Niagara Frontier Vibration Institute Chapter meeting Nov. 21, 2002
Impact Measurement for Reciprocating Compressors
Dr. George Zusman
Metrix Instrument Co.
1711 Townhurst Drive
Houston TX 77043-2899
Jon E. Palm
Metrix Instrument Co.
13094 Old Barona Road
Lakeside CA 90240-1537
Abstract:
While vibration is an excellent, and in many cases, sufficient tool for diagnosing the health of rotating equipment, there are certain reciprocating machine health issues that will not affect the overall machine vibration until damage has progressed beyond any early detection stage. This paper includes actual data that compares vibration and “impact” measurements on reciprocating compressors which illustrates the difference in these two measurements.
Keywords:
Impact: Action resulting from a loose mechanical connection between two parts on a reciprocating compressor “hammering” together at load reversal.
Transmitter: Self contained electronic circuit designed to convert a measured physical parameter into a standard 4-20 mA signal for connection to a PLC, DCS or other data logging instrument.
Introduction:
For years we have been applying rotating machinery vibration monitoring and analysis methodology to reciprocating engines and compressors. Our results have been less than satisfactory and even confusing at times. There are certain reciprocating machinery abnormal operating characteristics that can not be detected at an early stage. As a result they are not seen in the overall vibration level until damage has progressed beyond the simple adjustment stage.
The actual machine data reviewed in this paper compares vibration and “impact” measurements on a reciprocating gas compressor that clearly illustrates the difference in these two measurements. Mechanical problems such as looseness, cracked or broken parts, liquids in the process, and other causes, can be reliably detected by monitoring the impact that occurs due to these mechanical failures. While vibration is still a concern, mechanical looseness is of greater concern because the damage potential is significant.
Impact Measurement:
Impact monitoring has been used successfully to find mechanical looseness on large reciprocating compressors in recent years using empirically determined criteria. Mechanical impacts produce short duration, high amplitude “spikes” when measured using a piezoelectric accelerometer. These spikes are blurred by the traditional vibration monitoring techniques because of the signal processing methods and time constants that are used to accurately measure vibration. To measure impact, a special peak detection circuit is required that is has a fast rise time and will not blur these signal spikes or impact events.
Figures 1 and 2 show examples of signals measured using an accelerometer on a compressor cylinder. The signals were acquired using a transient capture mode on a spectrum analyzer. In the first case there is no significant impact events. This would be considered a normal time waveform for a compressor cylinder. Short duration, high amplitude spikes are present in the second case which is a direct result of mechanical
looseness.
Figure 1, Compressor cylinder time waveform with insignificant impact
There is no correlation to compressor piston position for these captured signals although looseness is more likely at points of rod reversal. For compressor protection the impact measurement is not synchronized with crank angle. For compressor analysis it is very important to synchronize the acquired data with rotation.
Figure 2, Compressor cylinder time waveform showing significant impact
More on Impact:
It is important to understand more about the methods in which signals from vibration sensors are processed to realize why the impact measurement works with a high degree of reliability. An industrial accelerometer is always used for impact measurements because it operates over a wide frequency range and therefore is capable of responding to impact events. The output of an accelerometer is measured in mV/g (g being a gravitational unit).
A conventional vibration monitor would process the signal from the accelerometer in such a way as to have a steady (not jumping around) and accurate measurement of the overall vibration level. Typically the detector circuits for this type of measurement are rms or peak type. The objective of the measurement is to produce a smooth signal to be displayed on a meter or output (4-20 mA) to a PLC, DCS, or such. The smoothing is accomplished with time averaging. Impact signals, such as seen in Figure 2, will not be accurately detected in such a circuit. If filtering or signal integration is added in order to display the measured signal in velocity vibration units (ips), the signal is further processed and altered. The result being even less capable in terms of detecting impact signals. For vibration measurement these circuits produce very accurate results and are recommended for repeatability.
For detecting impact signals from the same type of accelerometer, a special peak responding detector circuit with a fast response time is required. The detector circuit must be able to respond to short duration, high amplitude signals. Signal smoothing can not be used! One might ask, at this point; “Doesn’t this make a nervous responding measurement which can lead to false indications of machine problems?” The answer is, “Yes it can.” So something else must be done to qualify the impact signals. The next part of the impact measurement circuit counts impact events within a specified amount of time referred to as a reset time.
Figure 3, Trend graph of an overall vibration level and impact severity
Additionally, a threshold value, in g’s, is set so that impact events are not counted until the level of the measured signal crosses over that threshold. The impact count is based on the number of impacts that exceed the threshold level within the reset period. This counting method has proven to be very reliable. It separates transient operating characteristics, where impact events appear and go, from developing mechanical looseness, where impact events appear and stay.
The difference between measuring the overall vibration level and measuring impact severity can be seen in Figure 3. This is a trend graph of these two measurements. The impact severity trace appears as a “cityscape” while the overall vibration shows little variation. The graph covers a period of 60 minutes. Some level of mechanical looseness is evident. The impact severity trend shows a worsening condition as time progresses. The short interruption is a period where the compressor was stopped and soon restarted. What is important is to notice that when the impact severity went to the highest values, the overall level changes are hardly noticeable. This is evidence that you can not use overall vibration alone as an indicator for mechanical looseness.
Setting Impact Criteria:
It was stated that impact severity is the number of impact events that exceed the threshold level that occur within the reset period. By counting the impact events this way, it is possible to determine the severity of the mechanical problems. It should be pointed out that one impact signal might contain more than one impact event. That is, there might be more than one time where the signal crosses through zero and rises again as a result of “ringing”. If you look closely at the signal in Figure 2, for example, it may contain 3 or 4 events, depending on where the amplitude threshold level is set and the timing of the counter circuit. The latter is usually in the order of a few milliseconds.
There are three parameters that are set in the field; threshold amplitude, reset time, and how to respond in terms of alarm or trip. The threshold amplitude is set 3 or 4 times the baseline level in g peak assuming there are no impacts at the time. This is because when impacts occur the peaks will rise well above the baseline operating level. Nuisance alarms could result if set too low.
The next setting is the sampling time window or reset time. This is based on predominant compressor operating speed. A simple formula is used to determine the reset time.
Reset Time = 960 RPM
For a compressor that normally runs at 300 RPM, then:
Reset Time = 960 300 RPM = 3.2 seconds
Where to set the alarm (high alarm) or shut down (high high alarm) is based on the impact severity level. On traditional panel mounted monitoring systems, there usually is a separate relay function for the alarm level and shut down level. An example, for a smooth running compressor, would be to set the alarm for 3 or 4 counts and the shut down for 7 or 8 counts. These could be changed after having some running experience.
With the impact transmitter, that is described below, a similar logic applies however the PLC or DCS has to be programmed to perform the alarm and shut down tasks. Refer to the discussion on the impact transmitter output in the next section.
Impact Transmitter:
Recently there has been significant improvement in the reliability of this measurement as well as a lower cost per point. This is due to the development of a new impact transmitter that uses a patented technique to measure impact severity. The impact transmitter has a built-in piezoelectric accelerometer for the sensing element. It uses a special peak detector circuit and the same timing function described above as part of the severity determination. Only data meeting the preset amplitude threshold level is counted by the impact event counter. It is then stored in memory and the 4-20 mA output level is established. A simplified block diagram can be seen in Figure 4. During the start of the next reset period, the counter is cleared and a new count started.
Figure 4, Simplified block diagram of impact transmitter
Under normal operation the 4-20 mA signal is one reset period (time window) behind actual events. There are two cases where the current goes to 20 mA immediately. The first is when the counter reaches 16 events, even if the counting period is not completed. The second is if and when a single impact reaches a level of 50 g. In this case it is reasoned that critical breakage has occurred. Of course, the impact measurement is designed to provide early warning of developing mechanical looseness. Nevertheless, these two special cases were incorporated to provide a full scale output as quickly as possible for those unusual circumstances.
The impact transmitter combines the benefits of this measurement with the state-of-the-art 4-20 mA loop powered sensor technology. All of this is contained in one easy to mount stainless steel housing. The impact transmitter has a hazardous area certification.
Since it is a two-wire, 4-20 mA loop powered end device, it is wired like any other such field transmitter. A simple wiring diagram is shown in Figure 5.
Figure 5, Connection diagram for normal operation, the voltage polarity can be reversed
The impact transmitter has two modes of operation; a normal mode and a setup mode. The modes are set by reversing the polarity of the loop power that is applied to the transmitter. The reversing of the polarity is possible due to the new loop power technology being used in this transmitter.
Normal Mode – With the +24 Vdc applied to Pin A, the transmitter functions normally. The 4-20 mA output from the transmitter is proportional to impact severity. Each impact event that is counted produces 1 mA. Two events would result in a current output level of 6 mA (2 mA plus 4 mA).
Setup Mode – With the –24 Vdc applied to Pin A, the transmitter has a steady output of 12 mA for loop verification. Additionally the reset timing signal and an overall vibration signal is superimposed on the current loop. This signal is not seen by the PLC, or other data logging instrument and is only present in the setup mode. It can be “stripped off” by measuring ac voltage across a load. The overall vibration signal has an output sensitivity of 50 mV/g (based on a 250 ohm load) and can be measured by a portable field scope or other field instruments.
The reset time and amplitude threshold adjustments are made by removing the small cap screws located on the top of the transmitter. When these screws are removed, small potentiometers become visible. These adjustments can be made in the shop prior to installing the transmitter. They can also be made in the field as long as there is access to them.
As a rule-of-thumb the alert and shutdown for the impact transmitter should be set as follows: The early warning or alert (high alarm) should be set to respond to a current value of 8.0 mA (4 impact events). The urgent warning or shutdown alarm (high high alarm) should be set to respond to a current level of 12.0 mA (8 impact events). Operating experience might provide data supporting some variance from these values. Also, the threshold level that is set will affect the number of impact events that get counted. If set low, then set the count criteria for alarms higher.
Sensor Placement:
The impact transmitter is designed to detect mechanical looseness, not vibration. Therefore it is mounted with its center bolt perpendicular to the direction of rod motion, up on top of the crosshead or distance piece, where it will be out of the way of routine inspection or maintenance. A typical location is shown in Figure 6.
Figure 6, Sketch of compressor cylinder showing Impact Transmitter location
It is possible to detect mechanical problems and detonation on integral gas fired engines and diesel engines as well. The impact transmitter can be placed on the side of an engine (each side in the case of a large “V” type engine). At the time of this writing there are a number of engines being monitored using the impact transmitter. The focus of this writing is on measuring impact in the compressor cylinder area on a reciprocating compressor.
Other Important Measurements for Compressors and Other Equipment:
While impact monitoring provides operators with valuable compressor running condition information, it should not replace the monitoring that is in place. Impact should be used in conjunction with that monitoring to provide added protection.
If there is no monitoring on a compressor, deciding what parameters to monitor can be a difficult decision. There are good sound choices today. If your company has experience with monitoring systems, the choice/benefits are a little clearer. If not, you will have to choose which parameters to monitor based on their merit and your budget. Many compressors only have vibration switches installed on them for protection. While they provide a trip in an emergency, they do nothing towards assessing the running condition of the machinery. Some vibration switches have a 4-20 mA signal out that is proportional to velocity vibration. This type of switch does allow for vibration trending.
Type / Location / MeasurementImpact / Compressor cylinders / Mechanical looseness
Rod drop / Compressor rods / Rider band wear
Velocity / Side of engine
Side of crank case
Fin-Fans
Turbocharger
Auxiliary equipment / Running condition
Vibration switch / Side of engine
Side of crank case / Critical condition
Impact or velocity / Side of engine
Engine main bearings / Mechanical looseness
Detonation
Running condition
Temperature / Suction and discharge gas / Valve condition
Temperature / Bearing temperatures
Motor windings / Running condition
Pressure / Compressor cylinders / Operating efficiency
Table 1, Summary of measurement types and their benefit
Table 1 above provides a summary of monitoring choices. It identifies the type of measurement and the part of the compressor or other machinery that is applicable.
Impact monitoring will not replace the need for compressor performance analysis. An increasing trend in measured impact severity could provide the basis for performing an analysis.
Another important compressor monitoring parameter is rod drop monitoring. This measurement is intentionally not reviewed in this writing. If rider band wear is a problem, rod drop monitoring is a measurement that should be investigated.
Of course, pressure and temperature are also important measurements for reciprocating compressors. Pressure and temperature are also not reviewed in this writing.
The purpose of any monitoring is to protect machinery by providing operating condition data that operators can use to make run/don’t run decisions. This data is also used to maximize operating efficiency. The operating condition data can also be used to help mechanical equipment engineers assess plant equipment availability. Trending these measurement parameters adds another dimension to the running condition data.
Case History:
A rebuilt 6 cylinder compressor was being put into service as part of an expansion project in a gas plant. This compressor is driven with a 3000 HP electric motor running at 300 RPM. This plant had a policy of monitoring and trending velocity vibration on most of their plant equipment, including reciprocating compressors. They decided to install impact transmitters on each compressor cylinder on this machine.
At startup they were knocked off line by one of the transmitters. Upon an attempt to restart the machine the impact transmitter knocked them off line again. At this point they investigated and found that the retaining bolts on the high pressure packing case had not been tightened.
As with rotating machinery, many problems are found at startup following an overhaul or other work being done for maintenance. Reciprocating machinery is no different in this regard. Reciprocating machinery can develop mechanical looseness after periods of running. If left alone, looseness only gets worse and can lead to more costly breakage.
Originally prepared for the Vibration Institute 26th Annual Meeting - Page 1