Dec-2015

Rapid response

Flow (dP) transmitter speed of response to meet API 670 5th edition requirements.

Medhat Zaghloul
Compressor Controls Corporation

Article Summary

The new API Standard 670, that covers Machinery Protection Systems, was updated in the 5th edition in November 2014 to include a section on surge detection for centrifugal and axial compressors.Moreover, in the standard’s informative Annex K, there is a well-written distinction between an antisurge control system and a surge detection system.

One common aspect of both the surge control and surge detection applications is the selection of the differential pressure transmitters used to provide the compressor flow signal, and which must have a sufficiently fast response. In some cases, the two systems (surge detection and surge control) may share the same transmitter signals.

This article analyses a typical behavior of the differential pressure transmitter signal from a flow measuring device during compressor surge and derives practical guidlines for selecting adequate transmitter dynamic response characteristics.

Antisurge Control System
API mandates the installation of an antisurge control system on all axial compressors and most centrifugal compressors. The main function of the antisurge system is to reduce the possibility of machinery damage due to surge events.

Most modern antisurge control systems incorporate a proximity-to-surge calculation algorithm that uses a variety of process variable transmitters around a compressor stage and produces a measure of where the operating point lies with reference to the Surge Control Line.

If the operating point approaches or crosses the Surge Control Line, the antisurge control system is usually designed to modulate (open) an appropriately sized and located antisurge valve that either recycles gas around the compressor stage or blows it off to the atmosphere.

When the compressor stage’s operating point moves back to the right of the Surge Control Line the antisurge control system is usually designed to allow the antisurge valve to be ramped closed in order to eliminate the energy waste due to unnecessary recycle or blow-off.

In order to reduce energy costs associated with recycle and blow-off (when it becomes necessary), the owners/operators of turbocompressors should demand that the antisurge control system include provisions to keep the Surge Control Margin as small as is required while still ensuring adequate protection against surging.

Surge Detection System
In the event that the antisurge control system fails to protect the compressor from surging, API mandates that an independent surge detection system be installed for all axial compressors and, if specified, be installed for centrifugal compressors.

In contrast to an antisurge control system (where the proximity-to-surge is usually the main variable that is calculated), a surge detection system, according to the API standard article 9.4.1.1.1, “shall be capable of detecting each surge cycle”.

In article 9.4.3.1, the new API standard mandates that an alarm output shall be generated whenever a surge (cycle) is detected.

Optionally, and if specified, the surge detection system may be required, as per article 9.4.3.2, to initiate further actions, such as the fast opening of the antisurge valve modulated by the surge control system, or the shutdown of the main driver. These “further actions” should be initiated after a specified number of surge cycles have been detected within a user-defined time window.

The Flow Transmitter
In both the antisurge control and surge detection systems, the signal that represents volumetric flow through the compressor plays a very important role. The purpose of this article is to discuss various aspects of this signal, as generated by modern digital transmitters.

Figure 1 shows an analog recording of the differential pressure (dP) signal from a flow measuring device during surge, covering a period of about 4 seconds in which two surge cycles occur.

The analog recoding shows that the dP signal drops from about 71.2% of span (or 0.712 normaised) to zero in about 67 ms, i.e an equivalent drop of approx. 106% per 100ms (or 1.06 normalised).

Let us assume that any sudden drop in the flow signal of more than 20% per 100 ms (or 0.20 normalised per 100 ms) is the threshold value used to deduce that a surge event is happening. This is shown as the red lines in Figure 1. This setting should provide the logic solver, for surge detection purposes, with a sufficient margin so as not to produce spurious surge events, which could cause nuisance trips of the compressor train.

Most modern flow transmitters are digital devices which produce an output value every time the actual process variable is sampled (as illustrated in Figure 2).

The first factor we will consider in evaluating a flow transmitter that is appropriate to detect surge is the sampling rate and whether it has any significant impact on the usefulness of its digital values in terms of both Surge Control as well as Surge Detection.

Figure 2 shows the same Surge Flow signal profile, but sampled at a rate of 5 samples per second and sent as a digital signal value with no damping or dead time.

Most controllers calculate a derivative value based on several samples, and using various technioques that are well-described in lterature. When the discretisation of the signal approaches the useful frequency in the signal, the controller cannot accurately reconstruct the signal.
Discretisation of 5 samples per second leads to a significant loss of resolution and sensistivity in the calculation of the derivative of the signal value.

Thus the system (logic solver) will not be able to detect a rate of change that is faster than that which occurs within the discretisation period (which for 5 samples per second is 200 ms). In this case the system cannot detect a rate of change that is faster than 71.2% per 200 ms or 35.6% per 100 ms (0.356 normalised per 100 ms).

With the digital transmitters available in the marketpkace that are suitable for industrial flow measurement using differential pressures, there is always some amount of signal lag. This lag is dominated by a first order lag component  that renders the digital signal profile somewhat different than what is illustrated in Figure 2.