How to perform alarm rationalization

Disadvantages are long-term online analysis requires long online time; automatic correlation with operational, process, or equipment events may not be available or may have to be entered manually; and some desired alarm metrics may not be available from the system.

It should be noted that not all aspects of alarm analysis/rationalization may be provided by one method, so a combination of static and dynamic methods may be required.

Benchmarking can include identification of the values or states of the non-alarm system attributes such as operational, environmental, safety, and commercial goals. This benchmarking should also include interviews with operators and operating staff before rationalization.

The second part pf benchmarking is to analyze individual alarm operational performance to see what the alarm is doing on a day-to-day basis and to find individual bad actors or groups of bad actors.

This benchmarking step should not only give alarm system metrics but should identify alarm conditions such as alarm floods, operator alarm overload, standing alarms, transient alarms, nuisance alarms, chattering alarms, redundant alarms, obscuring alarms, alarm (symptom) vs. process or equipment activity, ignored alarms, defeated (shelved) alarms, process issues related to alarms, faulty field devices, etc.

4.  Identify and Analyze Individual Alarms
In alarm rationalization, an important step is identifying all the alarms in the system and their parameters (setpoints, directions, deadbands, test intervals, priorities, etc.).

Identifying individual alarms is only a beginning step because we also need to know the functionality of the alarm, the alarm's relationship to other alarms (correlation), and its relationship to process and equipment conditions. An alarm isn't simply exceeding a limit, then providing a visual and audible response. Defining an alarm like this can get you into trouble fast.

Analysis of the alarm functionality includes determining what the purpose of the alarm is, what the consequences and risk associated with the alarm are, how important the alarm is, what the alarm indicates, how the alarm is organized and presented, what other alarms will be active at the same time (situations where the alarm and other alarms will activate), alarm correlation, alarm patterns, what action is expected of the operator, how will the operator know what to do, how much time is there to detect and accomplish the action (and if it is reasonable), what the expected action to bring the process to the desired state is, process or equipment conditions that affect the desired performance of the alarm, how the alarm fits in to troubleshooting, etc.

Data gathered on the alarm in the initial benchmarking step will be used in this step to characterize individual alarm performance and correlation with other alarms, the process, and equipment for this analysis.

5.  Prioritize Alarms
Here you determine the importance or significance of the alarm through a ranking scheme. This is normally done by risk analysis to determine the importance that the operator detect and perform the expected action when the alarm occurs. Prioritization helps ensure the operator knows the importance of the alarm itself as well as the importance of the alarm in relationship to other alarms.

Figure 2: Prioritize in Proportion

According to EEMUA Guideline 191 (see sidebar, "Few Solid Guidelines"), the number of alarms of each priority--high, medium, low--should be proportioned as shown.

The prioritization scheme is generally limited to the control system's capabilities and any third-party alarm management software on the system. The number of prioritization levels should be kept to a minimum to minimize operator confusion. The number of alarms prioritized in each category (high, medium, low) generally can be visualized as a triangle (Figure 2) with EEMUA Guideline 191 alarm proportions (see sidebar, "Few Solid Guidelines"). Informational alerts should be kept out of the prioritization scheme if possible.

6.  Rationalize the Alarm Database
This workhorse step is where the alarm management procedure and other methods are applied to the information gathered in the previous steps to optimize the alarm system. Other methods can be statistical analysis; alarm filtering and suppression; bad actor(s) identification; logic, event, and fault trees; alarm organization and presentation; operator dialogs; adaptive alarm techniques; and expert systems.

This step considers alarms not only individually but on a system level as well as correlated to process, equipment, and system events and abnormal conditions. This step is further complicated by the fact that the process is not static--the process that the operator and the alarm system see may change based on operating conditions, and case scenarios may need to be developed against which you can analyze the alarm database.

7.  Implement the Rationalization
This is where the alarm rationalization is implemented on the control system. While one might assume this is a simple step of modifying the control system as the result of the previous steps, it is not that simple. Since the rationalization may add or remove alarms; change presentation, organization, or setpoints; add or modify procedures; modify or add training; etc.; it is necessary to have an implementation plan that involves the operators and operating staff and other appropriate personnel.

Failure to involve the operating staff can lead at the least to some upset people or loss of some of your expected gains with alarm rationalization. At worst it can result in a hazardous event due to the operators not understanding the new system (one of the things that alarm rationalization is supposed to improve).