Periodic Preventive Maintenance

From run-to-failure, we progress to periodic preventive maintenance, which is sometimes called "historical" maintenance. This is where the histories of each machine type are analyzed and periodic overhauls are scheduled to occur before the statistically expected problems occur. It has long been known that most groups of similar machines will exhibit failure rates that are somewhat predictable if averaged over a long time. This gives rise to the so-called "Bathtub Curve" which relates failure rate to operating time, as follows:

If this curve applied to all machines of the group, and if the shape of the curve is known, preventive maintenance could be used advantageously, but unfortunately, this is not the case in practice. Preventive maintenance also includes such activities as changing lube oil and filters, periodic cleaning and inspecting, etc. Maintenance activity may be scheduled on the basis of calendar time, machine operating hours, number of parts produced, and so on. Preventive maintenance became very popular in the early 1980s when small computers began to be used for planning and tracking maintenance work. In a famous study of preventive maintenance by United and American Airlines, it was found that for a large class of rotating machines, the failure rate greatly increased just after theperiodic overhauls -- in other words, the overhaul reduced the reliability of the machines. It is as if the machine reverts to the beginning of the bathtub curve after each overhaul. From this study and subsequent observations, it was found that periodic overhauls result in 20 % to 25 % of startup failures. About ten percent of these can be attributed to defective new bearings. It is obvious that preventive maintenance is an inefficient use of resources for most machines; however, there are cases where it can be used to good effect. Examples are machines which exhibit wear related to use such as rock and ore crushers, and machines that are subject to corrosion such as equipment handling caustic substances.

Predictive Maintenance
The next improvement in maintenance technology was the advent of predictive maintenance, which is based on the determination of a machine's condition while in operation. The technique is dependent on the fact that most machine components will give some type of warning before they fail. To sense the symptoms by which the machine is warning us requires several types of non-destructive testing, such as oil analysis, wear particle analysis, vibration analysis, and temperature measurements. Use of these techniques to determine the machine condition results in a much more efficient use of maintenance effort compared to any earlier types of maintenance. Predictive maintenance allows plant management to control the machinery and maintenance programs rather than vice versa. In a plant using predictive maintenance, the overall machinery condition at any time is known, and much more accurate planning is possible. Predictive maintenance utilizes many different disciplines, by far the most important of which is periodic vibration analysis. It has been shown many times over that of all the nondestructive testing that can be done on a machine, the vibration signature provides the most information about its inner workings. Certain machines, which would affect plant operations adversely if they were to fail, can be subjected to continuous vibration monitoring, in which an alarm is sounded if the vibration levelexceeds a predetermined value. In this way, rapidly progressing faults are prevented from causing catastrophic failures. Most modern turbine-driven equipment is monitored in this way. Oil analysis and wear particle analysis are important parts of modern predictive programs, especially in critical or very expensive equipment. Thermography is the measurement of surface temperature by infrared detection, and is very useful in detecting problems in electrical switchgear and other areas where access is difficult. Motor current signature analysis is another technique that is very useful in detecting cracked or broken rotor bars while the motor is in operation, and electrical surge testing of motor stators is used for detecting incipient electrical insulation failure.

Benefits of Predictive Maintenance

The major benefit of predictive maintenance of industrial mechanical equipment is increased plant readiness due to greater reliability of the equipment. The trending over time of developing faults in machines can be carefully done so as to plan maintenance operations to coincide with scheduled shutdowns. Many industries report from two to ten percent productivity increases due to predictive maintenance practices. Similar percentages of increased mission readiness are expected in shipboard systems. Another benefit of predictive maintenance is reduced expenditures for spare parts and labor. Machines that fail while in service often cost ten times as much to repair than if the repair were anticipated and scheduled. A great many new machines fail soon after startup due to built-in defects or improper installation. Predictive techniques can be used to assure proper alignment and overall integrity of the installed machine when first brought into service. Many plants base the acceptance of new machine installations on a clean bill of health as determined by vibration measurements.

Predictive maintenance reduces the likelihood of a machine experiencing a catastrophic failure, and this results in an improvement in worker safety. There have been many cases of bodily injury and even death due to sudden machine failures.

Pro-active Maintenance
The latest innovation in the field of predictive maintenance is so-called pro-active maintenance, which uses a variety of technologies to extend the operating lives of machines and to virtually eliminate reactive maintenance. The major part of a pro-active program is root cause failure analysis, which is the determination of the mechanisms and causes of machine faults. The fundamental causes of machine failures can thus be corrected, and the failure mechanisms can be gradually engineered out of each machinery installation. It has been known for a long time that imbalance and misalignment are the root causes of the majority of machine faults. Both of these conditions place undue forces on bearings, shortening their service life. Rather than continually replacing worn bearings in an offending machine, a far better policy is to perform precision balance and alignment on the machine, and then to verify the results by careful vibration signature analysis.

Precision Alignment
It has been reported in the TAPPI journal that precision alignment resulted in extending bearing life by a factor of eight in a large class of rotating machines. Other reported benefits were a seven percent savings in overall maintenance costs and a twelve percent increase in machine availability. Machine breakdowns attributed to misalignment were cut in half. Another benefit of precision alignment is a power saving. A recent study documented an average of eleven percent power saving by precision alignment in a group of simple pumpmotor assemblies. This is because less power is expended in flexing the coupling, vibrating the machine, and heating the bearings. The dollar saving in this case due to reduced power consumption was more than twice the maintenance costs on these machines!

New Installations
It is also well known that many machines contain defects when newly installed. These defects can range from improper installation, caused by poor footings and poor alignment, to defective parts within the machine, such as bad bearings, bent shaft, and so forth. A proactive maintenance program will include testing on new installations for the purpose of certification and verification that the performance is held to a rigid standard. The same standards are applied to rebuilt and overhauled equipment. This type of testing can also lead to the establishment of specific performance specifications that in many cases are more stringent than the equipment manufacturer's specifications and tolerances. An essential part of a pro-active approach is the training of maintenance personnel in the application of the underlying principles.

Benefits of Pro-active Maintenance

A successful pro-active maintenance program will gradually design the problems out of the machines over a period of time, resulting in greatly extended machine life, reduced down time, and expanded production capacity. One of the best features of a pro-active approach is that the techniques are natural extensions of those used in a predictive program, and they are easily added to existing programs. It is apparent today that we need a balanced approach to maintenance, including the appropriate use of preventive, predictive, and pro-active methods, and these elements are not independent, but should be integral parts of a unified maintenance program.

Case Studies
The USN aircraft carriers in the PACFLT have been implementing and using a vibration based predictive maintenance program since 1975, and since 1986, the ships' force have been collecting the vibration data. DLI has a very large database containing the entire history of the program. It is instructive to consult this database and to look at some of the details of its history.

What is Vibration?
In its simplest form, vibration can be considered to be the oscillation or repetitive motion of an object around an equilibrium position. The equilibrium position is the position the object will attain when the force acting on it is zero. This type of vibration is called "whole body motion", meaning that all parts of the body are moving together in the same direction at any point in time. The vibratory motion of a whole body can be completely described as a combination of individual motions of six different types. These are translation in the three orthogonal directions x, y, and z, and rotation around the x, y, and z-axes. Any complex motion the body may have can be broken down into a combination of these six motions. Such a body is therefore said to possess six degrees of freedom. For instance, a ship can move in the fore and aft direction (surge), up and down direction (heave), and port and starboard direction (sway), and it can rotate lengthwise (roll), rotate around the vertical axis (yaw), and rotate about the port-starboard axis (pitch). Suppose an object were restrained from motion in any direction except one. For instance, a clock pendulum is restricted from motion except in one plane. It is therefore called a single degree of freedom system. Another example of a single degree of freedom system is an elevator moving up and down in an elevator shaft. The vibration of an object is always caused by an excitation force. This force may be externally applied to the object, or it may originate inside the object. It will be seen later that the rate (frequency) and magnitude of the vibration of a given object is completely determined by the excitation force, direction, and frequency. This is the reason that vibration analysis can determine the excitation forces at work in a machine. These forces are dependent upon the machine condition, and knowledge of their characteristics and interactions allows one to diagnose a machine problem.

Equations of Motion

If the position, or displacement, of an object undergoing simple harmonic motion is plotted versus time on a graph as shown above, the resulting curve is a sine wave,or sinusoid, and is described by the following equation:

where d = instantaneous displacement, D = maximum, or peak, displacement = angular frequency, = 2f t = time This is the same curve that the sine function from trigonometry generates, and it can be considered the simplest and most basic of all possible repetitive wave forms. The mathematical sine function is derived from the relative lengths of the sides of a right triangle, and the sine wave is actually a plot of the value of the sine function versus angle. In the case ofvibration, the sine wave is plotted as a function of time, but one cycle of the waveform is sometimes considered to equal 360 degrees of angle. More will be said about this when we consider the subject of phase. The velocity of the motion described above is equal to the rate of change of the displacement, or in other words how fast its position is changing. The rate of change of one quantity with respect to another can be described by the mathematical derivative, as follows:

where v = instantaneous velocity. Here we see that the form of the velocity function is also sinusoidal, but because it is described by the cosine, it is displaced by 90 degrees. We will see the significance of this in a moment. The acceleration of the motion described here is defined as the rate of change of the velocity, or how fast the velocity is changing at any instant:

where a = instantaneous acceleration. Note here also that the acceleration function is displaced by an additional 90 degrees, as indicated by the negative sign. If we examine these equations, it is seen that the velocity is proportional to the displacement times the frequency, and that the acceleration is proportional to the frequency squared times the displacement. This means that at a large displacement and a high frequency, very high velocities result, and extremely high levels of acceleration would be required. For instance, suppose that a vibrating object is undergoing 0.1 inch of displacement at 100 Hz. The velocity equals displacement times frequency, or , Acceleration equals displacement times frequency squared, or a = 0.1 x 10000 = 1000 inches per second per second. One G of acceleration equals 386 inches per second per second, so this acceleration is

Now, see what happens if we raise the frequency to 1000 Hz: , and

Thus, we see that in practice, high frequencies can not be associated with high displacement levels.

Dynamics of Mechanical Systems

A small compact physical structure, such as a marble, can be thought of as simply a mass. It will move in response to an external force applied to it, and Newtons laws of motion will govern its movement. Simply put, Newton's laws dictate that if the marble is at rest, it will remain at rest unless acted on by an external force, and if in motion it will continue in motion unless acted on by an external force. If it is subjected to an external force, its acceleration will be proportional to that force. Most mechanical systems are more complex than a simple mass, and they do not necessarily move as a whole when subjected to a force. Mechanical systems, such as rotating machines, are not infinitely rigid, and have varying degrees of flexibility at different frequencies. As we will see, their motion in response to an external force is dependent on the nature of that force and the dynamic characteristics of their mechanical structure, and is often difficult to predict. The disciplines of Finite Element Modeling (FEM) and Modal Analysis are dedicated to predicting how a structure will respond to a known force. We will not discuss these fields further, for they are very complex, but it is instructive to look into how forces and structures interact if we are to understand the usefulness of vibration analysis of machines.

Vibration Amplitude Measurement

The following definitions apply to the measurement of mechanical vibration amplitude.

Peak Amplitude (Pk) is the maximum excursion of the wave from the zero or equilibrium point. Peak-to-Peak Amplitude (Pk-Pk) is the distance from a negative peak to a positive peak. In the case of the sine wave, the peak-to-peak value is exactly twice the peak value because thewaveform is symmetrical, but this is not necessarily the case with all vibration waveforms, as we will see shortly. Root Mean Square Amplitude (RMS) is the square root of the averageof the squared values of the waveform. In the case of the sine wave, the RMS value is 0.707 times the peak value, but this is only true in the case of the sine wave. The RMS value is proportional to the area under the curve -- if the negative peaks are rectified, i.e., made positive, and the area under the resulting curve averaged to a constant level, that level would be proportional to the RMS value.

The RMS value of a vibration signal is an important measure of its amplitude. As mentioned before, it is numerically equal to the square root of the average of the squared value ofamplitude. To calculate this value, the instantaneous amplitude values of the waveform must be squared and these squared values averaged over a certain length of time. This time interval must be at least one period of the wave in order to arrive at the correct value. The squared values are all positive, and thus so is their average. Then the square root of this average value is extracted to get the RMS value. Average Amplitude, which is simply the arithmetic average of the signal level over time, is not used invibration measurements, and we will not consider it further.

The RMS value must be used in all calculations regarding power or energy in a waveform. An example of this is the 117 volt AC line. The 117 volts is the RMS value of the voltage, and it is used in calculations of the wattage (power) drawn by devices connected to it. Remember that the RMS value of a sine wave is 0.707 times the peak value, and this is the only wave form where this is true. We will see shortly that this is important.

The Concept of Phase

Phase is a measure of relative time difference between two sine waves. Even though phase is truly a time difference, it is almost always measured in terms of angle,

either degrees or radians. This represents normalization to the time taken by one cycle of the wave in question, without regard to its true time period. The phase difference between two waveforms is often called a phase shift. A phase shift of 360 degrees is a time delay of one cycle, or one period of the wave, which actually amounts to no phase shift at all. A phase shift of 90 degrees is a shift of 1/4 of the period of the wave, etc. Phase shift may be considered positive or negative, i.e., one waveform may be delayed relative to another one, or one waveform may be advanced relative to another one. These conditions are called phase lag and phase lead respectively.

In this example, the lower curve is shifted 90 degrees with respect to the upper curve. This is a time lag of one-fourth of the period of the wave. You could also say the upper waveformhas a 90 degree phase lead. Phase can also be measured with reference to a particular time. An example of this is the phase of an imbalance component in a rotor with reference to a fixed point on the rotor, such as a key way. To measure this phase, a triggerpulse must be generated from a certain reference point on the shaft. This trigger can be generated by a tachometer or some type of optical or magnetic probe that senses a discontinuity on the rotor, and is sometimes called a "tach" pulse.

Phase of a Rotor The phase angle can be measured from the reference position either in the direction of rotation or opposite to the direction of rotation, i.e., phase lag or lead, and different equipment manufacturers use different conventions. In the DLI Balance program software for the DC-7, either direction may be selected at the operator's preference.

Vibration Units
So far, we have been looking at the displacement of a vibrating object as a measure of its vibration amplitude. The displacement is simply the distance from a reference position,

or equilibrium point. In addition to varying displacement, a vibrating object will experience a varying velocity and a varying acceleration. Velocity is defined as the rate of change ofdisplacement, and in the English system is usually measured in units of inches per second. Acceleration is defined as the rate of change of velocity, and in the English system, is usually measured in units of G, or the average acceleration due to gravity at the earth's surface. The displacement of a body undergoing simple harmonic motion is a sine wave as we have seen. It also turns out (and is easily proved mathematically), that the velocity of the motion is sinusoidal. When the displacement is at a maximum, the velocity will be zero because that is the position at which its direction of motion reverses. When the displacement is zero (the equilibrium point), the velocity will be at a maximum. This means that the phase of velocity waveform will be displaced to the left by 90 degrees compared to the displacement waveform. In other words, the velocity is said to lead the displacement by a 90-degree phase angle. Remembering that acceleration is the rate of change of velocity, it can be shown that the acceleration waveform of an object undergoing simple harmonic motion is also sinusoidal, and also that when the velocity is at a maximum, the acceleration is zero. In other words, the velocity is not changing at this instant. Then, when the velocity is zero, the acceleration is at a maximum -- the velocity is changing the fastest at this instant. The sine curve of acceleration versus time is thus seen to be 90 degrees phase shifted to the left of the velocity curve, and therefore acceleration leads velocity by 90 degrees. These relationships are shown here:

' Note here that the acceleration is 180 degrees out of phase with the displacement. This means the acceleration of a vibrating object is always in the opposite direction to thedisplacement! It is possible to define another parameter that is the rate of change of acceleration, and it is called "jerk". Jerk is what you feel when your car comes to a stop if you maintain a constant brake pedal pressure. It is really the sudden cessation of the deceleration. Elevator manufacturers are interested in measuring jerk, for it is the variation in acceleration that elevator passengers are especially sensitive to.

Summary of Amplitude Units:

In the English system of measurements, displacement is usually measured in mils (thousandths of an inch), and the peak-to-peak value is used by convention.

Velocity is usually measured in inches per second, and the convention is to use the peak value or the RMS value. The peak value is the most commonly used, not because it is better, but because of long tradition. Acceleration is usually measured in Gs, where 1 G is the acceleration due to gravity at the earth's surface. The G is not actually an acceleration unit -- it is simply an amount of accelerationwe experience as inhabitants of the earth. Acceleration is sometimes measured in inches per second per second (in/sec2), or m/sec2, that are true units. One G is equal to 386 inches/sec2 or 9.81 meters/sec2. The process of converting a signal from displacement to velocity or velocity to acceleration is equivalent to the mathematical operation of differentiation. Conversely, the conversion fromacceleration to velocity or velocity to displacement is mathematical integration. It is possible to perform these operations in vibration measuring instruments and thus to convert from any system of units to any other one. From a practical standpoint however, differentiation is an inherently noisy process, and is seldom done. Integration, on the other hand, can be done very accurately with inexpensive electrical circuitry. This is one reason that the accelerometer is the de facto standard transducer for vibration measurement, for its output is easily integrated once or twice in order to display velocity or displacement. Integration is not suitable, however, for signals of very low frequencies (Below 1 Hz), for in this region the noise level increases and the accuracy of the integration process itself suffers. Most commercially available integrators operate correctly above one Hz, which is sufficiently low for almost all vibrationapplications.

Complex Vibration
In a linear mechanical system, all the vibration components will exist together, and none will interfere with any other. In the case of a non-linear system, thevibration components will interact and generate new components which are not in the forcing function. See also the section on linear systems in the Machine Monitoring chapter. Vibration is the motion resulting from an oscillating force, and for a linear mechanical system, the vibration frequency will be the same as the forcing frequency. If there are several forcing frequencies occurring at the same time, then the resulting vibration will be a summation of the vibration at each frequency. Under these conditions the resulting waveform of the vibration will not be a sinusoid, and may be very complex.

Certain machines, especially very slow speed In the diagram, the high frequency and the ones, produce vibration wave forms that are low frequency vibration add together to relatively easy to interpret directly. See also the make the complex waveform. In simple section on Time Domain Analysis in the cases like this, it is relatively easy to find Machinery Monitoring chapter. the frequencies and amplitudes of the two components by examination of the wave form, but most vibration signals are far more complex than this, and can be extremely difficult to interpret. In a typical rotating machine, it is often hard to get very much information about the inner workings of the machine by looking at the vibration wave form, although in certain cases wave form analysis is a powerful tool, as will be discussed in the chapter on machine vibration monitoring.

Examples of some wave forms and their spectra

Following are some waveforms and spectra that illustrate some important characteristics of frequency analysis. While these are idealized in the sense that they were made from an electronic function generator and analyzed with an FFT analyzer, they do show certain attributes that are commonly seen in machine vibration spectra.

A sine wave consists of a single frequency only, and its spectrum is a single point. Theoretically, a sine wave exists over infinite time and never changes. The mathematical transform that converts the time domain waveform into the frequency domain is called the Fourier transform, and it compresses all the information in the sine wave over infinite time into one point. The fact that the peak in the spectrum shown above has a finite width is an artifact of the FFT analysis, which will be discussed later. A machine with imbalance has an excitation force that is a sine wave at 1X, or once per revolution. If the machine were perfectly linear in response, the resulting vibration would be a pure sine wave like the one shown above. In many poorly balanced machines, the waveform does resemble a sine wave, and there is a large vibration peak in the spectrum at 1X, or one order.

Here we see that a harmonic spectrum results from a periodic waveform, in this case a "clipped" sine wave. The spectrum contains equally spaced components, and their spacing is equal to 1 divided by the period of the waveform. The lowest of the components above zero frequency is called the fundamental, and the others are called harmonics. This waveform came from a signal generator, and it can be seen that it is not symmetrical about the zero line. This means it has a "DC." component, and this is seen as the first line at the left in the spectrum. This is to illustrate that a spectrum analysis can go all the way to zero frequency, or in common terminology, to DC. In vibration analysis of machinery, it is not usually desirable to include such low frequencies in the spectrum analysis for several reasons. Most vibration transducers do not have response to DC, although there are accelerometers that are used in inertial navigation systems that do have DC response. For machine vibration, the lowest frequency that is generally considered of interest is about 0.3 orders. In some machines this will be below 1Hz. Special techniques are required to measure and interpret signals below this frequency. Note that because this spectrum consists of discretepoints, the signal is by definition deterministic! It is not uncommon in machine vibration signatures to see a waveform which is clipped something like the one shown above. What this usually means is there is looseness in the machine, and something is restricting its motion in one direction.

The signal shown above is similar to the previous one, but it is clipped on both positive and negative sides, resulting in a symmetrical waveform. This type of signal can occur in machinevibration if there is looseness in the machine and motion is restricted in both directions. The spectrum seems to have harmonics, but they are actually only the oddnumbered harmonics. All the even-numbered harmonics are missing. Any periodic waveform that is symmetrical will have a spectrum with only odd harmonics! The spectrum of a square wave would also look like this. Sometimes the vibration spectrum of a machine will resemble this if there is extreme looseness and the motion of the vibrating part is restricted at each extreme of displacement. An unbalanced machine with a loose hold-down bolt is an example of this.

Modulation Effects
Modulation is a non-linear effect in which several signals interact with one another to produce new signals with frequencies not present in the original signals. Modulation effects are the bane of the audio engineer, for they produce "intermodulation distortion", which is annoying to the music listener. There are many forms of modulation, including frequency and amplitudemodulation, and the subject is quite complex. We will now look at the two primary types of modulation individually.

It is rare to see frequency modulation by itself; most machines will

Frequency modulation (FM) is the varying in

produceamplitude modulation at the same time asfrequency modulation!

frequency of one signal by the influence of another signal, usually of lower frequency. The frequency being modulated is called the "carrier". In the spectrum shown above, the largest component is the carrier, and the other components which look like harmonics, are called "sidebands". These sidebands are symmetrically located on either side of the carrier, and their spacing is equal to the modulating frequency.

Frequency modulation occurs in machine vibration spectra, especially in gearboxes where the gear mesh frequency is modulated by the rpm of the gear. It also occurs in some sound system loudspeakers, where it is called FM distortion, although it is generally at a very low level.

This example shows amplitude modulation at about 50% of full modulation Notice that the frequency of the waveform seems to be constant and that it is fluctuating up and down in level at a constant rate. This test signal was produced by rapidly varying the gain control on a function generator while recording the signal. This type of signal is often produced by defective bearings and gears, and can be easily identified by thesidebands in thespectrum. The spectrum has a peak at the frequency of the carrier, and two more components on each side. These extra components are the sidebands. Note that there are only two sidebands here compared to the great number produced by frequency modulation. The sidebands are spaced away from the carrier at the frequency of the modulating signal, in this case at the frequency at which the control knob was wiggled. In this example, the modulating frequency is much lower than the modulated or carrier frequency, but the two frequencies are often

close together in practical situations. Also these frequencies are sine waves, but in practice, both the modulated and modulating signals are often complex. For instance, the transmitted signal from an AM radio station contains a high-frequency carrier, and many sidebands resulting from the carrier modulation by the voice or music signal being broadcast. A vibration and acoustic signature similar to this is frequently produced by electric motors with rotor bar problems.

Beats

It is almost impossible to tell beating This waveform looks like amplitude fromamplitude modulation by looking at modulation, but is actually just two sine thewaveform, but they are fundamentally wave signals added together to form beats. different processes, caused by different Because the signals are slightly different in phenomena in machines. The spectrum tells the frequency, their relative phase varies from story. zero to 360 degrees, and this means the combined amplitude varies due to reinforcement and partial cancellation. The spectrum shows the frequency and amplitude of each component, and there are no sidebands present. In this example, the amplitudes of the two beating signals are different, causing incomplete cancellation at the null points between the

maxima. Beating is a linear process -- no additional frequency components are created. Electric motors often produce sound and vibration signatures that resemble beating, where the beat rate is at twice the slip frequency. This is not actually beating, but is in factamplitude modulation of the vibration signature at twice the slip frequency. Probably it has been called beating because it sounds somewhat like the beats present in the sound of an out of tune musical instrument. The following example of beats shows the combined waveform when the two beating signals are the same amplitude. At first glance, this looks like 100% amplitude modulation, but close inspection of the minimum amplitude area shows that the phase is reversed at that point.

' This looks like 100% amplitude modulation! This example of beats is like the previous one, but the levels of the two signals are the same, and they cancel completely at the nulls. This complete cancellation is quite rare in actual signals encountered in rotating equipment.

Earlier we learned that beats and amplitude modulation produce similar waveforms. This is true, but there is a subtle difference. These waveforms are enlarged for clarity. Note that in the case of beats, there is a phase change at the point where cancellation is complete.

Logarithmic Frequency Scaling

So far, the only type of frequency analysis discussed has been on a linear frequency scale, i.e., the frequency axis is set out in a linear fashion. This is suitable for frequency analysis with a frequency resolution that is constant throughout the frequency range, commonly called "narrow band" analysis. The FFT analyzer performs this type of analysis. There are several situations where frequency analysis is desired, but narrow band analysis does not present the data in its most useful form. An example of this is acoustic noise analysis where the annoyance value of the noise to a human observer is being studied. The human hearing mechanism is responsive to frequency ratios rather than actual frequencies. Thefrequency of a sound determines its pitch as perceived by a listener, and a frequency ratio of two is a perceived pitch change of one octave, no matter what the actual frequencies are. For instance if a sound of 100 Hz frequency is raised to 200 Hz, its pitch will rise one octave, and a sound of 1000 Hz, when raised to 2000 Hz, will also rise one octave in pitch. This fact is so precisely true over a wide frequency range that it is convenient to define the octave as a frequency ratio of two, even though the octave itself is really a subjective measure of a sound pitch change. This phenomenon can be summarized by saying that the pitch perception of the ear is proportional to the logarithm of frequency rather than to frequency itself. Therefore, it makes sense to express the frequency axis of acoustic spectra on a log frequency axis, and this is almost universally done. For instance, the frequency response curves that sound equipment manufacturers publish are always plotted in log frequency. Likewise, when frequency analysis of sound is performed, it is very common to use log frequency plots. The vertical axis of an octave bandspectrum is usually scaled indB. The octave is such an important frequency interval to the ear that so-called octave band analysis has been defined as a standard for acoustic analysis. The figure below shows a typical octave band spectrum where the ISO standard center frequencies of the octave bands are used. Each

octave band has a bandwidth equal to about 70% of it center frequency. This type of spectrum is called constant percentage band because each frequency band has a width that is a constant percentage of its center frequency. In other words, the analysis bands become wider in proportion to their center frequencies.

It can be argued that the frequency resolution in octave band analysis is too poor to be of much use, especially in analyzing machine vibration signatures, but it is possible to define constant percentage band analysis with frequency bands of narrower width. A common example of this is the one-third-octave spectrum, whose filter bandwidths are about 27 % of their center frequencies. Three one-third octave bands span one octave, so the resolution of such a spectrum is three times better than the octave band spectrum. Onethird octave spectraare frequently used in acoustical measurements. A major advantage of constant percentage band analysis is that a very wide frequency range can be displayed on a single graph and the frequency resolution at the lower frequencies can still be fairly narrow. Of course, the frequency resolution at the highest frequencies suffers, but this is not a problem for some applications such as fault detection in machines. In the chapter on machine fault diagnosis, it will be seen the narrow band spectra are very useful in resolving higher-frequency harmonics and sidebands, but for the detection of a machine fault, no such high resolution is required. The vibration velocity spectra of most machines will be found to slope downwards at the highest frequencies, and a constant percentage band (CPB) spectrum of the same data will usually be more uniform in level over a broad frequency range. This means that a CPB spectrum takes better advantage of thedynamic range of the instrumentation. One-third octave spectra are sufficiently narrow at low frequencies to show the first few harmonics of run speed, and can be used effectively for the detection of faults if trended over time. The use of constant CPB spectra for machine monitoring is not very well recognized in industry with a few notable exceptions such as the US Navy submarine fleet.

Overview
An early vibrationtransducer is the human finger! An earlier, and much more sensitive one is the lateral line organ of the fishes. The vibration transducer is a device that produces an electric signal that is a replica, or analog, of the vibratory motion it is subjected to. A good transducer should not add any spurious components to the signal, and should produce signals uniformly over the frequency range of interest.

Different types of transducers respond to different parameters of the vibration source, as shown in the following table: Name: Sensitive To:
Proximity Probe Velocity Probe Accelerometer Displacement Velocity Acceleration

On the following pages, we will examine the characteristics of these transducers

The Proximity Probe

One very common type The Proximity Probe, also called an "Eddy Current Probe" or ofproximity probe is known "Displacement Transducer", is a permanently mounted unit, and commercially as a requires a signal-conditioning amplifier to generate an output "Proximiter", which is a voltage proportional to the distance between the transducer end trademark of the Bentley and the shaft. It operates on a magnetic principle, and is thus Nevada Company. sensitive to magnetic anomalies in the shaft -- care should be taken that the shaft is not magnetized to assure the output signal is not contaminated. It is important to realize that the transducer measures relative displacement between the bearing and the journal, and does not measure total vibration level of the shaft or the housing. The displacement transducer is very commonly installed in large machines with journal bearings where it is used to detect bearing failure and to shut the machine down before catastrophic failure occurs.

These transducers are frequently used in pairs oriented 90 apart, and can be connected to the vertical and horizontal plates of an oscilloscope to display the "orbit", or path of the journal as it migrates around in the bearing. The frequency response of the displacement transducer extends from DC (0 Hz) to about 1000 Hz.

The Velocity Probe

Velocity Transducer Some velocity transducers are made with a moving coil outside a stationary magnet. The principle of operation is the same. Another type of velocity transducer consists of anaccelerometer with a built-in electronic integrator. This unit is called a "Velometer", and is by all accounts superior to the classic seismic velocity probe The velocity probe was one of the first vibration transducers to be built. It consists of a coil of wire and a magnet so arranged that if the housing is moved, the magnet tends to remain stationary due to its inertia. The relative motion between the magnetic field and the coil induces a current that is proportional to the velocity of motion. The unit thus produces a signaldirectly proportional to vibration velocity. It is self-generating and needs no conditioning electronics in order to operate, and it has a relatively low electrical output impedance making it fairly insensitive to noise induction. In spite of these advantages, the velocity transducer has many disadvantages that make it nearly obsolete for new installations, although there are many thousands of them still in use today. It is relatively heavy and complex and thus expensive, and it has poor frequency response, extending from about 10 Hz to 1000 Hz. The spring and the magnet make up a low-frequency resonant system with a natural frequency of about 10 Hz. This resonance needs to be highly damped to avoid a large peak in the response at this frequency. The problem is that the damping in any practical design is temperature sensitive, and this causes the frequency response and phase response to be temperature dependent.

The Accelerometer

Piezo-Electric Accelerometer The compression-type accelerometer, diagrammed here, was the first type to be developed. The shear type, which is arranged so the active element is subjected to shear forces, is generally preferred. There are also other designs for accelerometers The piezo-electric accelerometer can be considered the standard vibration transducer for machine vibration measurement. It is made in several different configurations, but the illustration of the compression type serves to describe the principle of operation. The seismic mass is clamped to the base by an axial bolt bearing down on a circular spring. The piezo-electric element is squeezed between the mass and the base. When a piezoelectricmaterial experiences a force, it generates an electric charge between its surfaces. There are many such materials, with quartz being one of the most commonly used. There are also synthetic ceramic piezo materials that work well, and in some cases, work at higher temperatures than quartz is able to do. If the temperature of a piezo material is increased, finally the so called "curie point", or "curie temperature" is reached, and the piezoelectric property is lost. Once this happens, the transducer is defective and not repairable. When the accelerometer is moved in the up and down direction, the force required to move the seismic mass is born by the active element. According to Newton's second law, this force is proportional to the acceleration of the mass. The force on the crystal produces the output signal, which is therefore proportional to the acceleration of the transducer. Accelerometers are inherently extremely linear in an amplitude sense, meaning they have a very large dynamic range. The smallest acceleration levels they can sense are determined only by the electricalnoise of the electronics, and the highest levels are limited only by the destruction of the piezo element itself. This range of acceleration levels can span an amplitude range of about 108, which is 160 dB! No other transducer can match this performance. The piezo-electric accelerometer is very stable over long periods of time, and will maintain its calibration if it is not abused. The two ways that accelerometers can be damaged are subjecting them to excessive heat and dropping onto a hard surface. If dropped more than a few feet onto a concrete floor or steel deck, the accelerometer should be re-calibrated to be sure the crystal is not cracked. A small crack will cause the sensitivity to be reduced and also will greatly affect the resonance, and thus the frequency response. It is a good idea to calibrate accelerometers about once a year if they are in service with portable data collectors.

The frequency range of the accelerometer is very wide, extending from very low frequencies in some units to several tens of kilohertz. The high-frequency response is limited by theresonance of the seismic mass coupled to the springiness of the piezo element. This resonance produces a very high peak in the response at the natural frequency of the transducer, and this is usually somewhere near 30 kHz for commonly used accelerometers. A rule of thumb is that an accelerometer is usable up to about 1/3 of its natural frequency. Data above thisfrequency will be accentuated by the resonant response, but may be used if the effect is taken into consideration. When using an ICP accelerometer, care must be taken not to subject it to accelerationlevels where the output voltage will exceed several volts. Otherwise, the internal preamplifier will be overloaded and datadistortion will result! Most accelerometers used in industry today are of the "ICP" type, meaning they have in internal integrated circuit preamplifier. This preamp is powered by a DC polarization of the signal lead itself, so no extra wiring is needed. The device the accelerometer is connected to needs to have this DC power available to this type of transducer. The ICP accelerometer will have a low-frequency roll-off due to the amplifier itself, and this is usually at 1 Hz for most generally available ICP units. There are some that are specially designed to go to 0.1 Hz if very low frequency data is required.

When an ICP accelerometer is connected to the power source, it takes a few seconds for the amplifier to stabilize, and during this time, any data the unit is collecting will be contaminated by a slowly varying voltage ramp. For this reason, there must be a time delay built into data collectors to assure the unit is stable. If the delay is too short, the timewaveform will have an exponentially shaped voltage ramp superimposed on the data, and the spectrum will show a rising very low-frequency characteristic sometimes called a "ski slope". This should be avoided because the dynamic range of the measurement is compromised. The resonant frequency of an accelerometer is strongly dependent on its mounting. The best type of mounting is always the stud mount -- anything else will reduce the effectivefrequency range of the unit.

When mounting an accelerometer, it is important that the vibration path from the source to the accelerometer is as short as possible, especially if rolling element bearing vibration is being measured.

Spectrum Analysis
Spectrum analysis, which is defined as the transformation of a signal from a time-domain representation into a frequency-domain representation, has its roots in the early 19th century, when several mathematicians were working on it from a theoretical basis. But it took a practical man, an engineer with a good mathematical background, to develop the rationale upon which almost all our modern spectrum analysis techniques are based. That engineer was Jean Baptiste Fourier, and he was working for Napoleon during his invasion of Egypt on a problem of overheating cannons when he derived the famous Fourier Series for the solution of heat conduction. It may seem a far cry from overheating cannons to frequency analysis, but it turns out that the same equations apply to both cases. Fourier later generalized the Fourier series into the Fourier Integral Transform. The advent of digital signal analysis naturally led to the so-called Discrete Fourier Transform and the Fast Fourier Transform or FFT

Forms of the Fourier Transform

There are four forms of the Fourier Transform, as follows: Fourier Series -- Transforms an infinite periodic time signal into an infinite discrete frequency spectrum. Fourier Integral Transform -- Transforms an infinite continuous time signal into an infinite continuous frequency spectrum Discrete Fourier Transform (DFT) -- Transforms a discrete periodic time signal into a discrete periodic frequency spectrum Fast Fourier Transform -- A computer algorithm for calculating the DFT They will be discussed in more detail in the next section.

The Fourier Series

The Fourier Series operates on a time signal that is periodic, i.e., a time signal whose waveform repeats over and over again out to infinite time. Fourier showed that such a signal is equivalent to a collection of sine and cosine functions whose frequencies are multiples of the reciprocal of the period of the time signal. The rather unexpected result is that any wave shape whatsoever, as long as it is not infinite in length, can be represented, as the sum of a collection of harmonic components, and the fundamental frequency of the harmonic series is 1 divided by the length of the wave shape. The amplitudes of the various harmonics are called the Fourier coefficients, and their values can be calculated easily if the equation for the wave shape is known. They can also be calculated graphically from the wave shape itself. A certain physics class is known to have done this with the silhouette of Marilyn Monroe. They posted the MM coefficients on the bulletin board as an "in" joke.

Fourier Coefficients
The calculation of the Fourier coefficients is defined as a mathematical transformation from the time domain to the frequency domain. One important fact emerges from the Fourier Series, and that is that the original waveform can be reconstructed from the frequency coefficients; in other words it is possible to transform from the frequency domain back to the time domainwithout loss of information. The Fourier series is perfectly adequate for performing frequency analysis on periodic waveforms; that is to say on deterministic signals.

The Fourier Integral Transform

The natural extension of the Fourier series to encompass time signals of infinite length, i.e., non-repetitive continuous signals, is the Fourier Integral Transform, or more simply the Fourier Transform. This integral will transform any continuous time signal of arbitrary shape into a continuous spectrum extending to infinite frequency. An interesting characteristic of the Fourier Transform is that an event encompassing a short time interval will be spread out over a wide frequency range and vice versa. This was seen in the Introduction to Vibration chapter where aspectrum of a short impulse is shown.

The Discrete Fourier Transform

Neither the Fourier Series nor the Fourier Transform lends itself easily to calculation by digital computers. To overcome this hurdle, the so-called Discrete Fourier Transform, or DFT was developed. Probably the first person to conceive the DFT was Wilhelm Friederich Gauss, the famous 19th century German mathematician, although he certainly did not have a digitalcomputer on which to implement it. The DFT operates on a sampled, or discrete, signal in the time domain, and generates from this a sampled, or discrete, spectrum in the frequency domain. The resulting spectrum is an approximation of the Fourier Series, an approximation in the sense that information between the samples of the waveform is lost. The key to the DFT is the existence of the sampled waveform, i.e., the possibility of representing the waveform by a series of numbers. To generate this series of numbers from an analog signal, a process of sampling and analog to digital conversion is required. The sampled signal is a mathematical representation of the instantaneous signal level at precisely defined time intervals. It contains no information about the signal between the actual sample times. If the sampling rate is high enough to ensure a reasonable representation of the shape of the signal, the DFT does produce a spectrum very close to a theoretically true spectrum. Thisspectrum is also discrete, and there is no information between the samples, or "lines" of the spectrum. In theory, there is no limit to the number of samples that can be used, or the speed of the sampling, but there are practical limitations we must live with. Most of these limitations are the result of using a digital computer as the calculating agent.

The Fast Fourier Transform

In order to adapt the DFT for use with digital computers, the so-called Fast Fourier Transform (FFT) was developed. The FFT is simply an algorithm for calculating the DFT in a fast and efficient manner. Cooley and Tukey are credited with the discovery of the FFT in 1967, but it existed much earlier, although without the digital computers needed to exploit it. The FFT algorithm places certain limitations on the signal and the resulting spectrum. For instance, the sampled signal to be transformed must consist of a number of samples equal to a power of two. Most FFTanalyzers allow 512, 1024, 2048, or 4096 samples to be transformed. The frequency range covered by FFT analysis depends on the number of samples collected and on the sampling rate, as will be explained shortl

Analog to Digital Conversion

The first step in performing an FFT analysis is the actual sampling process, which is illustrated here:

Analog to Digital Conversion The sampling is an analog, not digital, process and is accomplished with a "sample and hold" circuit. The output of this circuit is a sequence of voltage levels that are fed into an analog todigital converter (ADC). Here the voltage levels are converted into digital words representing each sampled level. The accuracy of the sampled levels depends in part on the number of bits in the digital words. The greater the number of bits, the lower the noise level and the greater the dynamic range will be. Most FFT analyzers use 12-bit words and this produces adynamic range of about 70 dB (3,100:1). Fourteen bit words can achieve 80 dB (10,000:1) dynamic range. It can be seen here that the sampling rate determines the highest frequency in the signal that can be encoded. The sampled waveform cannot know anything about what happens in thesignal between the sampled times. Claude Shannon, the developer of the branch of mathematics called information theory, determined that to encode all the information in a signal being sampled, the sampling frequency must be at least double the highest frequency present in the signal. This fact is sometimes called the Nyquist criterion.

Aliasing
It is important that there is no information in the sampled waveform near the sampling frequency to avoid a problem called aliasing.

Aliasing Here the actual signal is represented in black and the sampled representation of it is in gray. The vertical lines represent the sampling frequency. Note that if the sampling frequency is the same as the sampled frequency, each sample is the same size, and the output of the sampling circuit will be a constant direct voltage -- obviously having no relation to frequency of the input signal. Now note what happens if the actual signal is higher in frequency than the sampling frequency. The sampler output looks like a very low frequency, and again it is not a correct representation of the actual signal. This phenomenon is called aliasing, and it can lead to gross errors unless it is avoided. The best way to avoid aliasing is to pass the input signalthrough an analog low-pass filter whose cut-off frequency is less than one-half the sampling frequency. In most modern FFT analyzers, the sampling frequency is set to 2.56 times thefilter cut-off frequency. The filter must have a very sharp cut off characteristic, or roll off, and this means it will also have Phase Shift that can affect the data if one needs phaseinformation near the upper end of the frequency span of the analyzer. To avoid this, select a frequency span so the frequency in question is in the lower half of the frequency range. This is important in performing balancing with an FFT analyzer, where phase of the 1X vibration signal is needed. Aliasing also occurs in other media, such as motion pictures. For instance, sometimes in western movies the wagon wheel spokes may appear stopped, or rotating backward. This is optical aliasing, for a movie is a sampled representation of the original motion. Another example of optical aliasing is the stroboscope, which is set to flash at a rate equal to or near the rotation rate of the object being observed, making it appear stationary or slowly turning. Sampling Rules for Digital Signal Analysis The data path must contain an analog Anti-Aliasing low-pass filter You must sample at least twice as fast as the highest frequency to be analyzed The Frequency Response of the analysis depends on the sampling frequency

These rules apply to all FFT analysis, and the analyzer automatically takes care of them. The anti-aliasing filter is internally set to the appropriate value for each frequency range of the analyzer. The total sampling time is called the time record length and the nature of the FFT dictates that the spacing between the frequency components in the spectrum (also called thefrequency resolution) is 1 divided by the record length. For instance, if the frequency resolution is one Hz, then the record length is one second, and if the resolution is 0.1 Hz, then the record length is 10 seconds, etc. From this it can be seen that in order to perform high resolution spectrum analysis relatively long times are required to collect the data. This has nothing to do with the speed of the calculations in the analyzer; it is simply a natural law of frequency analysis.

Leakage
The FFT analyzer is a batch processing device; that is it samples the input signal for a specific time interval collecting the samples in a buffer, after which it performs the FFT calculation on that "batch" and displays the resulting spectrum If a sinusoidal signal waveform is passing through zero level at the beginning and end of the time record, i.e., if the time record encompasses exactly an integral number of cycles of thewaveform, the resulting FFT spectrum will consist of a single line with the correct amplitude and at the correct frequency. If, on the other hand, the signal level is not at zero at one or both ends of the time record, truncation of the waveform will occur, resulting in a discontinuity in the sampled signal. This discontinuity is not handled well by the FFT process, and the result is a smearing of the spectrum from a single line into adjacent lines. This is called "leakage"; it is as if the energy in the signal "leaks" from its proper location into the adjacent lines. The shape of the "leaky" spectrum depends on the amount of signal truncation, and is generally unpredictable for real signals.

Windows
In order to reduce the effect of leakage, it is necessary to see to it that the signal level is zero at the beginning and end of the time record. Multiplying the data samples by a socalled windowing or weighting function, which can have several different shapes, does this. The most common forms of windows and their uses are considered next.

If there is no windowing function used, this is called "Rectangular", "Flat", or "Uniform" windowing. In the figure above, the effect of the data truncation can be seen as discontinuities in the windowed waveform. The FFT analyzer only knows what is in the time window, or time record. It assumes the actual signal contains the discontinuities, and they are the cause of theleakage seen in the previous figure. Leakage could be avoided if the input waveform zero crossings were synchronized with the sampling times, but this is impossible to achieve in practice.

Windowing for Transient Signals

In the case where the input signal is a transient, it will by definition begin and end at zero level, and as long as it is entirely within the time record, no truncation will occur, and the analysis will be correct because the FFT sees the entire signal. It is very important that the entire transient fit into the record, and the record length is dependent upon the frequency range of the analysis. Most FFT analyzers allow the user to see the time record on the screen, so it can be assured that this condition is met.

The Hanning Window

The Hanning window, after its inventor whose name was Von Hann, has the shape of one cycle of a cosine wave with 1 added to it so it is always positive. The sampled signal values are multiplied by the Hanning function, and the result is shown in the figure. Note that the ends of the time record are forced to zero regardless of what the input signal is doing. While the Hanning window does a good job of forcing the ends to zero, it also adds distortion to the wave form being analyzed in the form of amplitude modulation; i.e., the variation inamplitude of the signal over the time record. Amplitude Modulation in a wave form results in sidebands in its spectrum, and in the case of the Hanning window, these sidebands, or side lobes as they are called, effectively reduce the frequency resolution of the analyzer by 50%. It is as if the analyzer frequency "lines" are made wider. In the illustration here, the curve is the actual filter shape that the FFT analyzer with Hanning weighting produces. Each line of the FFT analyzer has the shape of this curve -only one is shown in the figure. If a signal component is at the exact frequency of an FFT line, it will be read at its correct amplitude, but if it is at a frequency that is one half of delta F (One half the distance between lines), it will be read at an amplitude that is too low by 1.4 dB. The illustration shows this effect, and also shows the side lobes created by the Hanning window. The highest-level side lobes are about 32 dB down from the main lobe.

The measured amplitude of the Hanning weighted signal is also incorrect because the weighting process removes essentially half of the signal level. This can be easily corrected, however, simply by multiplying the spectral levels by two, and the FFT analyzer does this job. This process assumes the amplitude of the signal is constant over

the sampling interval. If it is not, as is the case with transient signal, the amplitude calculation will be in error, as shown in the figure below.

The Hanning window should always be used with continuous signals, but must never be used with transients. The reason is that the window shape will distort the shape of the transient, and the frequency and phase content of a transient is intimately connected with its shape. The measured level will also be greatly distorted. Even if the transient were in the center of the Hanning window, the measured level would be twice as great as the actual level because of the amplitude correction the analyzer applies when using the Hanning weighting. A Hanning weighted signal actually is only half there, the other half of it having been removed by the windowing. This is not a problem with a perfectly smooth and continuous signal like asinusoid, but most signals we want to analyze, such as machine vibration signatures are not perfectly smooth. If a small change occurs in the signal near the beginning or end of the time record, it will either be analyzed at a much lower level than its true level, or it may be missed altogether. For this reason, it is a good idea to employ overlap processing. To do this, two time buffers are required in the analyzer. For 50% overlap, the sequence of events is as follows: When the first buffer is half full, i.e., it contains half the samples of a time record, the second buffer is connected to the data stream and also begins to collect samples. As soon as the first buffer is full, the FFT is calculated, and the buffer begins to take data again. When the second buffer is filled, the FFT is again calculated on its contents, and the result sent to the spectrumaveraging buffer. This process continues on until the desired number of averages is collected.

Overlap Processing
Overlap processing can only be achieved if the time required to calculate the FFT is shorter than the time record length. If this is not the case, the spectral calculations will lag behind

the data acquisition leaving gaps of unanalyzed signal. See also the paragraph on real time speed later in this section.

If the overlap is 2/3, i.e., 66.7%, then the overall time weighting of the data will be flat, and there is no advantage to using a greater overlap. Most data collection for machinery analysis uses 50% data overlap, which provides adequate amplitude accuracy for most vibration work. Here is a summary of the relationship between sampling rate, number of samples, time record length, and frequency resolution that affect FFT analysis. The sampling rate in samples per second, times the time record length T in seconds, equals the number of samples N. In the FFT analyzer, the number of samples N is constrained to a power of two.

FFT Fundamentals

The FFT algorithm, operating on N samples of time data produces N/2 frequency lines. Thus a time record of 512 samples will generate a spectrum of 256 lines. FFT analyzers generally do not display the upper spectral lines because of the possibility of their being contaminated

by aliased components. This is because the anti-aliasing filter is not perfect, and has a finite slope in its cut-off range. Therefore, a 256 line spectrum will be displayed as a 200 line spectrum, and a 512-line spectrum will be displayed as a 400 line spectrum, etc. The frequency resolution, DF, is equal to the frequency span divided by the number of lines, and this is equal to 1/T. Conversely, the time record length T equals 1/DF. From this it can be seen that as the frequency resolution increases (smaller DF), the time record length also increases in proportion. For this reason, to create a high-resolution spectrum requires a relatively long time to acquire the data.

The Picket Fence Effect

As has been mentioned before, the FFT spectrum is a discrete spectrum, consisting of estimates of what the spectral level is at specific frequencies. These frequencies are determined by the analysis parameters that are set up in the analyzer, and have nothing to do with the signal being analyzed. This means there may be, and probably are, peaks in the true spectrum of the signal that are between the lines of the FFT analysis. This also means that in general, the peaks in an FFT spectrum will be measured too low in level, and the valleys will be measured too high. Moreover, the true frequencies where the peaks and valleys lie will not be those indicated in the FFT spectrum.

This phenomenon is called resolution bias error, or more commonly, the picket fence effect. In other words, looking at an FFT spectrum is a little like looking at mountain range through a picket fence.

Averaging
One of the important functions of the FFT analyzer is that it is easily able to do averaging of spectra over time. In general, the vibration signal from a rotating machine is not completelydeterministic, but has some random noise superimposed on it. Because the noise is unpredictable, it alters the spectrum shape, and in many cases can seriously distort the spectrum. If a series of spectra are averaged together, the noise will gradually assume a smooth shape, and the spectral peaks due to the deterministic part of the signal will stand out and their levels will be more accurately represented. It is not true that simply averaging FFT spectra will reduce the amount of the noise -- the noise will be smoothed but its level will not be reduced. There are two types of averaging in general use in FFT analyzers, called linear averaging and exponential averaging. Linear averaging is the adding together of a

number of spectra and then dividing the total by the number that was added. This is done for each line of the spectra and the result is a true arithmetic average on a line-by-line basis. Exponential averaging generates a continuous running average where the most recently collected spectra have more influence on the average than older ones. This provides a convenient form to examine changing data but still have the benefit of some averaging to smooth the spectra and reduce the apparent noisiness of them.

Time Synchronous Averaging

Time synchronous averaging, also called time domain averaging, is a completely different type of averaging, where the waveform itself is averaged in a buffer before the FFT is calculated. In order to do time domain averaging, a reference trigger pulse must be input to the analyzer to tell it when to start sampling the signal. This trigger is typically synchronized with an element of the machine that is of interest. The average gradually accumulates those portions of the signal that are synchronized with the trigger, and other parts of the signal, such as noise, are effectively averaged out. This is the only type of averaging which actually does reduce noise. More information on applications of time synchronous can be found in the next chapter on Machine Vibration Monitoring.

Pitfalls in the FFT

This is a summary of the pitfalls that plague the FFT analysis technique. This is not to say that FFT analysis is no good -- on the contrary, it has revolutionized the analysis of vibrationdata. The important fact is that the problems with FFT analysis can be overcome by proper technique, and the residual effects that remain can be reduced to insignificant levels. Sampling causes aliasing Time limitation causes leakage Discrete frequencies in the calculated spectrum causes the picket fence effect.

Introduction
It has been shown many times over that the vibration signature of an operating machine provides far more information about the inner workings of the machine than any other type of non-destructive test. A bearing that has a small developing defect will cause a telltale change in the machine vibration, as will an imbalance condition, a misalignment, or any of a myriad of other faults. Vibration analysis, properly applied, allows the technician to detect small developing mechanical defects long before they become a threat to the integrity of the machine, and thus provides the necessary lead-time to schedule maintenance to suit the needs of the plant management. In this way, plant management has control over the machines, rather than the other way around. Vibration measurement and analysis is the cornerstone of Predictive Maintenance, which stands in sharp contrast to the historical "run-to-failure" type of maintenance practice. Numerous studies, such as those conducted by the Electric Power Research Institute (EPRI), have shown that on average, the cost to industry for maintenance will be reduced by more than 50% if a predictive maintenance program is used instead of run-to-failure.

History of Vibration Analysis used for Machinery Maintenance

The first vibration meters were introduced in the 1950s, and they measured the overall, or "broad band" level of machine vibration, either in peak-to-peak mils (thousandths of an inch) of vibratory displacement, or in inches per second (IPS) of vibration velocity. A little later, tunable analog filters were added to the meters in order to discriminate between different frequencycomponents, and thus to produce a sort of vibration spectrum. The 1970s brought forth the personal computer and the advent of digital signal processing that led to the FFT analyzer, and it made quick work of calculating a frequency spectrum from a recorded vibration signal. The first such analyzers were quite bulky, weighing as much as 75 pounds, and this made them more suited as laboratory instruments than portable units for field use. The 1980s saw the exploitation of the microprocessor on a single silicon chip, and the battery-powered truly portable digital signal analyzer quickly followed this. It is this device, coupled with a computer program that stores the data and takes care of the logistics of vibration data collection that has revolutionized the application of vibration analysis to machinery diagnostics.

Test Point Location

In general, it is desirable to locate the test transducer as close as possible to the bearing with solid metal between the bearing and the sensor. Avoid bearing caps, which are of thin metal and are thus poor conductors of vibration energy. If possible, pick test point locations so that there is no metal-to-metal joint between the bearing and the sensor. The joint between the end bell and stator housing of a motor is an example of this. Fan housings on the ends of motors are also to be avoided.

Vibration Sensor Orientation

In any machinery-monitoring program, it is extremely important that the data is collected in exactly the same manner each time a measurement is taken. This is to assure that the data is repeatable and can be trended over time. For this reason, it is not recommended that hand-held transducers be used. By far the most reliable data is collected when the transducer is stud mounted to the machine surface.

In general, it has been found that for motors of less than about 50 HP, one test point is adequate, but for motors over 50 HP, each bearing should have its own test point. In any machines that are especially sensitive to bearing damage, and bearing problems should be detected as early as possible, each bearing should have its own test point. Another consideration in the integrity of the path between the bearing itself and the transducer: If the motor and bell is a solid casting, it will effectively transmit vibration with little loss of high frequencies, but if it contains one or more metal to metal connections, the high frequencies will be significantly distorted.

Triaxial Measurements
To assist in the determination of machine problems, it is very helpful to have vibration data from each measurement point in three directions. These directions are called Axial, Radial, andTangential. Axial is the direction parallel to the shaft in question, radial is the direction from the transducer to the center of the shaft, and tangential is 90 degrees from radial, tangent to the shaft.

Alignment of Vibration Axes

Orientation Examples

The following diagram shows the six possible orientations of the sensor for a horizontal machine.

For vertical machines, 'R' is Radial, 'T' is Tangential, and 'A' is vertical:

Sensor Mounting Pads -- "Blocking"

When using a triaxial accelerometer, it is extremely important that it be installed in exactly the same location each time the data is collected, and also that it be oriented in the same direction. One way of assuring this is to use permanently affixed mounting blocks on the machine. The cylindrical mounting block, or "pad", is a bronze disc with a central tapped hole and a key way at the edge that receives an indexing pin on the transducer itself. The transducer that is sensitive along the axis of the mounting screw is channel No. 1, the axis in the direction of the key way is channel No. 2, and the axis perpendicular to this is

channel No. 3. The pad is normally attached to the machine with a hard, strong adhesive such as Versilok type 204 structural adhesive. As was mentioned above, it is very important that the orientation of the block is known by the software, and if a block is replaced, the new one must be oriented in the same direction. The VTAG states the proper orientation of each block. The installation of the mounting blocks is sometimes referred to as "blocking" a machine.

Sensor Mounting Pad

Test Conditions
The vibration signature of a machine is strongly dependent on the operating parameters as well as its physical condition. These operating parameters include such things as running speed, load, pump discharge pressure, and compressor delivery pressure. The machine must be in its normal operating condition when vibration data is collected. If this is not the case, the vibration signature will not match the vibration signatures previously recorded, and trending vibration levels over time becomes impossible. Running speeds of induction motors depend on the load, and should not vary from one collection time to the next by more than a few percent. This means that load conditions must be as nearly as possible the same. The vibration level contributed by extraneous sources, such as nearby machines, must also be the same for each data collection time. Do not collect data with adjacent machines turned off if the previous spectra were recorded with them running. This is especially true with strong background vibration levels, as in the engine room of a ship. Propulsion diesels must be operating at the same speed for each data collection session!

Operating Conditions
It is imperative that when collecting data, the test RPM is very near the RPM that was used for the previous tests. In turbine-driven equipment, the speed should be verified by the use of a portable stroboscopic or other tachometer, and it must be running at a constant, not varying, speed! Gauge pressures should reflect normal operating conditions. Pump testing with discharge valves closed is discouraged, but if a pump must be tested in a recirculating condition, the recirculation valve may be partially closed to attain a normal discharge pressure.

Warm-up

All machines should be tested in a fully warmed-up condition. Machine temperature will affect alignment and operating clearances due to thermal expansion. A cold machine will have a different vibration signature than a warmed-up machine, sometimes extremely different.

Visual Inspection
Visual inspection of an operating machine while vibration testing is important, for valuable clues to machinery condition can often be uncovered. RPM and discharge pressure, etc., should be noted. The following items should be checked: Are there any unusual noises present? Do any bearings feel hotter than normal? Can you feel any excessive vibration level? Is there anything unusual about the operation of the machine? Are there any fluid or steam leaks obvious? Do the gauge readings look normal? Does the machine operator have any comments on machine condition?

Vibration Measurement Parameters

As we saw in the Introduction to Vibration chapter, it is possible to examine the same vibration signal in terms of Acceleration, Velocity, or Displacement. It is seen that velocity at anyfrequency is proportional to the displacement times the frequency, and the acceleration at any frequency is proportional to velocity times frequency, which means it is also equal todisplacement times frequency squared.

Machinery Testing Schedule

It is important to begin a vibration-monitoring program of manageable size and then gradually expand it as you gain experience. The most important machines to monitor should be those that are critical to the plant's productivity and/or have a poor maintenance record. Variable speed machines, extremely complex machines, and reciprocating machines should not be included at first. For a successful monitoring program, machinery measurements must be carried out on a scheduled periodic basis. Most equipment should be tested monthly, with certain less important machines on a 3-month schedule. Weekly testing is common for critical machines. In any case, it is important to tailor your measurement schedule to suit the machines and

their condition. As experience is gained, it will be easy to revise the testing schedule accordingly

Trending of Vibration Data

Trending is the storage of vibration signatures recorded at specific time intervals and plotting the changes in vibration levels at the forcing frequencies vs. time. An upward trend in levelindicates a developing problem. The simplest way to utilize the concept of vibration trending is to establish a representative vibration spectrum of a normally operating machine as a reference, and compare this reference to spectra measured at later times on the same machine. The comparison of the spectra is made possible by order normalization, which will be discussed shortly. When performing the spectral comparison there are several important points that need to be addressed: The operating conditions of the machine when measuring the new vibration data must match as closely as possible the conditions under which the reference spectrum was recorded. Otherwise, the spectra will not be comparable and gross errors can be made. The vibration data must be recorded in exactly the same way that the reference data were measured. The transducer must be mounted in exactly the same location, and its calibration must be accurate. If possible, the same transducer should be used for all successive measurements on the machine. When taking vibration data with an FFT analyzer, or data collector, it is important to average several instantaneous spectra together to reduce random variations and the effects of extraneousnoise in the measured signal. The number of spectral averages recorded to produce the spectra must be sufficient to produce a uniform and steady signature. Usually from six to ten averages will do this, but on some machines with a relatively high random noise content in their vibration signature, longer averaging times may be needed. A rule of thumb is to record a spectrum with several averages and then immediately record another one with twice as many averages. If the spectra are significantly different, the number of averages should be doubled

again and anotherspectrum recorded. If the latter two spectra are similar, then the previous number of averages is adequate for this machine.

The Reference Spectrum

When performing trending, it is extremely important to be sure that the reference spectrum to which the subsequent test spectra will be compared is truly representative of the machine.

Averaged Vibration Signatures

Long experience has shown that an excellent way to generate a meaningful reference is to average several spectra together from machines of the same type. If there are a number of similar machines in a plant the statistical average of their reference spectra is a good indication of the overall characteristics of that particular machine. A series of similar machines in good working order will produce vibration spectra that are similar to one another, but will have random variations in level. The spectra of the machines are averaged together and the standard deviations in level at each significant frequency are calculated. Some types of machines are so individualistic that when averaged together, the standard deviation between the vibration magnitudes is so great that the average is essentially meaningless. In this case, each machine must be used by itself to produce a meaningful reference by averaging a series of measurements over a fairly long time period, and generating a mask from this average reference spectrum. There are many situations where a large selection of similar machines is not available, and in this case, the averaged reference spectra are taken on the same machine at different times. When averaging spectra from a group of machines to make a reference spectrum, care must be taken to see that the spectra to be averaged are valid and that the machines they come from are not defective. One of the most important jobs of the vibration analyst is to be sure that the average reference spectra are valid and representative of the machines in question. Do not confuse reference spectrum averaging to produce a reference signature with spectrum averaging done at the time of vibration data collection, as described above.

The Spectrum Mask

As we have seen, healthy machines will show minor deviations in their vibration spectra because of small load variations, temperature variations, line voltage variations, and background noise level fluctuations. These variations in vibration signatures can cause false alarms to be generated if the raw spectrum is directly compared to a valid reference spectrum. For this reason, it is desirable to generate a socalled mask spectrum from the reference spectrum. The mask is a new spectrum made by increasing the levels in the reference spectrum by various amounts at different frequencies. For instance the mask might be 6 dB above the reference at 1X, but only 4 dB above the reference at 2X. A good staring point for establishing the mask is to add one standard deviation in level at each spectral peak to the averaged reference spectrum. A large class of machines will be found to produce averaged spectra with fairly small standard deviations, and with these machines in particular it is a good idea to perform the spectrum averaging and then generate the mask by adding one standard deviation to the average spectrum at each frequency. A group of machines which exhibits large standard deviations in level when making the reference will be more difficult to deal with in generating the mask, and the mask levels will have to be higher than one standard deviation above the reference. The determination of the shape of the mask spectrum can be fairly complicated, and it depends on the machine in question and normal variation in its vibration spectral levels at different frequencies. This can only be determined by looking at a series of historical spectra and applying good judgment and a good knowledge of the machine itself.

Order Normalization
Order normalization is performed by most condition monitoring software, and under certain conditions it is possible for the software to select the wrong peak as the 1X component. For this reason it is important for the analyst to verify that the normalization was correctly done if a spectrum looks vastly different than other spectra taken from the same machine. In such a case, the analyst must re-normalize the spectrum. Following is a non-order normalized spectrum, scaled from zero to 30,000 RPM.

Conventional Vibration Spectrum

Note that many peaks appear to be equally spaced, but it may be difficult to tell which one near 20,000 CPM is a shaft harmonic. The next figure is a normalized spectrum scaled from 0 to 10 orders. Note that the harmonics of turn speed are integers on the frequency scale, and that the peak below 7X is immediately seen as a non-synchronous component.

Order Normalized Spectrum

Order normalization of spectra has the following advantages: The fundamental turn speed is instantly recognizable at 1.0 order. Harmonics of the turn speed will be integers A second shaft in a gear-driven machine will have an order equal to the gear ratio Excitation frequencies, such as gear mesh and pump vane pass, are readily recognized because their order is equal to the number of elements Bearing tones will be non-integer, often the only major noninteger components Sidebands around bearing tones will be easily recognized because they will be at the tone order 1, 2, etc. Most important: Because machine speed is almost never exactly the same from test to test, the peaks in the spectrum will not be at the same frequencies, and the spectra cannot be averaged. Normalized spectra have the peaks at the same orders from test to test, and they can be averaged without smearing.

Evaluating Machine Vibration Spectra

Most machines have a relatively simple set of vibration forcing frequencies, determined by the geometry of the machine and its speed. The existence of other frequencies than the forcing frequencies, such as harmonics of 1X, in the vibration signature of the machine indicates non-linearities, and the combined magnitude of these new frequencies is a good indicator of the overall health of the machine. As a machine wears, its clearances typically become greater and its vibration signature becomes more complex due to generation of harmonics and sidebands. In trending the vibration level of a machine over time, a rise in the level of a forcing frequency indicates a change in the mechanism causing that particular forcing frequency, but does not necessarily indicate any damage to the machine. For instance, an increase in 1X at a motor bearing indicates an increasing imbalance condition, but if harmonics of 1X begin to appear, this indicates damage, such as bearing clearance increases, looseness, or cracking of the structure. Therefore, a strong 1X vibration means the rotor should be balanced, but the appearance ofharmonics of 1X means the bearing and surrounding structure should also be inspected for damage.

Introduction
The steps in manual (non-automatic) machine vibration analysis are: Identifying vibration peaks in the spectrum and relating them to forcing frequencies Determining the severity of machine problems from the amplitudes and relationships between the vibration peaks.

Making the appropriate repair recommendations based on the severity of the machine problem. In order to do a proper job of vibration analysis, several tools are needed: If the vibration spectra are being analyzed on a computer, a calculator and Vibration Test and Analysis Guide (VTAG) for the machine in question are required. If the vibration spectra have been printed on paper, then a straight edge and ten-point divider are desirable. Previous vibration data and average vibration data are also helpful if available. he VTAG contains important information about the design of the machine, the test points and their locations, the frequency ranges to be tested, and the forcing frequencies to be expected. The VTAG should be consulted before any vibration analysis is attempted. Following is an example of a VTAG:

Checking for Data Validity

After determining the shaft rotation rate and locating it on the spectrum (it will be the first order in a normalized spectrum), the vibration analyst must check the validity of the spectrum. Data validity can be corrupted by such things as incorrect labeling of accelerometer orientation or position, improper accelerometer attachment, rapid accelerometer temperature changes, and incorrect machine operating conditions. When data are to be compared to previously collected data from the same point, similar test conditions must be maintained, especially machine speed, load, and operating temperature. The integrity of the accelerometer cable is crucial to the collection of valid data. If the central conductor in the cable is intermittent or open, the measured signal will consist mostly ofrandom noise, and if the cable shield is intermittent or broken, the data will be contaminated with 60 Hz noise and harmonics. (50 Hz in countries with 50 Hz power lines.) In electrically driven machines, the 60 Hz line frequency will produce a series of 120 Hz harmonics in the vibration spectrum, as explained in the section on electrically induced vibration. If an accelerometer is exposed continuously to a higher temperature than that for which it is rated, it will become desensitized, and the data it senses from then on will be worthless. Some accelerometers will operate up to 400 degrees F, but most give up the ghost at about 200 degrees F. Care must be taken that the accelerometer is not dropped onto a hard surface lest the piezo-electric element be damaged. If the element is cracked, the stiffness of the internal assembly will decrease, reducing the resonant frequency of the accelerometer, and this can greatly change its sensitivity at high frequencies.

Step-by-Step Analysis of Spectra

In preparation for the diagnostic techniques described in the next chapter, the first steps of analysis should be performed as follows: This procedure assumes the vibration spectra are printed on paper. When viewing spectra on the computer screen, similar procedures are used, as explained in the software instructions. Note that all the following steps are greatly simplified if the spectra are order normalized.

Identifying the First Order (1X) Peak

The first step in machine vibration analysis is to identify the spectral peak corresponding to shaft rotation rate, or the so-called 1X peak. This will be the 1X in a normalized spectrum. It is important to check to be sure the normalization was done correctly. It is also called the first order peak. In multiple-shaft machines, each shaft will have a characteristic 1X peak, and these are then located by the analyst.

Single-shaft machine
Mark the harmonics of 1X on the spectra. This is simplified if you use a ten-point divider. Identify the fan blade pass frequency and mark it on the spectra. This is the number of blades multiplied by the RPM. Note the harmonics of blade pass frequency if they are prominent. Look for bearing tones, which are between the harmonics of the 1X run speed and not synchronous with it. Mark them on the spectra. There are other machine components besides bearings that generate non-synchronous tones. Probably the most common one is belt drives.

Multiple-shaft machine

Identify and mark the 1X and harmonics of the pump on the spectra. The pump RPM can be found from the VTAG, or can be calculated from the motor speed and gear ratio as follows: If the motor is turning 1780 and the gear ratio is 2.3 to 1, then the pump speed is:

Identify and mark the pump vane pass frequency and harmonics, if any, on the spectra. The vane pass is the number of vanes times the pump RPM. Search the spectra for non-synchronous components that could be bearing tones, or consult the VTAG for bearing tone frequencies, and mark them on the spectra. After performing all these tasks, you are ready for the work described in next chapter on Machine Diagnostics.

The Waveform vs. the Spectrum

Time Domain Analysis is simply the use of the waveform instead of the spectrum to help diagnose machine problems. As we learned in the frequency analysis section of the VibrationFundamentals course, the spectra of an impulse or transient and of a random signal may look almost exactly alike. This is true even though the parent time signals are very different in character. The waveform immediately shows the difference, however, and therefore it is a good idea for the analyst to examine the waveform when the spectrum may not provide all the information needed to make a complete diagnosis.

FFT Analyzer Setup for Waveform Collection

When setting up an analyzer to store waveforms, an important point should be born in mind, and that is that the frequency range normally convenient for looking at a spectrum is usually not suited to looking at the waveform. Most FFT analyzers, with a few notable exceptions, do not allow you to set up specific sampling rates or time domain record lengths you must set them up in terms of frequency span and frequency resolution. Remember from the FFT Analysis chapter that the time record length used by the analyzer to calculate the spectrum is the reciprocal of the line spacing, or resolution, of the spectrum. Spectra are generally scaled so relatively wide frequency ranges can be examined, and the FFT analyzer of necessity acquires a short time record. For instance, a 400line spectrumextending from DC to 1000 Hz will have a line spacing of 1000/400, or 2.5 Hz. The time record length used to calculate this spectrum is 1/2.5, or 0.4 seconds. This time record, which is the actual waveform, will show details that happen in that 0.4-second time span, but in practice, when looking at a machine vibration waveform, we are often looking for events that occur over a much longer time than that. If we are looking for beats in the vibration signature of an electric motor, or of the combined vibration of two machines running at slightly different speeds, we need to see a waveform that is at least several seconds long. To acquire a waveform lasting five seconds, we need to set up a line spacing of 1/5 Hz, and this can be done by adjusting the number of lines of resolution and the frequency span to suitable values. To find out the sampling rate of the waveform, and thus it time resolution, again we need to get the information from the spectrum characteristics. The sampling frequency of the time record for most analyzers is 2.56 times the highest frequency in the spectrum. Thus

a frequency span of 100 Hz implies a sampling frequency of 256 samples per second, and a span of 1000Hz requires a sampling rate of 2560 samples per second. Remember that a meaningful time record contains many more data points that the usual spectrum, and therefore you need to take care that you have enough memory available in your data collector to store the waveform data. For this reason, it is best to use the lowest sampling rate and the shortest time record length that will provide the needed data. For example if you just want to resolve beats in waveform that only occur once in several seconds, the sampling rate need not be very high 50 samples per second is probably fast enough. This corresponds to a frequency span of 50/2.56, or 19.53 Hz. So you can select 20 Hz in the frequency span set up. On the other hand, if you want to examine a waveform that might have interesting glitches at 50 times per second, then you need to sample fast enough to resolve each glitch. You might sample at 1000 samples per second, and this requires a frequency span of 1000/2.56, or about 390 Hz. A good rule of thumb to memorize is that the time record length depends only on the line spacing of the FFT spectrum and the sampling rate depends only on the frequency range of theFFT spectrum, and they are independently adjustable. We will return to this subject of time resolution versus frequency resolution soon, when we look into Synchronous Averaging.

Acceleration vs. Velocity

When collecting a waveform, it is important to note that we generally want to see highfrequency information. When integrating a signal from acceleration to velocity, the highfrequency response is greatly reduced in level, and many small irregularities in the waveform are not visible. Therefore, it is a generally valid rule of thumb that we should use vibration accelerationrather than velocity as the measurement parameter. However there are cases, especially in low-speed machines, when low-frequency information is interesting to us, and then integration to velocity, or even to displacement, can be used. Of course, the integration must be done to the input signal from the accelerometer. It is not possible to digitally integrate the acceleration waveform.

Phase in the Time Domain

The concept of phase is interesting in that it can be considered a property of the time domain waveform or a property of the frequency domain spectrum. Phase is nothing more than the time difference between a reference time and a measured time. It usually is represented in units of milliseconds in the waveform, and as degrees of angle in the spectrum. One complete rotation corresponds to 160 degrees of phase angle. As we will see later in this book when we look at diagnostics of machine problems, we will see that there are many different areas where phase gives us important diagnostic information,balancing being one of the most important. Even though phase really is defined in the time domain, it is usually much easier to measure and display in the frequency domain. For this reason, we will not address it further here.

Orbit Plots
Orbit plots are plots that appear on an oscilloscope screen when two proximity probes are connected to the horizontal and vertical inputs, respectively. The proximity probes are installed in a sleeve bearing, one oriented vertically and one oriented horizontally. Under these conditions, vertical motion of the shaft center line will move the oscilloscope dot vertically, and horizontal motion of the shaft will move the dot horizontally. Thus, when the shaft is turning, the dot will be stationary if the center of the shaft is stationary in the

bearing. If the shaft is migrating around in the bearing, the dot will follow, and it will trace out the motion of the shaft center in real time. The instrumentation setup is shown schematically below:

Under normal operating conditions, the scope will show a circle, indicating the oil film has equal stiffness and thickness in all directions, and there is some imbalance causing the centerline to move in a circle.

The orbit at the left above shows an ideal condition of the shaft in the bearing, while the one on the right shows that the shaft is moving more in the vertical direction than it is in the horizontal direction. This may mean the bearing is worn in an oval pattern, with more vertical clearance than horizontal clearance.

An orbit pattern like the one shown above indicates shaft motion that moves in a figure 8 pattern. It is vibrating twice as fast in one direction than in the other one. This can be caused by excessive clearance in one direction, or a bearing worn into an oval shape. A pattern like this is a danger sign, for the journal is likely to develop metal-to-metal contact with the bearing, causing extensive damage to both.

Impacts vs. Random Noise

Impacts may be caused by rolling element bearings where the rollers encounter a crack or small spall in a race. If there is a lot of external noise present, the spectrum may not show a well-defined peak at the bearing tone frequency. However, the acceleration waveform will usually exhibit repetitive peaks with their repetition rate equal to the period of the bearing ball pass frequency.

Random of Non-Periodic Impacts

A loose machine component that hits something at a rate that is unrelated to machine speed will generate impacts that are usually not precisely periodic, and can be essentially random in their spacing. In the vibration spectrum of such a condition, the impacts will produce a continuous noise spectrum that can extend to very high frequencies. The spectrum could be confusing in that other sources or wide band noise will look almost the same. In the waveform, the impacts are clearly seen.

Truncation or flattening of the signal

In many cases of looseness, such as a bearing pillow block that rises up a little during part of the rotation and then contacts the base for the rest of the cycle, the waveform will be flattened on one side. This will result in harmonics in the spectrum, but other types of waveform distortion will also produce harmonics. The waveform gives a quick identification of this kind of looseness where motion is restricted in one direction.

Low-frequency events
In some cases the vibration signal might have a discontinuity once in a while that when transformed into the frequency domain will be so low in frequency it will not be clearly seen in thespectrum. An example of this is a low-speed gearbox which has one broken or cracked tooth on the large gear. A waveform recorded over a long time will show the discontinuities.

Crest Factor
The Crest factor, also sometimes called the "peak-to-RMS-ratio", is defined as the ratio of the peak value of a waveform to its RMS value. It is a pure number, without units. The crest factor of a sine wave is 2, or1.414; i.e. the peak value is 1.414 times the RMS value. A typical vibration signal from a machine with a large imbalance and no other problems will have acrest factor of about 1.5, but as the bearings begin to wear, and impacting begins to happen, the crest factor will become much greater than this. The reason that the crest factor is so sensitive to the existence of sharp peaks in the waveform is that the peaks do not last very long in time, and therefore do not contain very much energy. The RMS value is proportional to the amount of energy in the vibration signal. Examples of actual crest factors measured on a cam rider on a large low-speed machine are shown below. The cam rider contains a roller bearing. The vibration signature, which is scaled in velocity units (ips), shows a little irregularity that is no doubt due to some roughness in the bearing. There is essentially no low-frequency motion here because the accelerometer is oriented in the radial direction of the bearing shaft, which is at 90 degrees from the motion of the cam rider. Note the RMS value is 0.017 ips, and the crest factor is 8.9. In other words, the peak value is 8.9 times the RMS value.

Radial Velocity Waveform Without Fault

The next figure, below, shows the same measurement point, at a later time. Here, the waveform shows that the bearing has an obvious fault in one of the races. The RMS level remains low at 0.086 ips, but the crest factor has risen to 19. This shows that a simple vibration meter that is only sensitive to RMS level is not able to detect a defective bearing, at least in this case.

The next figure, below, was measured at the same point on the machine as the previous waveforms, but in the direction of motion of the cam rider rather than perpendicular to it. The measurement was made before the bearing developed the fault. Here, we see the lowfrequency content due to the movement of the cam. The small noise bursts are caused by the minor bearing damage as was shown in the first figure, above. Note the RMS value is 0.45 ips and the crest factor is 1.7.

Tangential Waveform Without Fault The figure below is from the same measurement point and direction as the one above, except it was recorded after the bearing fault developed, as in the second figure, above.

Tangential Velocity Waveform With Fault Note that the bearing fault is clearly visible in the sharp spikes similar to the ones in the Radial with fault figure above. Note also that the RMS value is 0.45 ips, the same as in the previous figure, indicating that the bearing fault did not add significantly more energy to the vibration signature. The interesting fact here is that the crest factor of 1.8 is only slightly higher than before, even though the sharp vibration spikes are present. In this case, the large low-frequency signal masks the spikes and they do not show up as an elevated crest factor. This is a good illustration that the crest factor alone can sometimes be misinterpreted unless the vibration waveform is actually observed. This condition does no occur very often in practice, however.

Synchronous Averaging
Synchronous averaging, also sometimes redundantly called Time Synchronous averaging, was discussed earlier as a method of background noise reduction in spectra of complex signals. Now, we will look at it as a means of greatly increasing the information we can extract from the time-domain vibration waveform. Synchronous averaging is a fundamentally different process than the usual spectrum averaging that is generally done in FFT analysis. It is used to greatly reduce the effects of unwantednoise in the measurement. The waveform itself is averaged in a time buffer before the FFT is calculated, and the sampling of the signal is initiated by a trigger pulse input to the analyzer. If the trigger pulse is synchronized with the repetition rate of the signal in question, the averaging process will gradually eliminate the random noise because it is not synchronized with the trigger. However, the signal that is synchronous with the trigger will be emphasized, as shown below:

When you do time domain averaging on the vibration signal from a real machine, the averaged time record gradually accumulates those portions of the signal that are synchronized with the trigger, and other parts of the signal, such as noise and any other components such as other rotating parts of the machine, etc., are effectively averaged out. This is the only type of averaging that actually does reduce noise. An important application of time synchronous averaging is in the waveform analysis of machine vibration, especially in the case of gear drives. In this case, the trigger is derived from a tachometer that provides one pulse per revolution of a gear in a machine. This way, the time samples are synchronized in that they all begin at the same exact point in the angular position of the gear. Consider a gearbox containing a pinion with 13 teeth and a driven gear with 31 teeth. If a tachometer is connected to the pinion shaft, and its output is used to trigger an analyzer capable of time synchronous averaging, the averaged waveform will gradually exclude vibration components from everything except the events related to the pinion revolution. Anyvibration caused by the driven gear will be averaged out, and the resulting waveform will show the vibration caused by each individual tooth on the pinion.

Note that in the figure above, the lower averaged waveform indicates one damaged tooth on the pinion.

Analyzer Set-Up for Synchronous Averaging

Here, the FFT analyzer is used not as a frequency analyzer, but as a time-domain averager. Many, but not all, FFT analyzers have this capability. We will see that the setup parameters forsynchronous averaging are generally quite different from those needed for spectrum analysis. There are two important considerations in setting up a synchronous averaging test, the Time Record Length, and the Time Domain Resolution. The length of the time record is

usually set up to be at least a little longer than one revolution of the of the shaft you are looking at, or one cycle of the event you are studying. There is usually no reason to use a longer time record, since you simply lose resolution in the time domain. The fixed number of samples is spread over a longer time; so short events during the motion you are analyzing could be lost.

Time Record Length

Time record length in an FFT analyzer, as was discussed in the frequency analysis chapter, is the reciprocal of the frequency resolution of the spectrum. In other words, a spectrum withfrequency resolution (or line spacing) of 10 Hz is generated from a time sample lasting 1/10th second. So, suppose in the example in the figure above the pinion is rotating at 1800 RPM and we want to look at one revolution. 1800 RPM is 30 Hz (30 revolutions per second), so one revolution will take 1/30th of a second. To set up an FFT analyzer to collect a 1/30th second time record requires that its frequency resolution, or line spacing, must be 30 Hz. In order to acquire a little more than one revolution, you might use 1/25th second as the time record length. There are many theoretical combinations of frequency range and resolution that could be used to do this - i.e., a span of 5000 Hz with 200 lines resolution, a span of 2500 Hz and 100 lines of resolution, or a span of 250 Hz at 10 lines resolution.

Time Domain Resolution

The resolution in the time domain depends on the sampling rate of the D/A converter in the analyzer - the higher the sampling rate, the greater the resolution. The FFT requires that the sampling rate be 2.56 times the maximum frequency being analyzed. Therefore, a span of 5000 Hz will set the A/D converter to 5000 x 2.56 = 12,800 samples per second. If our time record length is 1/25 second, then the time resolution is 1/25th of 12,800, or 512 samples in the time record. This means we will get 512 snapshots of the gear in one revolution. Since there are 13 teeth on the pinion in question, our 512 samples per revolution will produce for us 512/13 or about 39 snapshots of each tooth. This is more than adequate to examine each tooth in the averaged waveform. On the other hand, had we used 250 Hz at 10 lines of resolution, we would have only 250 x 2.5 = 640 samples per second, and 1/25 second will only contain about 25 samples, or about 2 samples per tooth of the gear, which would probably not be adequate to examine the gear in enough detail. This same basic setup can be used to examine the driven gear rather than the pinion. If it is not possible to put a tachometer on the shaft of the large gear, a frequency divider can be placed in the tachometer output, and if its division ratio is 13:31, its output pulses will coincide with each revolution of the large gear. Then, the time-averaged waveform will show the teeth of the driven gear rather than the pinion. The setup is shown in the figure below:

Since the driven gear is running slower than the pinion, the analyzer must be set up to acquire a time record length corresponding to a little more than 1 revolution of the large gear. If the averaged waveform were also subjected to a frequency analysis with the analyzer, the predominant frequency in the spectrum would be 13X, which is the gear mesh frequency, as in the previous case with the pinion. The technique can be applied to gearboxes with multiple gears as long as the numbers of teeth on each gear are known, and the frequency divider has the capability of performing the proper ratio multiplication.

Case Histories using Synchronous Averaging

The following data are from actual industrial machines whose vibration waveforms were analyzed by synchronous averaging as well as conventional vibration spectrum analysis. In these examples, the averaged waveform presented a much clearer and unequivocal picture of the nature of the machine faults than the vibration spectrum did. However, it must be emphasized that synchronous averaging is not suitable for use in finding modulation effects such as sidebands and other conditions where high resolution in the frequency domain are required.

Low-speed Radar Gearbox

This example is the acceleration spectrum of a large low speed (4 RPM) gear driving a longrange radar antenna. The frequency range is from 0Hz to 6Hz. Note that the spectrum is noisy, and it is difficult to tell it contains anything else but noise.

The next graph shows the synchronous-averaged waveform from which this spectrum came :

Note that the time record length is 90 seconds. There is a sharp peak every 15 seconds. Since the gear speed is 4 RPM, 15 seconds is the time it takes to make one revolution. The peaks are the result of one defective tooth in the gear. Each time the tooth contacts the pinion gear an impulse is generated. The information is immediately assimilated form the waveform, while the spectrum is nowhere near as informative.

Tabletting Press Example

Following is a simplified diagram of the drive mechanism for a large tabletting press. The speed of the large ring gear is 56 RPM. This machine was moved from one building to another, and encountered an accident while en route. It was decided to perform a vibration analysis on it to find any hidden damage.

The following vibration spectra were measured eight days apart. The lower spectrum was the first to be collected. Note the gear mesh frequency at 3640 RPM, with the two prominentsidebands spaced at the pinion RPM from the GMF. This indicated modulation of

the ring gear speed at the pinion rate, and is probably caused by a bent pinion shaft or uneven pinion tooth wear.

It was then decided to collect a synchronous averaged waveform with a time span of about 3 ring gear revolutions. The waveform from the earlier measurement is shown below:

This is not remarkable in that it does not show any significant repetitive pattern. Now, note the upper spectrum in the previous figure, made 8 days later. Here we see sidebands around the GMF at the ring gear turning speed. This indicates the ring gear vibration levelis fluctuating once per revolution. This could indicate a localized fault in the ring gear that causes momentary level changes. Again, a synchronous averaged waveform was collected, and is shown below. The time duration is about 3 seconds, so the plot encompasses 3 revolutions of the ring gear. Note the very definite repeating pattern of a high vibration level. The conclusion was that the ring gear was damaged by the accident, probably by impacting the pinion, and the damage increased rapidly as the machine ran.