Home › Technical articles

Đọc bằng tiếng Việt

Alarm Rationalization Study: alarm philosophy, team, priority matrix and the master alarm database

TDL Industrial Automation · Last updated:

From Rockwell, Emerson, exida and Siemens white papers: what the alarm philosophy must settle, who sits in the team, what to gather, how each alarm is reviewed, building the priority matrix, the pace the sources report, one case study's results, loading the master alarm database into Ignition, FactoryTalk and AVEVA, and checks after go-live.

Alarm rationalization is the stage of the ISA-18.2 and IEC 62682 alarm management lifecycle in which a team reviews every alarm against the site's alarm philosophy, keeps only the alarms that need an operator response, and records each decision in the master alarm database. This article covers how the study is prepared, run and loaded into the control system. The whole lifecycle, the KPI targets and an example rationalization record are in ISA-18.2 alarm management.

An alarm rationalization study, from philosophy to monitoringSix steps from top to bottom. Alarm philosophy: criteria for what counts as an alarm, the priority method based on consequence severity and time to respond, classes, roles, change control, KPI targets. Data and documents: the current alarm list with limit, priority, deadband and delay; P&IDs, PFDs, cause-and-effect charts; PHA and LOPA results, operating procedures; alarm and process history. Team review, one alarm at a time: validity, consequence, cause, confirmation, action, time to respond, priority from the matrix, class, limit, attributes, special handling. Master alarm database: the authorised list, with the basis of each alarm kept. Configuration: import or bulk edit from the database, operator help from the same records. Monitoring after go-live: KPIs over at least 30 days, new bad actors, running settings compared with the database. At the bottom: every later change goes through management of change and then into the master alarm database.An alarm rationalization study, from philosophy to monitoring1. Alarm philosophyrules agreed firstalarm criteria: what counts as an alarmpriority method: severity and time to respondclasses, roles, change control, KPI targets2. Data and documentsgathered before the meetingscurrent alarm list with limit, priority, deadband, delayP&IDs, PFDs, cause-and-effect chartsPHA and LOPA results, operating proceduresalarm and process history3. Team reviewone alarm at a timevalid? consequence, cause, confirmation, actiontime to respond, priority from the matrixclass, limit and other attributesspecial handling, e.g. by operating state4. Master alarmdatabasethe authorised listevery rationalized alarm with its attributescause, consequence and action kept as the basis5. Configurationloaded, not retypedimport or bulk edit from the databaseoperator help built from the same records6. Monitoringafter go-liveKPIs over at least 30 daysnew bad actors reviewedrunning settings compared with the databaseEvery later change: management of change, then the master alarm database is updatedSteps and items paraphrased from Rockwell PROCES-WP015B, Siemens (Process Online, 2014),Emerson and exida (2012), Emerson's Alarm Rationalization white paper (2019) andexida (2021, 2022). Figure: drawn by TDL.
Figure 1: An alarm rationalization study, from philosophy to monitoring. On a phone, swipe sideways to see the whole figure.

1. The alarm philosophy comes first

Rationalization applies the philosophy's rules to each alarm, so the rules must exist before the first meeting. ISA has a technical report for each: ISA-TR18.2.1-2018 (alarm philosophy) and ISA-TR18.2.2-2016 (identification and rationalization). The public sources list what the philosophy should settle:

Item What it settles Source
Alarm criteria What qualifies as an alarm; every alarm needs a defined operator response Emerson and exida
Priority method Priority from consequence severity and time available to respond Emerson and exida, Siemens
Classes, HMI colour How alarms are grouped and how colour shows them Siemens
Change control Including whether operators may disable alarms or change limits from the HMI Emerson and exida, Siemens
Roles, targets Who does each task; KPIs such as acceptable alarm load Emerson and exida, Siemens

For a new plant, Siemens says the philosophy should be approved before commissioning. exida adds that it should define which types of corrective action are valid, and that rules for whole groups of similar alarms, such as diagnostic alarms, let them be handled outside the meeting. Display rules belong to the HMI style guide: ISA-101 explained.

2. Who sits in the team

exida compares rationalization to a HAZOP workshop: a small core team full time, specialists part time. Rockwell cites exida for an optimum size of four to five people; exida adds that more people bring more opinions and slow the work.

Member What they bring Source
Operators Experience of running the unit through the HMI; preferably from different shifts Rockwell, Emerson 2019
Production or process engineer The process, its economics and the control system Rockwell, Emerson 2019
Control system engineer Responsible for alarm configuration; can trace an alarm through the logic Rockwell, exida 2022
Facilitator Knows the method and the philosophy, keeps decisions consistent; no stake in the unit reviewed exida 2022, Emerson 2019
Part time Safety and environment, maintenance, instrumentation; a management sponsor Rockwell, Emerson 2019

exida also gives core members side jobs, among them a scribe who records decisions in the master alarm database and someone who pulls up process history. It calls operator participation essential; the results can later be shown to junior operators in an alarm help display.

3. Preparing the study

Emerson's white paper lists what to gather before the first meeting:

  • for an existing system, a complete list of current tags and alarms with settings such as priority, limit, hysteresis and delay;
  • current P&IDs and PFDs, and cause-and-effect charts for the process, ESD, SIS and fire and gas systems (see cause-and-effect matrix and SIS testing);
  • PHA reports, LOPA results and safe operating limits;
  • critical operating procedures and a list of alarms repeated on external annunciator panels.

Rockwell places rationalization of an existing system after monitoring and benchmarking, and suggests statistical analysis of process history for choosing limits. In exida's case study the benchmark used 30 days of alarm data from two operator positions. Emerson states that filling the master alarm database with all tags and relevant information before the team meets has been shown to double the pace.

4. Reviewing one alarm

Rockwell's nine steps and an example record (Figure 2) are in the ISA-18.2 article. Rockwell stores the result as alarm type, priority, class, limit, operator action, consequence of inaction and need for advanced alarming. exida's list of common mistakes adds rules for the fields:

  • Cause: the root cause, not the symptom ("high level" is not a cause; a closed drain valve is), and only the two or three most likely.
  • Consequence: the direct consequence the operator's action prevents, not the ultimate one from the hazard study; assume other protection layers, such as the SIS or a relief valve, work.
  • Corrective action: it must correct the condition; "investigate" or calling the control engineer does not count, and a vague action may mean the alarm is not justified.
  • Time to respond: from the alarm to the last moment the action still prevents the consequence, not the time the operator should or will take.

Emerson adds removing duplicates, settings for different operating states, and reusing the rationale for similar alarms. Siemens notes that a typical analog point can carry six or more alarms where rationalization may justify one. Keep the basis of removed alarms too: exida describes a loss of containment after a "non-alarm" was removed that operators used to know when to stop filling a tank.

5. Building the priority matrix

The matrix is part of the philosophy. Rockwell's example grades consequence severity separately for personnel, environment and financial loss (none, minor, major, severe) against operator urgency: not urgent (more than 30 minutes), prompt (15–30), rapid (5–15) and immediate (less than 5). Each cell gives a priority; Rockwell recommends no more than three or four levels. Emerson uses the same two inputs, notes that response time is estimated rather than calculated, and points to an example matrix in ISA-TR18.2.2, subclause 7.1.

VanCamp lists three causes of a skewed priority distribution: priorities left at default values, priority used to sort alarms by impact category, and priorities set on ultimate rather than direct consequences, which pushes alarms towards critical. The target distribution is in the ISA-18.2 article.

6. Pace and scheduling

The sources count in different units (tags, control modules or alarms):

Source Reported pace
Emerson and exida, 2012 20–30 or more process tags per day, about 100–200 alarms
exida, 2022 100–200 alarms per day as the common benchmark; 300–400 with good prework; listed projects from 427 to 2,919 alarms per day
Emerson, 2019 80–100 alarm tags per day in pre-rationalization; about 300 modules per day in owner review meetings
exida case study about 27,000 alarms in 60 sessions of up to four hours; up to 900 alarms per day

exida names what moves the pace, including brownfield or greenfield, the facilitator's preparation, the tool, identical units whose results can be copied, and rules in the philosophy. In its worked example, 10,000 alarms at 100–200 per day meant 50–100 days; narrowing to 3,000 enabled process alarms, with diagnostic alarms handled by rule and credit for duplicate equipment, cut the estimate to 10 days. Emerson budgets up to eight weeks per typical unit for pre-rationalization, then about a week for the owner's review.

7. What the sources report as outcomes

exida's case study of a specialty chemical plant, which finished the study itself after training and initial workshops with exida, reports:

Measure Before After
Enabled alarms 17,188 4,773
Configured priority, high / medium / low 85.5 / 12.3 / 2.2 % 5.6 / 21.2 / 73.2 %
Time at no more than 1 alarm per 10 min, operator 1 / operator 2 67.6 / 52.5 % 77.5 / 82.1 %

The plant also moved 3,141 alerts and prompts to a separate list. These are one plant's results as exida reports them, not a forecast.

8. From the master alarm database to the configuration

Rockwell defines the master alarm database as the authorised list of rationalized alarms and their attributes, and names PlantPAx Alarm Builder among the tools for exchanging data between it and the control system. Emerson and exida say it can be a spreadsheet or a commercial tool, and that changes made in the control system must be written back to keep both in step. Siemens recommends importing settings instead of retyping them, and notes that a control system with user-defined fields can hold the database itself. The platforms document their own bulk paths:

  • Ignition 8.3 imports tags from JSON, XML or CSV; the CSV format carries no alarm configuration.
  • FactoryTalk View SE 16.00 help uses the Alarm Import Export Wizard to import an XML or Excel file into a Tag Alarm and Event Server.
  • AVEVA Application Server 2026 exports instances to CSV, one column per configurable attribute, for editing in Excel and re-import, optionally updating only changed attributes.

In exida's case study the changed parameters were loaded from a configuration file through the control system's bulk edit, and alarm help faceplates were filled automatically with cause, consequence, action and time to respond. Shelving, journals and KPI reports per platform: ISA-18.2 alarm management software.

9. Verifying after go-live

Siemens bases the metrics on at least 30 days of data and recommends comparing the running configuration with the master alarm database regularly, to catch changes made outside management of change. Emerson and exida suggest a monthly review for new bad actors, one subsystem at a time, most critical first. KPI definitions and targets: ISA-18.2 article; the data comes from the alarm log or a process historian.

What we check first

This is TDL's practice, not a requirement of ISA-18.2.

  • We check that an approved alarm philosophy with a priority matrix exists before any meeting is booked.
  • We export the alarm configuration from the running system and compare it with what operators see.
  • We check how long alarm history is kept and whether it shows what reached each operator position.
  • We ask which P&IDs and cause-and-effect charts are current.
  • We agree with operations which operators can join, from which shifts, for how many sessions.
  • We check how the target platform imports alarm settings, so the database loads without retyping, and list alarms from packaged units and the SIS.

How TDL approaches it

TDL Industrial Automation is a System Integrator in the Rockwell Automation PartnerNetwork, a System Integrator member of the AVEVA Partner Ecosystem and an Ignition Certified Integrator. Related services: SCADA and historian and ICSS and safety systems. The alarm lifecycle in a SCADA replacement: SCADA upgrade with ISA-101 and ISA-18.2.

Frequently asked questions

What is an alarm rationalization study?

It is the stage of the ISA-18.2 and IEC 62682 lifecycle in which a team reviews every alarm against the site's alarm philosophy, keeps only the alarms that need an operator response, and records each decision in the master alarm database. The settings are then loaded into the control system and checked after go-live.

Who should take part in alarm rationalization?

A core team of four to five people: operators, preferably from different shifts, a production or process engineer, a control system engineer and a facilitator with no stake in the unit. Safety, environmental, maintenance and instrumentation specialists join part time.

How many alarms can a team rationalize per day?

exida gives 100–200 alarms per day as the common benchmark and 300–400 with good prework, and lists projects from 427 to 2,919 alarms per day. Scope, preparation, the tool and identical units whose results can be copied make the difference.

What is a master alarm database?

It is the authorised list of rationalized alarms and their attributes, such as type, priority, class, limit, operator action and consequence of inaction. It can be a spreadsheet or a commercial tool, and settings should be imported from it into the control system instead of being retyped.

How is the alarm priority matrix built?

One axis grades the severity of the direct consequence for personnel, environment and financial loss; the other is the time available to respond, in bands such as more than 30, 15–30, 5–15 and less than 5 minutes. Each cell gives a priority, with no more than three or four priority levels.

Sources

Read 15 November 2026. This article paraphrases the sources and does not quote them; read the original documents before designing or working on a system.

Related services:
ICSS & safety systems SCADA & historian