The occurrence of surge is inferred from a rapid change in any measurement that is a direct function of compressor flow or head. Compressor flow, discharge pressure, and inlet temperature are signals that can be used for surge detection. However, based on experience, suction pressure can also exhibit surge induced oscillations.

The Surge Detection System (SDS) should have several analog input channels available that can accept any of the above measurements. Since compressor flow dP will exhibit the most telling surge signature in almost all applications, the flow measurement should always be considered for surge detection.

This article contains excerpts from the paper, “Compressor load sharing control and surge detection techniques” by Wayne Jacobson, Medhat Zaghloul, Michael Tolmatsky and Serge Staroselsky of Compressor Controls Corporation, as well as Jeff McWhirter, an independent consultant.

Other measurements such as discharge and suction pressure or suction temperature are selected based on the evaluation of the specific installation. In some instances, flow measurement may not be available. In these cases, motor current or power may be used instead of flow dP, provided that the signal produces a measurable surge signature. The SDS needs to accommodate any combination of these signals.

The SDS must be able to measure the rate of change for each analog input channel.  Each channel will have a rate of change threshold that represents a surge condition. The rate-of-change thresholds could be negative or positive to depending on the observed surge signature. In some instances, a compressor may not recover after a surge cycle, due to, for example, failure of the check valve, or mechanical damage. In order to protect against such eventuality, the SDS should also include minimum flow protection. If flow dP remains below configured minimum level after the first cycle, the surge counter is incremented every configured time period. The time period is an estimate of the surge cycle duration.

The SDS should be flexible on how it uses the different analog inputs for triggering surge detection.  Depending on selected surge detection mode, detection may be triggered by monitoring either one or a combination of its analog input signals.  When a combination of inputs is used, surge can be detected when any one input exceeds their rates of change threshold or only when all inputs exceed their individual rate of change thresholds within a specified time period. When using the mode where any one input is used for surge detection, a specified time period between surge counts is needed in order to avoid duplicate surge counts from the different inputs.  If more than two inputs are used, more complex arrangements can be designed into the system.

Detection should be disabled during compressor startup and shutdown in order to avoid false detection at low signal levels, where the machine characteristics are not well defined.

It is imperative then that SDS can differentiate between normal process changes, signal failures, and surge. For this purpose, using a combination of signals listed above can increase reliability of the detection. For example, concurrent reduction of flow and discharge pressure on a running machine signifies surge with a larger degree of certainty than just a flow reduction.

The SDS shall be equipped with two surge detection counters. Each time a surge is detected by the SDS, two counters are incremented. One for counting the number of surges during a particular surge event and one for accumulated surge counts over time.  The event surge counter resets to zero when the surge alarms are reset.  The accumulated surge counter is not reset when the surge alarms are reset and is the total number of detected surges.

A surge alarm is required on the detection of the first surge cycle. Additionally, a Surge Event should be generated every time the surge count is incremented. It is recommended that an additional alarm for excessive surge is generated if the event surge counter exceeds a configurable threshold within a configurable time (for example 3 surges detected in 10 sec).  The excessive surge discrete output is typically used as a condition to trip the compressor.

For centrifugal compressors, the excessive surge threshold is typically set to three surge counts within 10 to 20 seconds.  Axial compressors may be set to trip on the first or second surge detected depending on the compressor manufacturer’s requirements.   Surge and Excessive Surge alarms as well as Surge Events are captured in a log that is viewable by operators. In addition, the log needs to include parameter values and their rates of change at the time of the alarm/event. It is desirable to have surge event data accessible for external analysis.

An Event Recorder is necessary to archive each surge event and store the data in a buffer. At a minimum, the data consists of input channel values and their rates of change as well as the Surge and Excessive Surge alarms. The data should be recorded with a sampling rate of no more than 20ms.  The recorder should capture the data at least 10 seconds before the event and 10 seconds after the event. For each recorded event archived, a file is generated which allows the data to be viewed in a standard format, such as a csv file and can be uploaded to a computer.By TMI Staff & Contributors · On September 18, 2019