initial orientation steps, the primary tag (corresponding to the variable that is directly by the disturbance) is the most used while the trend pane is ranked higher during later phases that correspond to diagnosis and execution. These insights into the cognitive behavior of control room operators provide a new approach to tackling human error in a proactive fashion.


Chu, Rose, et al. "Qualitative user aiding for alarm management (QUALM): an integrated demonstration of emerging technologies for aiding process control operators." Systems, Man, and  Cybernetics, 1994. Humans, Information and Technology., 1994 IEEE International Conference on. Vol. 1. IEEE, 1994.

Cooke, L., 2005. Eye tracking: How it works and how it relates to usability. Technical Communication 52 (4), 456–463.

Embrey, D., and Zaed, S. (2010). A Set of Computer Based Tools Identifying and Preventing Human Error in Plant Operations. In 6th Global Conference on Safety, AIChE (pp. 1–11). San Antonio.

Just, M. A., and Carpenter, P. A. (1976). Eye fixations and cognitive processes. Cognitive Psychology, 8(4), 441–480. http://doi.org/10.1016/0010-0285(76)90015-3 26

Kidam, K., Hurme, M., and Hassim, M. H. (2010). Technical analysis of accident in chemical process industry and lessons learnt. Chemical Engineering Transactions, 19, 451–456.

Leveson, N. (2004). A new accident model for engineering safer systems. Safety Science, 42, 237–270. http://doi.org/10.1016/S0925-7535(03)00047-X

Marsh (2014). The 100 Largest Losses 1974–2013. Marsh and McLennan Companies. Marshall, S. P. (2007). Identifying cognitive state from eye metrics. Aviation, Space, and Environmental Medicine, 78(5 Suppl), B165–B175.

Matos, R. (2010). Designing eye tracking experiments to measure human behavior. Merriënboer, J. J. G. van, and Sweller, J. (2005). Cognitive load theory and complex learning: Recent developments and future directions. Educational Psychology Review (Vol.17). http://doi.org/10.1007/s10648-005-3951-0

Munger, S., Smith, R., Payne, W., 1962. An index of electronic equipment operability: Data store. Tech. Rep. AIRC43-l/62-RP(l), American Institute for Research, Pittsburgh.

Extended Abstract: File Not Uploaded