Plant Condition Monitoring Techniques

A review of some Condition Monitoring techniques available for reciprocating machinery.

Reciprocating machinery can be found in diverse industries. These machines take many forms e.g. internal combustion engines, compressors, pumps etc. One thing they have in common is the conversion of linear motion of the piston(s) to rotary motion of the crank or vice versa. This reciprocating motion introduces different forces and hence different vibration signatures to those encountered in rotating machinery so that different detection and analysis is required if condition monitoring is to be valid.

This technique uses an eddy current probe to measure the relative position of the compressor rod. For horizontal rods, the probe is mounted vertically above or below the rod so that a change in vertical position of the rod is measured as a change in probe driver output voltage. Ideally the gap should remain the same over the entire travel of the rod. In practice the rod position will vary due to running clearances and may even bow when subjected to the compression and tension forces. Nevertheless by plotting the probe gap on a continuous basis and looking for changes in value it is possible to identify faults such as rider band wear.

Rod drop monitoring has the advantage that it only detects those faults associated with changes in rod behaviour, enabling the engineer to target a well-defined area of the machine when investigating problems. One disadvantage is that for retrofits the machine must he stripped to install the probes.

Because reciprocating machines generate uneven forces, they often vibrate more than a comparable rotating machine. Measuring vibration velocity can be unreliable as the increase in velocity from incipient failures is usually small and will be buried in the larger signal due to machine movement. By the time the fault has been detected major secondary damage may have already occurred.

Impact monitoring overcomes this problem by taking raw acceleration and counting the number of excursions that exceed an alarm threshold in a set time. If there are less than the counter preset, the count is cleared and begins again. For a faulty machine there will be more excursions counted per cycle than for a healthy machine, so the count is likely to exceed the preset in the specified time and the alarm will annunciate. Judicious setting of the threshold, count and count time will permit reliable monitoring without excessive false alarms.

This method will detect various faults provided that increased excitation of machine resonances is generated by the developing fault. This is normally the case when items crack, become loose or operate with excess clearance. Faults in the crankcase may be too far away to be detected by the crosshead casing accelerometer and should be catered for by other techniques or additional Impact measurements. It has been suggested that this method is not an effective early warning system since it only works when damage has already occurred. This is true for the health of the damaged component, however the ability of this technique to prevent major secondary damage and extensive loss of production considerably outweighs this argument.

Figure 1

Figure 1 is the acceleration trend from a compressor with a cracked crosshead. As the reciprocating forces vary, the crack opens and closes causing additional impacts. These impacts then increase the acceleration trend above normal. Here the level rises from healthy (green) to warning (yellow) and finally to danger (red) over 24 hours. Following a rebuild the trend returned to healthy.

Performance monitoring revolves around the measurement of cylinder pressure(s) and crank angle. These parameters are plotted on a graph with markers indicating fixed events such as valves opening and closing and the location of top dead centre (TDC).

Other useful information includes the rate of change of pressure versus crank angle and the pressure versus volume. Comparison of the latest PV curve against a baseline enables the engineer to identify variations in performance caused by for example, gases passing the piston rings, excessive valve bounce or incorrect valve timing.

Traditionally, performance-monitoring systews for reciprocating machinery have been off-line using temporary sensors and portable instruments. This is mainly due to the philosophy that performance degrades slowly over a long period and therefore intermittent monitoring is adequate for this situation.

The frequency spectrum for a reciprocating machine can be complex. Each cylinder's vibration contains multiple frequency components ranging from a few Hertz to several kilohertz. Some of these will be harmonics of running speed whilst others will be from apparently unrelated sources such as lube oil pumps and structural resonances.

A typical reciprocating compressor vibration time signature

One useful technique is the application of alarm bands, however unlike rotating machinery a more complex spectrum precludes the use of a separate band for each spectral component. The alarm level is based upon energy within the band rather than the value of any single harmonic.

Resonances excited by faults appear as increases in amplitude of a group of spectral components and consequently cause an increase in energy For one or more bands. The difficulty with this technique is deciding the bandwidth, how many bands to apply and what threshold levels to set.

The disadvantages of using RMS velocity have been touched on under impact monitoring, however the raw vibration time signature does contain data which can yield valuable information about the health of the machine. As with performance and rod drop, reliable analysis requires the data for each cycle to be synchronised using a reference pulse from the crank. This allows repetitive events such as rod reversal, valve operation and TDC to be recognised. Health assessment is accomplished by comparing one cycle with another, and treating major differences with silicion. Faulty valves will produce vibration that may be higher in amplitude, take longer to decay and occur early or late, whilst a worn wrist pin may generate excessive vibration at TDC and BDC.

Measuring increases in temperature is a simple method for monitoring the condition of valves. Measurements are obtained from platinum resistance thermometers (PRT) or thermocouples mounted in the valve covers, or from infrared non-contact measurements taken on an intermittent basis.

Which technique(s) to use will depend, to some extent on priority, reliability or efficiency. If reliability is paramount then techniques capable of detecting incipient mechanical faults such as impact and rod drop should be chosen. Of course detection of malfunctions affecting reliability is not exclusive to these techniques, indeed if a valve is damaged performance monitoring will highlight this by identifying the drop in efficiency. In an off-line system, however, this problem may not become apparent for some time. All of these techniques have the potential to reduce costs whether through increased reliability or energy saving resulting from optimised performance.

From an article by Ray Hopcraft, InterCorr International Ltd.
Tel: +44(0) 1349 861048 and +44(0) 121 764 5444


For details of Condition Monitoring Suppliers, click here...
For details of Vibration Sensor Suppliers, click
For details of Temperature Sensors Suppliers, click
Home - Website - Search - Suppliers - Links - New Products - Catalogues - Magazines Problem Page - Applications - How they work - Tech Tips - Training - Events