http://www.BetaMachinery.

com
E:\Marketing Communications\Papers and Presentations\Technical Papers\New Technical Papers\Fault Detection and
Diagnosis#1.doc

FAULT DETECTION AND DIAGNOSIS IN

RECIPROCATING MACHINERY

Bryan R. Long, Ph.D., P. Eng. And David N. Schuh, P. Eng.

1. The Need for Analysis

Reciprocating machinery in industry is found in the following forms:
• Diesel engines
• Spark ignited engines
• Reciprocating compressors
• Plunger pumps

Machines of these types perform important roles in a number of industries, including

transportation, pipelines, petrochemical and gas gathering.

Faults in reciprocating equipment fall into two general areas: Deterioration in performance and
deterioration in mechanical condition. Improvements in either area generate significant benefits,
including increased machine availability, reduced energy costs and optimum utilization of
maintenance resources.

This paper will explore current developments which help achieve these benefits. We will
concentrate on spark ignited engines, but the principles are applicable to other reciprocating
machines.

2. Analysis Techniques
Analysis of reciprocating equipment is more complex than analysis of rotating equipment. Useful
signals are:
• conditioned vibration versus crankshaft position
• cylinder pressures as a function of crankshaft position
• ultrasonic energy versus crankshaft position
• ignition voltages versus crankshaft position
• process type data: pressures, temperatures, flow rates

2.1 Vibration Patterns

Conventional broad-band vibration waveforms and spectra are of little value. Normal vibration in
a reciprocating machine tends to obscure information associated with faults. Instead,
appropriately conditioned vibration signals, presented against crankshaft position, can be used to
detect and diagnose a variety of problems.

The vibrational energy associated with faults in cylinders arises from:

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
2

• mechanical impacts such as valve closure or chattering rings

• gas leakage such as cylinder blowby and leaking valves

These energies tend to be relatively broad band in frequency content. By filtering out low
frequencies, the normal vibration will be excluded, leaving only the energy associated with the
events in which we are interested. We then have a signal like that illustrated in Figure 1.

Figure 1 – This filtered vibration signal is a normal patterns representing the events happening in
the cylinder of a two cycle engine

We are dealing with patterns in which some energy content is normal; the challenge is to be able
to determine what is normal, what is abnormal, and the nature of any abnormality. Figure 2
presents some abnormal signals.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
3

Figure 2 – The analyst must evaluate vibration patters to determine which are normal, and which
are abnormal.

The patterns are complex. Even more problematic, they vary with operating conditions.
Considerable art is involved, and experience is critical, but an expert can get a reasonable
indication of the condition of the cylinder and valves.

2.2 Ultrasonic Patterns

Ultrasonic energy patterns can be used to enhance the information in the vibration patterns.
Ultrasonic energy is collected from higher frequencies than vibration; usually in the range of 20
kHz to 100 kHz.

These patterns will be more sensitive to gas leakage and less sensitive to mechanical impact
events. If we take both ultrasonic and vibration signals on a cylinder, problem detection becomes
more sensitive and reliable, and diagnosis is also more reliable.

Figure 3 shows an ultrasonic pattern along with the corresponding vibration signal.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
4

Figure 3 – Mechanical impacts tend to dominate vibration traces, while gas leakage shows up
more clearly in ultrasonic traces

2.3 Cylinder Pressure

Knowledge of the pressure in the cylinder as a function of crank angle allows:

• detection of performance problems, such as poor air/fuel ratio and excessive firing
pressures
• detection of mechanical problems such as low compression pressure and carbon
build-up in ports
• interpretation of anomalies through correlation with vibration and ultrasonic signals

Examples of power cylinder pressure-time curves are shown in Figure 4, where traces for several
cylinders are superimposed on the same picture. Several cycles were recorded for each cylinder,
to illustrate the variation in peak firing pressures.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
5

Figure 4 – Power cylinder pressure-time curves photographed from an oscilloscope display

2.4 Ignition Voltages

In spark ignited engines, the ignition system is a source of many of the problems which arise. It is
common practice, therefore, to measure the ignition primary waveform and/or the ignition
secondary waveform. An example is shown in Figure 5.

Figure 5 – Ignition primary voltage patters

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
6

3. Historical Perspective
About twenty years ago, special purpose electronic instruments called “engine analyzers” were
developed. These devices provided the sensors and signal conditioning to produce the “displays
present in Section 2.

The original versions were simply data acquisition and display devices, with no analysis function.
More recent versions have some analytical capabilities, such as the ability to calculate and
display power.

Current engine analyzers have inherent limitations:

• they are complex instruments, requiring considerable expertise to collect data

• data collection is labour intensive
• there is no systematic methodology for dealing with signals which are inherently
variable
• anomaly detection is a purely manual process
• problem diagnosis is more art than science, based on the training and experience of
the analyst
• there is no mechanism for retaining the expertise which an analyst develops; you
have to start from zero with every new analyst and, by consensus, allow two years for
his development

These problems have made true engine and compressor analysis very difficult. In practice, the
analysis effort is often concentrated on machines which have been identified as suspect, because
it is too time-consuming to analyze all machines regularly.

4. Meeting the Need

A new analyzer system called RECIP-TRAP has been developed. This system uses a micro-
processor based portable data collector (Figure 6) and a permanent data-base on a personal
computer. The major advances are:
• Data collection has been de-skilled
• Data collection is much faster
• Anomaly detection has been automated
• Decision-making has been formalized
• Expertise is built into the system

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
7

Figure 6 – The RECIP-TRAP Instrument is small and simple to use

About twenty RECIP-TRAP systems are now in use, with an average age of about one year.
Experience to date indicates the following:
• tradesmen can be trained to collect data in hours
• data collection requires approximately ¼ the time
• computerized detection of anomalies is reliable when there is enough data on file;
analyst time for anomaly detection is reduced to zero
• the formal decision-making process and convenient information access enhance the
likelihood that good decisions will be made and some appropriate action taken
• personnel can become effective with the system in about six months, starting with no
analysis expertise and no computer literacy
• the time saved during data collection and anomaly detection is used in the critical
areas of decision-making and feedback, where real benefits can be achieved
• it is cost-effective to monitor machines not previously analyzed because of reduced
time requirement
• it is feasible to monitor machines more often, and therefore failures between visits
are much less likely

5. Improved Fault Detection and Diagnosis

This section shows you how the difficulties in reciprocating equipment analysis described above
can be reduced with modern technology. The principle is to depend on the computer for data
organization, recall, and anomaly detection, and to use the person for detailed analysis and
decision-making.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
8

5.1 Detailed Illustration

A sample study is presented here for an Ingersoll Rand KVSR-12.

5.1.1 Engine Combustion Statistics

Figure 7 shows the peak firing pressure statistics, which give the analyst an overall picture of the
performance of the engine. Combustion statistics indicate power cylinder balance, how evenly
each cylinder fires, and the presence of misfires or detonation.

Figure 7 – Part 1 of the POWERHEALTH Report presents the results of a statistical analysis of
64 cycles in the cylinder

Engine balance is indicated by the cylinder DELTA readings. A cylinder which has a large
positive or negative DELTA value is considered to be in poor balance.

ANGLE is the angle after top dead centre at which the peak firing pressure occurs. ANGLE is
affected by air/fuel ratio, mechanical condition, and ignition timing.

Compression pressure is used to provide an indication of the mechanical health of the cylinder
and IHP (indicated horsepower) tells how the load is distributed to each cylinder.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
9

5.1.2 Mechanical Condition Indicators

One of the techniques that is used for automatic anomaly detection in the RECIP-TRAP System
is the comparison of the most recent data to the baseline data. The baseline data is assumed to
represent the unit in a healthy condition. Any deviation from the baseline is considered to be a
result of deterioration in health. Figure 8 gives a summary of this information for the KVSR-12.

Figure 8 – The POWERHEALTH Report, Part 2 picks out cylinders in which significant changes in
mechanical condition have occurred

A severity index, SEV, is used to indicate the magnitude of the deviation between new and
baseline vibration data. Any SEV that is between – 1.0 and 1.0 is not considered to be
anomalous and is not presented in the POWERHEALTH Report.

The POWERHEALTH Report gives you:

• the names of the cylinders for which the newest data is significantly different from the
baseline data
• the angular window in which the difference is found
• the Severity Index of the anomaly
• an indication that the anomaly is based on a comparison of peaks (PEAK) or area
under the curves (SUM)
• a list of possible faults associated with each anomaly

A review of this report shows that power cylinder P8 has a severity index of 2.35 in the window
360 – 450 degrees based on a comparison of the new and baseline peaks. Possible faults in this
region are bad exhaust valve lifters and excessive valve lash. (P7 and P10 are also significantly
different than on previous occasions and require further analysis.)

The analyst uses the CODA Decision-Making Model, which follows, to evaluate each of the
cylinders identified.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
10

5.1.3 The CODA Decision-Making Model

Experience has shown that good decision-making is the most difficult part of an analysis program.
Analysts may become overwhelmed with the amount of data available to review, and find it
difficult to actually decide what should be done.

The RECIP-TRAP system overcomes this difficulty, first by having the computer detect which
signals are anomalous, and second, by providing CODA, for the analysis of those signals.

CODA begins with the CODA form, as shown in Figure 9. This is printed automatically for all the
anomalies so that they are not overlooked. Thereafter, the system provides several analytical
displays, described on the following pages, to aid the analyst in making his decision.

Figure 9 – The CODA Form is produced automatically for each anomaly

CODA stands for Check, Observe, Decide, Assess. It is a step by step process developed to
ensure that analysts consider all the relevant information and commit themselves to a decision.

Everyone who is involved can be trained to understand CODA, so that when a recommendation
is made, they understand it significance and urgency. CODA becomes a shortcut for
communication, because the CODA Report reminds everyone of the detailed analysis that has
gone into this recommendation.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
11

5.1.4 The Engine P-T Parade

The first analytical display used is often the pressure-time Parade, shown in Figure 10.

Figure 10 – The Parade presents one representative pressure-time curve, plus combustion
statistics, for each cylinder. Compare this with the conventional display (Figure 4)

The pressure-time data is displayed along with the combustion statistics. Note that the
combustion statistics icon shows the average (long horizontal bar), average deviation (small
horizontal bars) and the maximum and minimum (extreme points of the vertical bar) peak firing
pressures. At a glance we can determine whether the cylinders are out of balance, detonating,
misfiring or having late combustion.

5.1.5 Power Cylinder Vibration Patterns

The power cylinder vibration display (Figure 11) shows up to eight vibration-time patterns. The
patterns should be generally the same, this display makes it easy to identify differences.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
12

Figure 11 – Power Cylinder Vibration Patterns are displayed in groups of 8 for easy comparison

The vertical dotted lines mark the bounds of the exhaust, intake and fuel injection events, as
detailed in Figure 13.

Recalling that the POWERHEALTH Report determined that P8 was anomalous in the window
360 – 450 degrees, we note that the pattern for P8 has an anomalous event just after exhaust
valve closure.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
13

Figure 12 – Cylinder vibration data can be plotted with respect a common reference rather than to
TDC of each cylinder

Figure 12 presents the same data as shown in Figure 11, illustrating the power and flexibility of
digital data processing. This display is used to show crosstalk to adjacent cylinders. It is often
used to shed light on the origin of smaller, otherwise unidentifiable signals. Crosstalk may be
seen in the traces for P10 and P12 from the closure of the fuel valve in P11.

5.1.6 The Power Cylinder Signature

Another analytical display, the power cylinder signature, combines the timing diagram with the
vibration-time and pressure-time curves. The analyst typically turns to this plot to determine the
relationship between the major events in the cylinder and the vibration and pressure data.

Recall that P8 was flagged as anomalous in the POWERHEALTH Report; Figure 13 confirms that
the exhaust valve is closing hard.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
14

Figure 13 – The power cylinder signature relates the timing of exhaust, inlet, and fuel injection
events to the vibration-time and pressure-time traces

5.1.7 Data Archives

After the data is processed, it is placed into the data archives where comparisons may be done.

Figure 14 contains the CODA display, where new, previous and baseline data are plotted on the
same graph. The analyst can place a cursor anywhere on the display, and can then see actual
levels at each degree along the plot.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
15

Figure 14 – The CODA Display lets you compare today’s vibration data with historical data

Two sets of readings were collected on 19 May, 1988 and both of them show that the exhaust
valve is behaving differently from 15 January, 1988. If we assume that the baseline data
represents the healthy state of the unit, we must conclude that the health of P8 has deteriorated.

The new and baseline data can be overlaid to detect more subtle changes, as shown in Figure
15. In this display, the regions where the new data exceeds the baseline data are emphasized.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
16

Figure 15 – The new over baseline display emphasizes differences

5.2 Liner and Ring Anomalies

This section presents selected results from the analysis of a large two-cycle engine. Part 2 of the
relevant POWERHEALTH Report (Figure 16) indicates several problems.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
17

Figure 16 – The POWERHEALTH Report, Part 2, identifies cylinders whose mechanical condition
has changed significantly since the baseline was recorded

The analyst then review the vibration envelope display in Figure 17, to confirm the problems.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
18

Figure 17 – Power cylinder vibration patterns indicating problems in liners and rings

The pattern for Cylinder 1L illustrates a ring blowby case. We can be confident that ring blowby
is the problem in the region of 0 – 45 degrees because the pressure in the cylinder rises to a
maximum and there are no major mechanical events in this region.

Cylinder 3R exhibits both ring blowby and cylinder liner scuffing during the power stroke. The
rounded pattern is indicative of the passage of gas and the sharp, narrow patterns tend to
indicate mechanical impacts.

Cylinder 4R exhibits ring clip as two rings pass the exhaust port at approximately 85 – 90
degrees.

5.3 Ignition Analysis

The IGNITION HEALTH Report in Figure 18 presents a statistical analysis of an ignition

secondary. The RECIP-TRAP records 64 timing angles, and determines the average, average
deviation, maximum and minimum of these. Statistics for the signal level and percentage of
misfires are also calculated.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
19

Figure 18 – The IGNITION HEALTH Report summarizes the condition of the ignition system

For this unit, note that the timing for P6 is earlier than for the other five cylinders, indicating that
the distributor board assembly switch for P6 is probably out of place.

Figure 19 shows ignition primary traces generated by the RECIP-TRAP system.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)

 http://www.BetaMachinery.com
20

Figure 19 – Ignition primary patterns can be used to evaluate the ignition system

6. Conclusion
Detection and diagnosis of faults in reciprocating machinery is viable. The benefits that can be
obtained justify the effort that is required.

Current methods require significant expertise, which is difficult to obtain and maintain. They also
require a significant amount of time. The RECIP-TRAP overcomes these limitations, primarily by
using computer based analysis.

Fault Detection and Diagnosis in Reciprocating Equipment (pp. 303 – 322)