Intelligent alarm management saves industry $$$

InTech, Oct 1997 by Hawman, Chris

Intelligent alarm management systems provide operators the right information to improve abnormal situation management.

Recent studies suggest that U.S. industry loses more than $20 billion per year to preventable process disruptions. Globally, the productivity impact is much greater. For this reason, the process industries are taking a considerable interest in abnormal situation management (ASM).

Much attention is being focused on systems that provide operator decision support and troubleshooting assistance during process disruptions. To respond to process disruptions, operations personnel often manage high volumes of process data to quickly filter nuisance information, diagnose the root cause of an alarm condition, and implement a corrective action. There is a demand for process control technologies that assist operators in analyzing process information and implementing corrective control strategies when abnormal situations occur.

These issues point to a demand within the process industries for intelligent alarm management solutions. This article explores applying expert systems to provide a powerful framework for alarm management. Expert systems add value to alarm systems by diagnosing process disruptions, generating more precise operator information, and assisting in implementing corrective actions. The result is a more robust alarm management system that helps operations personnel manage process disruptions safely and efficiently and reduces the impact on productivity.

Expert systems "reason"

Expert systems are applications of artificial intelligence (Al) that apply inference engine and fuzzy-logic technologies to reason in real time about events that occur in dynamic processes. The inference engine reasons about rules defined in a knowledge base that is derived from process experts' heuristics and domain-specific knowledge.

Functionally, real-time expert systems can be divided into two categories: advisory and supervisory control. In advisory applications, the expert system reasons about dynamic changes in process data, makes decisions based on process events, and presents conclusions and a rationale to the operator. The expert system acts as a supervisor with the ability to provide timely, accurate advice to the operator about process occurrences and impending problems. In more robust supervisory control applications, the expert system proactively assists operators by adjusting set points and switching discrete equipment on or off to resolve problems, optimize productivity, or attain other objectives defined in the knowledge base.

When applied to alarm management, expert systems can be integrated alongside an operator interface (OI) to supervise the process and provide online advisory or supervisory control functionality The expert system reasons about alarm points defined in a real-time database according to absolute tag values, dynamic changes among a group of tag values, or the state of other alarm points in the database. The flexibility to introduce meta alarms (alarms that reason about the state of other alarms) enables users to apply a higher level of logic that captures the dynamic changes occurring in their process environment.

Nuisances common in conventional systems Conventional alarm systems, which are typically advisory in nature, are tag-centric. They provide discrete alarms based on an absolute limit that is defined for a single process value. They mirror the function of earlier microprocessorbased annunciators by generating operator messages, annunciating the alarm condition, and logging the event with a date and time stamp.

The alarm is typically identified as HiHi, High, Low, or LoLo to advise the operator about the alarm condition urgency. However, this identification does not take into account the current process state (e.g., start-up, normal, or shutdown) and, as a result, nuisance alarms are common.

Discrete alarm systems also tend to point toward the symptoms of an abnormal situation instead of the root cause, which reduces the diagnostic assistance available to the operator. There is also no capacity for reasoning about the state of other alarm points in the database. As a result, alarm showers caused by a single parent alarm are possible, which flood the control console and seriously impair the operator's ability to correct the root process disruption. This presents safety risks and also reduces the operator's ability to efficiently return the process to a normal operating state.

Expert systems advance the state of the art in alarm management and add value to conventional discrete alarm systems by introducing advanced alarm management concepts, including sensitivity to process state, reasoning about alarm persistence, meta alarms, expanded concepts of alarm priority, and alarm inhibition. They also provide operators with intelligent assistance in managing abnormal situations by taking a user-definable action on alarm entry, acknowledgment, or exit.

With expert systems, users can, within their alarm management strategies, take advantage of the reality that in a dynamic process environment alarm priority is dependent on process state. For instance, a low-tank-level alarm may be critical during normal operating conditions but may be an anticipated condition during process start-up.