Reducing alert load via data analytics

Technology Update: Cloud-based data analytics can help plant operators focus on the emergencies that matter and improve overall productivity as well as reduce plant downtime.

By Ute Messmer August 5, 2017

If control systems can identify potential issues early, power plant operators can take action to avoid downtime and remain in compliance. However, achieving this is far from easy.

It doesn’t take much for operators to become overwhelmed by the alerts and event notification coming from sensors and operational components. As technology has evolved, the number of sensors added to industrial equipment has mushroomed. While in the early days there were only a few dozen sensors inside a gas turbine, there are now thousands-and the number continues to grow. The advent of the Internet of Things (IoT) means more sensors are being added to every component, machine, and auxiliary system.

These sensors will intensify an already overwhelming flood of alerts appearing on the screens of operators and plant managers. In one case, a utility had to deal with almost 300 alerts per hour. This is far from an isolated incident and such alert rates are not uncommon.

The bulk of these alerts are triggered by relatively unimportant factors. Yet operators must respond to all messages, even if they do no more than note the alert and take no further action. This absorbs too much of their time and makes it harder to address critical issues. 

Reduce alert volume

It is necessary, therefore, to free up the time of skilled personnel. The best way to do this is by reducing the number of alerts. This would enable them to zero in on priority action, thereby avoiding trips and load restrictions.

Some basic tools can support engineering staff in separating critical alarms from background noise. The lack of automation and analytics support means engineering time continues to be absorbed in evaluating the appropriate response to event notifications. By combining data from three sources, however, it becomes possible to analyze data accurately and hand engineers the information they need to make informed decisions far more rapidly.

These three data sources are:

Archive data. Historical data, based on an archive of alerts, can assist in matching current event notifications with underlying causes. While some monitoring support tools use this data today, its usefulness is limited when used in isolation.

Engineering data. Engineering data plays an essential role in successful plant operation. In the hands of skilled specialists, it enables them to operate complex systems. Yet such data typically remains confined to functional silos—it is available only to a few trained technicians. That has made it difficult to analyze and cumbersome to apply as a practical aid in decision-making. However, it is possible for engineering data to be automatically translated into other formats and displayed graphically to enhance understanding of ongoing alerts and notifications.

Knowledge base. Machinery original equipment manufacturers (OEMs) often have access to massive knowledge bases gathered from thousands of power plants and turbines running worldwide. Some of information data dates back many decades. Intelligent use of analytics enables the matching of event notifications and alarms with similar events logged elsewhere. 

The value of data analytics

Many systems on the market make good use of one of these three data sources. They offer operators valuable insight, but are limited in scope. Similarly, there are a few systems available which seek to combine two of these data streams, some more successfully than others. But integrating all three sources is a new idea.

Even then, bringing together archival, engineering, and knowledge base information is not enough. That only compounds the problem because of the added complexity and increased data volumes. Such an approach may help to identify real situations and determine root causes, but it would be a slow, manual, and laborious process. By the time the data had been combined, filtered, and evaluated, it would be too late to mitigate damager and avoid downtime.

The key to improved maintenance efficiency must go beyond the identification of root causes and be capable of reaching conclusions on time. It must also encompass a substantial reduction in alert volume while providing a means of effectively responding to those alerts deemed to be of the highest priority.

Cloud-based data analytics can provide that solution for engineers. When sophisticated analytics engines are used in tandem with the power of modern microprocessors, huge volumes of data can be synchronized, compared, and evaluated. This opens the door to alert volume reduction as well as faster and more accurate decisions.

Data analytics are now available that offer an accurate view of why event notifications are happening, and the precise causes involved (see Figure 1). This makes it possible to confirm what engineers might have suspected but could not prove, while offering previously unattainable insight.

A simple alert-load example concerns scheduled maintenance. Equipment downtime can lead to a major spike in the number of alarms. As generation systems move out of the established temperature norms when they cool down, a large number of components, systems, and machines suddenly will flood plant operators with alerts and notifications. Culturally, it may take a while for those involved to figure out that all of these notifications can be ignored and that routine maintenance is taking place (Figure 2).

Data analytics also can reveal what really happens when indicators and metrics move above or below normal performance levels. Such events may or may not be important. It takes detailed analysis based on multiple data sources to determine whether movements of this kind are a true cause for concern. If that is accomplished rapidly enough, operators have the time to take effective remedial action. 

Example of reducing alarms, workload

A large power generation company in Germany discovered that notification messages were being triggered by many different signals. In one case more than 1,600 notification were received, at a rate of 281 alerts per hour, yet as few as 10 signals were responsible for 43% of those alarms.

Worse, many alerts continued to take place during periods of routine shutdown. This forced maintenance staff to continue to monitor units that were offline and going through planned downtime. This wasted resources and deflected attention from more urgent maintenance priorities. By zeroing in on those 10 signals and dealing with the thousands of alarms, the maintenance staff’s workload dropped significantly.

Digital alert management highlighted which signals led to alarms and why. Visualization techniques presented essential management information in an easily digestible format. Engineers turned this analysis into action, addressed the correct problems, and slashed the number of event notifications by 90% (Figure 3).

Digital alert management eliminated these unnecessary notifications. Additionally, it generated practical recommendations for specific actions for optimizations. These ranged from the delay of notifications for a few seconds to altering certain manual procedures. Overall, these changes exerted a transformational impact on operational efficiency. That, in turn, improved compliance and profitability, while streamlining the actions of skilled engineers.

Digitization and the future

Digital alert management is an example of applied automation using analytics based on different sources of relevant data. When combined with hands-on expertise, this software brings plant operators into a whole new realm.

Ute Messmer is the head of the Digital Technology Center at Siemens Power Generation Services. Edited by Chris Vavra, production editor, Control Engineering, CFE Media,


Key Concepts

  • Many alerts that operators receive on the plant floor are low priority and do not require immediate attention.
  • Cloud-based data analytics can help operators prioritize alarms and focus on the emergencies that matter.
  • In one example, a company used cloud-based data analytics to reduce the number of event notifications by 90% and improve overall productivity.

Consider this

What other methods can engineers use to reduce the number of alarms they receive and improve productivity?