Oct-2017

The filter setting challenge

Important considerations when using crosshead acceleration and frame velocity to monitor and protect a reciprocating compressor

Robert Eisenmann, Jr. BP Refining Technology and Engineering
Oliver Franz Prognost

Article Summary

History has shown that within the fleet of reciprocating compressors numerous catastrophic failures result in significant losses. Although adequate machinery protection systems are available, they are not always in place or applied correctly to mitigate catastrophic failures.

API 670 5th Edition Machinery Protection includes an informative annex aimed to define the key components of reciprocating compressor protection and condition monitoring systems. A crucial aspect was to include crosshead guide acceleration in addition to the frame vibration shutdown recommend by API 618 5th edition. While these industry standards recommend monitoring and protection points they do not fully explain how the transducers should be monitored.

Numerous case studies with several different failure modes identify that appropriate frequency filter setting has major impact on the machinery protection systems ability to function as needed when critical conditions arise. Additionally, the application of RMS based analysis instead of peak analysis is also discussed.

This article discusses the most important aspects to consider when implementing machinery protection systems using crosshead acceleration and frame velocity on reciprocating compressors – specifically:
• API670 5th edition – Requirements and the Frequency Filter Setting Challenge
• Signal analysis consideration – Filter Setting / RMS / Peak
• Crosshead acceleration – Why - Where applied – Effective analysis approach
• Frame velocity – Why - Where applied – Effective analysis approach
• Conclusion

Knowledge surrounding proper low pass filter settings for acquisition systems performing critical shutdown function is very limited and often misapplied. This paper illustrates why the common practice of setting low pass signal filters at 2 kHz introduces risk that serious failure modes go undetected compromising plant safety, health and the environment.

Until the API 670 5th edition task force was formed, reciprocating compressors were not a major focus of the standard. This was rectified with the release of 5th edition and the inclusion of informative Annex P on reciprocating compressor monitoring.

Annex P recommends crosshead guide vibration as a shutdown parameter and monitoring frequencies (0-7 kHz) to detect relevant mechanical impacts over the minimum bandwidth of 0-2kHz.

Based on the above API recommendation which frequency range would you set the low pass filter for crosshead guide acceleration to provide the earliest and best representation of the compressor health? Figure 1 is an example of crosshead guide acceleration data from 0-7kHz during an event showing what data would be seem if you monitor only 0-2 kHz verse 0-7 kHz.

In the case studies investigated, we can conclude that frequency filter setting is vital in detecting failure modes such as wrist pin seizures, developing cracks in pistons and piston rods.

Signal Analysis Considerations
RMS: Modern machinery protection systems apply a fully continuous on-line RMS analysis of focus signals such as crosshead acceleration and velocity. This is a beneficial approach specifically on reciprocating compressors where an early indication of damaging impacts is of the essence and RMS values are best to describe the energy within a given signal as calculated by the equation in Figure 2.

While the proper representation of contained energy is a positive factor, there is a risk that individual, high XY2 data samples get lost in the average. This is specifically the case if the number of total samples “n” in the evaluation period is high e.g. when the Xrms equation above is evaluated over one entire revolution of the rotating machine or one second eventually containing hundreds or thousands individual data samples. Furthermore the reliable RMS computation and alarming in real time requires modern, redundant CPUs to handle this significant processor workload eventually occurring on multiple sensor channels in parallel.

Peak analysis: Some systems take a different path by employing peak analysis instead of RMS. This addresses the effect that RMS analyses may undervalue high individual samples when evaluation periods are relatively long (e.g. 1 compressor revolution or 0.2 sec @ 300 rpm) and compares maximum values detected in that period against a set of Alert and Shutdown limits.

Unfiltered peak analysis however leaves users vulnerable for nuisance alarms, e.g. caused by isolated, high frequency events, non-repetitive signal spikes and sensor glitches.

In order to address these nuisance alarms some users apply a low pass filter (e.g. @ 2 kHz) so only the 0-2 kHz frequency content is analysed for its peak vibration content.

While the above strategy reduces nuisance alarms when using peak analyses it also eliminates capability to detect many critical failure modes containing majority damaging energy in higher frequency ranges as will be discussed next.

Crosshead Acceleration: Looking at the working principal of reciprocating compressors the crosshead clearly is a focal point. Here, the rotating movement of the crank-shaft is transformed into a reciprocating (linear) movement of the piston rod. It is the central component where the major drive forces are transferred from the running gear to the crosshead and ultimately the piston rod assembly. In order to contain these forces into the right direction, the crosshead travels within the crosshead guide. The crosshead guide is the most direct connection of the running gear to the frame/crosshead guide and therefore is the best position to install vibration sensors. Therefore, API 670 5th edition – (Annex P.4.3.4.4) recommends crosshead accelerometers should be mounted in the vertical direction on the top or bottom of the crosshead guide as shown in Figure 3.

Practical examples and case study material
Case Study 1 – Seized Wrist Pin: During the commissioning and start-up of a new API 618 compressor in H2 service the machine was suddenly tripped by the machinery protection system. A first data review revealed that crosshead acceleration amplitudes reached the default protective limits having saved the asset from consequential damage or loss of containment. During a detailed analysis of the high resolution data available this first case study is an excellent example illustrating the importance of high frequency (0-7kHz) data for effective machinery protection with crosshead acceleration sensors.

The wrist pin seizures detected did not involve true mechanical impacts typical of loose components (showing lower frequency content primarily below 2 kHz). Comparison of Figure 4 “Good Condition” to Figure 5 “Bad Condition” shows the majority of failure related energy as well as relative signal change has higher frequency content above 2 kHz. In order to initiate the trip function prior to a catastrophic failure and potential loss of containment the full 0-7 kHz frequency spectrum suggested by API 670 must be monitored.