_JPG.jpg)
Data Visualization for Alarm Rationalization results
December 29, 2022
Alarm rationalization is the most important step in the ISA 18.2 Alarm Management lifecycle. This article aims to supplement the Youtube webinar and explain in detail how the Visualization of the Alarm Rationalization Workshop results can be fed into the detailed design of the Alarm System.
ISA 18.2
Management of Alarm Systems for the Process Industries
ISA 18.2 is the industrial standard for Alarm Management in Process Industries. Process Industries are manufacturing processes concerned with processing bulk resources into various products. They are usually related to chemical processes, i.e. petrochemical, oil and gas, food processing, and pharmaceutical.
The main problem with the current industry practice is the overloading of alarms for the operators.
Unlike hardwired alarms back in the 50s and 60s, alarms now exist on the software layer in an industrial control system. On top of that, there is a false belief that the more the alarm, the safer the industrial plant will be, industrial control system manufacturers usually install thousands of alarms by default.
Without proper organization and management of the alarm system, we can potentially end up in a very dangerous situation. One very notorious example is the Deep Water Horizon.
Please check out the wiki page below for the full story:
Rationalization
The flowchart below partially shows the Alarm Management Lifecycle based on ISA 18.2.
This Data Visualization is focusing on the results of the Rationalization process.
According to the ISA:
Rationalization encompasses several significant activities, including alarm justification, documentation, prioritization, and classification. In justification, existing or potential alarms are systematically compared to the criteria for alarms set forth in the alarm philosophy. If the proposed alarm meets the criteria, then the alarm type, set point, cause, consequence, and operator action are documented. The alarm is prioritized and classified according to the philosophy. Classification encompasses assigning alarms to a group or class defining certain administrative requirements.
The result of the Rationalization is a Database containing the configuration for every alarm in the control system. The Data Visualization process will be creating infographics based on the Database.
Tableau
Despite the fact that Industrial Control Systems can produce charts based on alarm configurations, the functionalities are very limited. I appreciate the versatility of Tableau in bringing complicated matrices into clean presentable visualizations, which can effectively communicate the pending actions and steps to the stakeholders and the higher management.
Priority Distributions
The priority is used to inform the operator of the importance of the triggered alarm. The distributions are looking at the proportional ratios between different priorities.
The golden rule for Priority Distributions is to control High and Medium priorities alarms to a limited number.
Priorities for the whole plant
The Industrial System for this project is a Delta V system. They use:
Critical for High Priority
Warning for Medium Priority
Advisory for Low priority
At the top left concern of the Tableau worksheet, it shows the overall priority distribution over the whole plant. It contains information on the distribution of each priority level and their targets set out by the Philosophy document.
The circular bar chart focuses on a new parameter I created in the database - Alarm Priority Number.
The Alarm Priority Number is the summation of priorities in each subunit.
For each priority, I assign the following number:
Critical as 3
Warning as 2
Advisory as 1
The Alarm Priority Number is, therefore, a combined indicator for both the number of alarms and the level of priority. We will be focusing on the units with high Alarm Priority Numbers.
Based on this database, the most problematic units are:
MEK
ROSE
Boilers
These units are isolated for studying so that new action plans can be derived for Detailed Design.
The histogram above shows the subunits of MEK, ROSE and Boilers.
We want to identify the problems in these units and derive strategies to troubleshoot.
There are two main strategies:
Implementing Advanced Alarming Techniques
Reviewing Alarm Rationalization for the specific unit
In this project, I propose:
For MEK-Main, ROSE-Main, and Boiler D, we need to assess, within each subunit, if the alarms with a higher level of priority belong to specific equipment or linked cascading chemical processes. Then we need to group the alarms together and when one of the alarms in the group is triggered, the rest of the alarms in the group are suppressed. There are other kinds of Advanced Alarming Techniques in the ISA 18.2, such as Eclipsing and State-Based Suppressions. Since the focus of this project is on data visualization, I will not dig too deep into Alarm Management arsenal.
For ROSE-LOOP, these are alarms related to the automatic functions in either the Basic Process Control Systems or Safety Systems. They do not need any actions from the operators. Therefore, we need to revisit the Rationalization and reassess their priorities.
Alarm type
Alarm type in the process industry indicates the functional basis of the alarms in the Control System. It is an efficient way to look at the makeup of certain alarm priorities.
The screenshots below show the relationship between alarm types and different kinds of alarms. I used two different kinds of presentations:
Word clouds for "real" alarms
Bubble charts for non-alarm priorities
 |  |
|---|
Real Alarms
In most control systems, priorities encompass all kinds of data points, no matter if they are "alarms" or "notifications". For example, "interlock", "bypass", and "advisory".
By "real alarms", we are looking only at data points with Advisory, Warning, and Critical priorities. Therefore, these alarms are mostly process alarms (high or low alarms) or any alarms which require operator actions.
Non-alarm priorities
Interlocks
Interlocks indicate the active state of an automatic interlock. They do not require actions from the Operator. Therefore, they should only exist as notifications in the system.
Usually, High High alarms and Low Low alarms are the activation threshold for interlocks. The rationalization result matches the assumption.
No alarm
No alarm is a data point in the database indicating the alarm is removed during the rationalization.
High High and Low Low Alarms are usually either duplicates of High or Low Alarms. They are sometimes interlocks too. Therefore, these alarms being removed from the system are compliant with the Alarm Management principles.
There are also a number of High, Low and Deviation Alarms removed from the system. Deviation alarms are alarms associated with any process (High/Low) alarms. This implies it is highly probable that this is the first time the site has done the alarm management.