286752 Proactive Abnormal Situations Management Using Anticipatory Alarms

Monday, October 29, 2012
Hall B (Convention Center )
Shichao Xu, Institute of Chemical and Engineering Sciences, Singapore , Singapore, Singapore and Rajagopalan Srinivasan, Department of Chemical Engineering, Indian Institute of Technology Gandhinagar, Gujarat, India


Modern chemical plants are built with large amount of integrated and interlinked process units. Such complexity ensures that these plants are operated in an optimized and profitable fashion. When an abnormal situation occurs, the automation systems in these plants alert the operators through alarms and help orient them to the new state. However, due to the coupling nature of chemical processes, the alarming system may generate many simultaneous alarms resulting in an alarm flood1. Such a deluge of alarms may disorient and cause information (or data) overload on the part of the operator.  To make matter worse, for inexperienced operators managing the plant this would lead to more confusion in the control room, delaying or preventing timely sense-making and eventually resulting in an emergency shut-down in the best case, or sometimes worse.

To address this issue, many alarm flood management techniques has been developed. For example, an alarm prioritization framework was introduced to ease the burden of operators with meaningless or false alarms2 and an intelligent alarm management system was proposed to suppress nuisance alarms in a chemical refinery to reduce the number of alarms to a more manageable level3,4. Other techniques also include the development of a refinery robot to assist operators in managing alarms in refineries during abnormal situations5 and the use of an intelligent operator decision support system, called Op-Aide, to assist the operator in quantitative diagnosis and assessment of alarms during abnormal situations6. While the above techniques have shown to improve alarm management especially during an alarm flood, none of them has proposed strategies to look at the sense-making facilities offer by the alarm system. As result, operators may still be disorient and overloaded with information caused by the number alarms during an alarm flood.      In this paper, we introduce a new type of alarms, called anticipatory alarms, that seeks to improve the sense-making facilities offered by the alarm system in the face of alarm floods.

The idea of anticipatory alarmsis to help plant operators, especially those that are less experienced, to manage alarms during an abnormal situation in a more proactive and organized manner. In most chemical plants, process variables are measured and monitored in real-time by tens of thousands of sensors. To ensure safe operation of the plant, thousands of alarms are configured on these process variables. In the event when a large disturbance or abnormal situation occurs in the process, operators’ attention is abducted by the occurring alarms which are also highlighted in the DCS schematic. Other variables that are moving towards their alarm limits but have not yet reached their alarm threshold would appear to be in a normal state. In the absence of trend information in today’s DCS schematics, the operator does not have the complete information required for inferring the true state of the plant or the process units. Anticipatory alarms seek to overcome this handicap of partial information offered by modern-day alarm systems. This type of alarms provide operators with anticipatory information on incipient alarms that are about to occur with a certain time-window. By having this information, operators can obtain a complete sense of the true state of the process which would help them, localize and rectify the problem early before the plant reaches a dangerous operating state. The availability of the anticipatory alarms in advance of their actual occurrence reduces the confusion in the control room caused by alarm floods and allows efficient handling of abnormal situations.

The proposed anticipatory alarms are built around an alarm anticipation algorithm, which utilizes dynamic models of the process to anticipate the alarms. These dynamic models, called anticipatory alarming models, are different to typical first-principle or data-driven models as they are able to (1) represent the dynamics of the process during an abnormal situation and (2) account for structural and parametric changes to the nominal model of the process so that they do not get less and less reliable during abnormal situations when the process deviates further from the nominal operating point. The alarm anticipation algorithm utilizes the anticipatory alarming models to predict the rate-of-change of process variables which are translated into predictions of time horizons for occurrence of various alarms. The time-prediction of alarms is then passed on to the operators through an anticipatory alarms display. The effectiveness of the anticipatory alarms is demonstrated using a case study of a depropanizer unit subjected to different abnormal situations.


[1]    J. Errington, D. Vernon Reising and C. Burns, “ASM consortium guidelines: Effective alarm management practices,” ASM Consoritum, Phoenix, 2009.

[2]    O.M. Foong, S.B. Sulaiman, D.R.B. Awang Rambli and N.S.B. Abdullah, “ALAP: Alarm prioritization system for oil refinery” In proceedings of the World Congress on Engineering and Computer Science (ISBN:978-988-18210-2-7), San Francisco, 2009.

[3]    J. Liu, K.W. Lim, W.K. Ho, K.C. Tan, R. Srinivasan and A. Tay, “The intelligent alarm management system”, IEEE Software, vol. 20, pp.  66-71, 2003.

[4]    J. Liu, K.W. Lim, R. Srinivasan, K.C. Tan and W.J. Ho, “The intelligent alarm management system”, Hydrocarbon Processing , pp.  47-53, 2004.

[5]    K. Krebsbach, and D. Musliner, “A refinery immobot for abnormal situation management,” Automated Reasoning Group, Honeywell Technology Center, 1998.

[6]    H. Vedam, S. Dash and V. Venkatasubramanian, “An intelligent operator decision support system for abnormal situation management”, Computers adn Chemical Engineering, vol. 23, pp. 577 – 580, 1999.

Extended Abstract: File Not Uploaded