25.10 Fault Detection

Back

In many discussions of machine condition monitoring and fault diagnostics, the distinction between

fault detection and fault diagnosis is not made. Here, they have been divided into separate sections in

order to highlight the differences and clarify why they should be treated as separate tasks. Fault detection

can be defined as the departure of a measurement parameter from a range that is known to represent

normal operation. Such a departure then signals the existence of a faulty condition. Given that

measurement parameters are being recorded, what is needed for fault detection is a definition of an

acceptable range for the measurement parameters to fall within. There are two methods for setting

suitable ranges: (1) comparison of recorded signals to known standards and (2) comparison of the

recorded signals to acceptance limits.

25.10.1 Standards

One of the best known sources of standards is the International Organization for Standardization

(ISO). These standards are technology oriented and are set by teams of international experts. ISO

Technical Committee 108, Sub-Committee 5 is responsible for standards for condition monitoring

and diagnostics of machines. This group is further divided into a number of working groups who

review data and draft preliminary standards. Each working group has a particular focus such as

terminology, data interpretation, performance monitoring, or tribology-based machine condition

monitoring.

While ISO is perhaps the most widely known standardization organization, there are several others

that are focused on specific industries. Examples of these include the International Electrical

Commission, which is primarily product oriented, and the American National Standards Institute

(ANSI), which is a nongovernment agency. There are also different domestic government agencies

that vary from country to country. National defense departments also tend to set their own

standards.

25.10.1.1 Standards Based on Machinery Type

Because different machines that are designed to

perform approximately the same task tend to

behave in a similar manner, it is not surprising

that many standards are set based on machinery

type. Figure 25.11 shows a generic plot separating

vibration amplitude vs. rotating speed into

different zones. For a specific type, size, or class

of machine, a plot like this can be used to

distinguish gross vibration limits relative to the

speed of operation. Machines are usually divided

into four basic categories:

1. Reciprocating machinery: These machines

may contain both rotating and reciprocating

components (e.g., engines, compressors,

pumps).

2. Rotating machinery (rigid rotors): These

machines have rotors that are supported

on rolling element bearings (usually). The

vibration signal can be measured from the

bearing housing because the vibration signal is transmitted well through the bearings to the housing

(e.g., electric motors, single-stage pumps, slow-speed pumps).

Vibration Amplitude

Rotational Speed

ZONE A

ZONE B

ZONE C

ZONE D

FIGURE 25.11 Normalized vibration amplitude vs.

probability density (zone A — new machine; zone

B — acceptable; zone C — monitor closely; zone D —

damage occurring).

Machine Condition Monitoring and Fault Diagnostics 25-21

© 2005 by Taylor & Francis Group, LLC

3. Rotating machinery (flexible rotors): These machines have rotors that are supported on journal (fluid

film) bearings. The movement of the rotor must be measured using proximity probes (e.g., large

steam turbines, multistage pumps, compressors). These machines are subject to critical speeds

(high vibration levels when the speed of rotation excites a natural frequency). Different modes of

vibration may occur at different speeds.

4. Rotating machinery (quasi-rigid rotors): These are usually specialty machines in which some

vibration gets through the bearings, but it is not always trustworthy data (e.g., low-pressure steam

turbines, axial flow compressors, fans).

25.10.1.2 Standards Based on Vibration Severity

It is an oversimplification to say that vibration levels must always be kept low. Standards depend on

many things, including the speed of the machinery, the type and size of the machine, the service

(load) expected, the mounting system, and the effect of machinery vibration on the surrounding

environment. Standards that are based on vibration severity can be divided into two basic

categories:

1. Small-to-medium sized machines: These machines usually operate with shaft speeds of between 600

and 12,000 rpm. The highest broadband RMS value usually occurs in the frequency range of 10 to

1000 Hz.

2. Large machines: These machines usually operate with shaft speeds of 600 to 1200 rpm. If the

machine is rigidly supported, the machine’s fundamental resonant frequency will be above the

main excitation frequency. If the machine is mounted on a flexible support, the machine’s

fundamental resonant frequency will be below the main excitation frequency.

While general standards do exist, there are also a large number of standards that have been developed

for specific machines. Figure 25.12 shows a table with generic acceptance limits based on vibration

severity.

25.10.2 Acceptance Limits

Standards developed by dedicated organizations are a useful starting point for judging machine

condition. They give a good indication of the current condition of a machine and whether or not a fault

exists. However, judging the overall condition of machinery is often more involved. Recognizing the

changing machinery condition requires the trending of condition indicators over time. The development

and use of acceptance limits that are close to the normal operating values for specific machinery will

detect even slight changes in condition. While these acceptance limits must be tight enough to allow even

small changes in condition to be detected, they must also tolerate normal operating variations without

Vibration Amplitude

Increasing

Vibration Severity for Separate Classes of Machines

ClassI Class II Class III Class IV

A

B

C

D

A

B

C

D

A

B

C

D

A

B

C

D

FIGURE 25.12 Acceptance limits based on vibration severity levels (zone A — new machine; zone B — acceptable;

zone C — monitor closely; zone D — damage occurring).

25-22 Vibration and Shock Handbook

© 2005 by Taylor & Francis Group, LLC

generating false alarms. There are two types of limits:

1. Absolute limits represent conditions that could result in catastrophic failure. These limits are

usually physical constraints such as the allowable movement of a rotating part before contact is

made with stationary parts.

2. Change limits are essentially warning levels that provide warning well in advance of the absolute

limit. These vibration limits are set based on standards and experience with a particular class of

machinery or a particular machine. Change limits are usually based on overall vibration levels.

It is important to note that the early discovery of faulty conditions is a key to optimizing

maintenance effort by allowing the longest possible lead-time for decision making. As well as the overall

vibration levels being monitored, the rates of change are also important. The rate of change of a

vibration level will often provide a strong indication of the expected time until absolute limits

are exceeded. In general, relatively high but stable vibration levels are of less concern than relatively low

but rapidly increasing levels.

An example of how acceptance limits may be used to detect faults and trend condition is provided

when the gradual deterioration of rolling-element bearings is considered. Rolling-element bearings

generate distinctive defect characteristic frequencies in the frequency spectrum during a slow,

progressive failure. Vibration levels can be monitored to achieve maximum useful life and failure

avoidance. Typically, the vibration levels increase as a fault is initiated in the early stages of

deterioration, but then decrease in the later stages as the deterioration becomes more advanced.

Appropriately, set acceptance levels will detect the early onset of the fault and allow subsequent

monitoring to take place even after the overall vibration level has dropped. However, rapid bearing

deterioration may still occur due to a sudden loss of lubrication, lubrication contamination, or a

sudden overload. The possibility of these situations emphasizes the need for carefully selected

acceptance limits.

It should also be noted that changes in operating conditions, such as speed or load changes, could

invalidate time trends. Comparisons must take this into consideration.

25.10.2.1 Statistical Limits

Statistical acceptance limits are set using statistical information calculated from the vibration signals

measured from the equipment that the limits will ultimately be used with. As many vibration signals as

possible are recorded, and the average of the overall vibration level is calculated. An alert or warning

level can then be set at 2.5 standard deviations above or below the average reading (Mechefske, 1998).

This level has been found to provide optimum sensitivity to small changes in machine condition and

maximum immunity to false alarms. A distinct advantage to using this method to set alarm levels is the

fact that the settings are based on actual conditions being experienced by the machine that is being

monitored. This process accommodates normal variations that exist between machines and takes into

account the initial condition of the machine.

25.10.3 Frequency-Domain Limits

Judging vibration characteristics within the frequency spectra is sometimes a more accurate method of

detecting and trending fault conditions. It can also provide earlier detection of specific faults because,

as mentioned previously, the frequency domain is generally more sensitive to changes in the vibration

signal that result from changes in machine condition. The different specific methods are listed and

described below.

25.10.3.1 Limited Band Monitoring

In limited band monitoring, the frequency spectrum is divided into frequency bands. The total energy

or highest amplitude frequency is then trended within each band. Each band has its own limits based

on experience. Generally, ten or fewer bands are used. Small changes in component-specific frequency

Machine Condition Monitoring and Fault Diagnostics 25-23

© 2005 by Taylor & Francis Group, LLC

ranges are more clearly shown using this strategy. Bandwidths and limits must be specific to the machine,

sensor type, and location. Narrowband monitoring is the same as limited band monitoring, except it has

finer definition of the bands.

25.10.3.2 Constant Bandwidth Limits

When limited band monitoring is practiced and the bands have same width at high and low frequencies,

the procedure is called constant bandwidth monitoring (see Figure 25.13). This technique is useful for

constant speed machines where the frequency peaks in the spectra remain relatively fixed.

25.10.3.3 Constant Percentage Bandwidth Limits

Constant percentage bandwidth monitoring involves using bandwidths that remain a constant

percentage of the frequency being monitored (see Figure 25.14). This results in the higher frequency

bands being proportionally wider than the lower frequency bands. This allows for small variations

Frequency

Amplitude

FIGURE 25.13 Constant bandwidth acceptance limits.

Frequency

Amplitude

FIGURE 25.14 Constant percentage bandwidth acceptance limits.

25-24 Vibration and Shock Handbook

© 2005 by Taylor & Francis Group, LLC

in speed without the frequency peaks moving between bands, which may have different

acceptance limits.

Навигация