Maintenance Quarterly: At last…Shock Monitoring For Reciprocating Compressors

Although overall vibration trending is an excellent tool for monitoring rotating machinery health, it is not generally effective for monitoring reciprocating compressors. Many typical faults on reciprocating machinery are characterized by mechanical looseness, which results in impacting or shock events in the machine. Impacts producing very short duration pulses in the vibration signal generally have little effect on the overall vibration level. Thus, conventional trending techniques do not detect these types of faults at an early stage. This is a significant problem, since critical damage or even catastrophic failure can occur in a very short period of time after the onset of faults.

Typical faults include:

• Loose or broken bolts
• Loose or cracked rod nuts
• Cracked connecting or piston rod
• Excessive crosshead/slipper clearance
• Excessive clearance in connecting pins
• Liquid or debris in the cylinder
• Scoring in the cylinder
• Other cracked or broken parts

A new generation
Impact transmitters have been successfully used to monitor compressor faults for many years. Their effectiveness, however, is quite sensitive to the shock threshold level selected, since they only monitor shock pulses that exceed a single threshold level. If the level is incorrectly selected (too high or too low), it results in either false trips, or insufficient warning and machine damage. Today, a new generation impact transmitter is available for use with reciprocating equipment. Although it is very sensitive to compressor faults in their early stages of development, it is less likely to give false trips than conventional transmitters. Known as the Reciprocating Machinery Protector (RMP), it improves on existing technology in the following ways:

1. Peak amplitude of routine data that does not exceed a shock threshold level is measured and can be trended.
2. Peaks are evaluated relative to two shock threshold levels, rather than one. This allows more flexibility in setting thresholds, resulting in earlier warning of faults without false trips.
3. Peak counts (i.e., peaks which exceed thresholds) are weighted based on their levels, to help better quantify vibration severity.
4. A “dead time” is used to eliminate false peak counts due to mechanical ringing of lightly damped structures caused by the impacts.
5. Monitoring parameters are programmable, so the process may be optimized for particular machines.
6. The monitor has a higher frequency response than existing units.

Impact data
Consider data taken on both a good and a bad compressor. While we typically would expect to see significant differences in peak amplitudes due to impacts, the overall vibration level would not change enough to reliably detect it. The RMP uses special high-speed peak detection circuitry to accurately measure the amplitude of each shock event that occurs within a preset sample time (typically 12 to 16 cycles of operation) and compares them with two preset shock threshold levels. Based on improved exceedance criteria developed from empirical data, a Reciprocating Fault Index (Rfi) is calculated to help determine machinery health. This index provides a better indication than is provided by conventional impact transmitters.

Overall trend vs. Rfitrend plots
The difference between monitoring overall vibration level and Rfion a reciprocating compressor is shown in Fig. 1. This is a trend plot, over a 60-minute period, that shows both measurements on the same compressor. (Note: Time runs from right to left on this plot.) The Rfitrace appears as a “cityscape” and shows significant increase in amplitude over the 60-minute period. The overall vibration trend, on the other hand, shows little change in amplitude and, in this case, did not trip an alarm. Some level of mechanical looseness is evident, and the Rfitrend shows a worsening condition with the progression of time. The short interruption in the data is a period where the compressor was stopped and then restarted. It is important to notice that when Rfiwas at the highest values, overall vibration level changes were minimal. This clearly shows that overall vibration level alone cannot be reliably used as an indicator for mechanical looseness.

Overall trend vs. Rfitrend plots
The difference between monitoring overall vibration level and Rfion a reciprocating compressor is shown in Fig. 1. This is a trend plot, over a 60-minute period, that shows both measurements on the same compressor. (Note: Time runs from right to left on this plot.) The Rfitrace appears as a “cityscape” and shows significant increase in amplitude over the 60-minute period. The overall vibration trend, on the other hand, shows little change in amplitude and, in this case, did not trip an alarm. Some level of mechanical looseness is evident, and the Rfitrend shows a worsening condition with the progression of time. The short interruption in the data is a period where the compressor was stopped and then restarted. It is important to notice that when Rfiwas at the highest values, overall vibration level changes were minimal. This clearly shows that overall vibration level alone cannot be reliably used as an indicator for mechanical looseness.

Protecting recip machinery
The RMP from IMI Sensors, like the one shown in Fig. 2, is a two-wire device that operates off of standard 24V loop power and has a 4-20 mA output signal proportional to the Rfi. Its output can be connected to a PLC, DCS or SCADA system, as well as to many other standard instruments accepting a 4-20 mA signal. The system used should have either dual relays or display functions and should be set to provide notification when the Rfiexceeds either the warning or critical alarm level. It also may be set to shut the machine down when the critical alarm level is reached.

Monitoring parameters can be factory set, based only on the rpm of the compressor and default settings. All RMP parameters are, however, user adjustable by incorporating an optional USB programmer.

Shock and vibration data are measured using an industrial accelerometer that operates over a wide frequency range, and thus responds accurately to impact events. For ease of installation, the RMP contains an integral accelerometer and is housed in a compact, hermetically sealed, industrial accelerometer-type unit. It is typically mounted to the crosshead or crosshead slipper, using a single ¼-28 mounting stud with sensing axis perpendicular to the piston rod motion. If the compressor does not have a crosshead, the unit is mounted to the crankshaft side of the cylinder. 

Once mounted, the RMP continuously monitors the embedded acceleration signal using a high-speed peak detector. Using an internal microprocessor, it compares each peak detected against low and high shock threshold levels, calculates Rfiand outputs a 4-20 mA signal proportional to the Reciprocating Fault Index.

If no peaks exceed either threshold, Rfiis simply equal to the peak amplitude detected. Thus, if a data logger is used with the system, trending can be implemented. If any peaks in the sample time exceed either threshold, the processor counts them, applies a weighting factor based on amplitude, and computes Rfi. The output of the RMP is typically routed to a meter, PLC or other device capable of tripping warning and critical alarms based on output. Default values for these alarms are provided with the unit.

Real-world performance
A rebuilt six-cylinder compressor was put into service as part of an expansion project in a gas plant. The compressor is driven with a 3000 hp electric motor and runs at 300 rpm. This plant routinely monitors and trends velocity vibration measurements on most of its equipment, including reciprocating compressors. Management decided to install an impact monitor on each compressor cylinder on this machine.

At startup, the transmitter alarm relay tripped and took the compressor offline. During attempts to restart the machine, the impact transmitter again tripped and took the machinery offline. Upon investigation, it was found that the retaining bolts on the high-pressure packing case had not been properly tightened. Had this error not been caught, the looseness would have grown worse and likely have led to catastrophic failure.

Dr. George Zusman is director of Product Development for the IMI Sensors division of PCB Piezotronics, overseeing all research and development of its industrial vibration monitoring instrumentation product lines. He has nearly 35 years experience in industrial vibration monitoring and was formerly director of Engineering, and later, president, of Metrix Instruments, Co./PMC-Beta. Prior to Metrix, he was president & CEO of ViCont Ltd, where he was responsible for all aspects of R&D, sales and customer service.

David A. Corelli is director of Applications Engineering, PCB Piezotronics. His nearly 35 years of experience in vibration analysis and instrumentation includes working as a test engineer for the Air Force Avionics Laboratory and as a field engineer for Hewlett Packard, Entek and IRD Mechanalysis. A Category IV Vibration Analyst in accordance with ISO 18436-2, Corelli serves on the Board of Directors of the Vibration Institute, as well as chairman of its Certification Committee.