Vibration Analysis on Rotating Equipment

The use of vibration analysis in the condition assessment of rotating equipment

Prepared by:

Ron Frend

COPYRIGHT RONALD FREND 2002

1

Vibration Analysis on Rotating Equipment

CONTENTS VIBRATION ANALYSIS - AN INTRODUCTION -------------------------------------------9
Vibration Examples------------------------------------------------------------------------------------------------- 9 Imbalance--------------------------------------------------------------------------------------------------------------- 9 Misalignment ---------------------------------------------------------------------------------------------------------10 Looseness--------------------------------------------------------------------------------------------------------------11 Rolling Element Bearing Defects----------------------------------------------------------------------------------13

VIBRATION THEORY--------------------------------------------------------------------------- 15
Simple Harmonic Motion-------------------------------------------------------------------------------------------15 RMS vs. PEAK-------------------------------------------------------------------------------------------------------18 Time Domain ---------------------------------------------------------------------------------------------------------19 The Frequency Domain ---------------------------------------------------------------------------------------------19 What is an FFT?------------------------------------------------------------------------------------------------------20 The FFT Analyzer ---------------------------------------------------------------------------------------------------20 Advantages of FFT Analyzers -------------------------------------------------------------------------------------21 Frequency Spans -----------------------------------------------------------------------------------------------------21 Measurement Basics -------------------------------------------------------------------------------------------------21 Spectrum---------------------------------------------------------------------------------------------------------------21 Parameter Selection------------------------------------------------------------------------------------------------ 22 Selecting displacement, velocity or acceleration----------------------------------------------------------------22 How does it work? ------------------------------------------------------------------------------------------------- 22 Accelerometers -------------------------------------------------------------------------------------------------------24 Acceleration Amplitude Demodulation ------------------------------------------------------------------------ 25 Theory -----------------------------------------------------------------------------------------------------------------25 The Demodulation Process -----------------------------------------------------------------------------------------26 Resonance Sources---------------------------------------------------------------------------------------------------28 A.C. Motor Example.------------------------------------------------------------------------------------------------29

FAILURE MODES-------------------------------------------------------------------------------- 32
Induction Motors--------------------------------------------------------------------------------------------------- 32 Mechanical or Electrical Effects-----------------------------------------------------------------------------------32 Armature Related Problems ----------------------------------------------------------------------------------------32 Stator Related Problems --------------------------------------------------------------------------------------------33 Broken Rotor Bars ---------------------------------------------------------------------------------------------------34 DC Motors----------------------------------------------------------------------------------------------------------- 34 How DC Power Is Created. ------------------------------------------------------------------------------------34 DC Systems and Controls-------------------------------------------------------------------------------------------36 DC Control Firing Cards--------------------------------------------------------------------------------------------37 S.C.R. problems ------------------------------------------------------------------------------------------------------38 Example of a Firing Card Fault ------------------------------------------------------------------------------------38

2

Vibration Analysis on Rotating Equipment
DC Control Comparitor Card --------------------------------------------------------------------------------------40 Example of a Comparitor Card Defect ---------------------------------------------------------------------------41 Importance of Exact RPM ------------------------------------------------------------------------------------------42 Example of Mechanical -v- Electrical Frequencies ------------------------------------------------------------43 Rotating Equipment ----------------------------------------------------------------------------------------------- 45 Imbalance--------------------------------------------------------------------------------------------------------------45 Vibration due to imbalance-----------------------------------------------------------------------------------------46 Misalignment ---------------------------------------------------------------------------------------------------------51 Looseness--------------------------------------------------------------------------------------------------------------54 Vibration due to aerodynamic forces--------------------------------------------------------------------------- 59 Aerodynamic cross coupling ---------------------------------------------------------------------------------------60 Surging-----------------------------------------------------------------------------------------------------------------60 Choking or Stone Walling ------------------------------------------------------------------------------------------61 Bearing Failures ---------------------------------------------------------------------------------------------------- 62 Elasto Hydrodynamic Lubrication --------------------------------------------------------------------------------62 First Stage of Bearing Failure--------------------------------------------------------------------------------------63 Second Stage of Bearing Failure ----------------------------------------------------------------------------------64 Third Stage of Bearing Failure-------------------------------------------------------------------------------------65 Fourth Stage of Bearing Failure -----------------------------------------------------------------------------------66 Bearing Defect Frequency Calculation ---------------------------------------------------------------------------67 Analysis of bearing defects -----------------------------------------------------------------------------------------70 Balancing ------------------------------------------------------------------------------------------------------------ 75 In-place Balancing ---------------------------------------------------------------------------------------------------75 Vibration Related to Imbalance------------------------------------------------------------------------------------78 How to Balance - Single Plane-------------------------------------------------------------------------------------79 Single Plane Vector Method of Balancing-----------------------------------------------------------------------80 Four-step Method of Balancing Single Plane -------------------------------------------------------------------82 Balancing in One Run -----------------------------------------------------------------------------------------------85

SINGLE CHANNEL ANALYSIS -------------------------------------------------------------- 87
Taking measurements --------------------------------------------------------------------------------------------- 87

POTENTIAL FAILURE ANALYSIS ---------------------------------------------------------- 91
A methodology for objective set up ----------------------------------------------------------------------------- 91 Introduction --------------------------------------------------------------------------------------------------------- 91 The PFA Tree ------------------------------------------------------------------------------------------------------- 91 Base cause-------------------------------------------------------------------------------------------------------------91 Failure type------------------------------------------------------------------------------------------------------------91 External manifestation-----------------------------------------------------------------------------------------------92 Technology------------------------------------------------------------------------------------------------------------92 Parameter --------------------------------------------------------------------------------------------------------------92 Analysis----------------------------------------------------------------------------------------------------------------92 Interval-----------------------------------------------------------------------------------------------------------------92 Setup -------------------------------------------------------------------------------------------------------------------92

3

114 What is resonance? ---------------------------------------------------------------------------------------------.Vibration Analysis on Rotating Equipment Developing a Potential Failure Analysis for Rolling Element Bearings --------------------------------.105 Two Channel Analysis------------------------------------------------------------------------------------------.106 Advanced functions ---------------------------------------------------------------------------------------------.115 Liquids & pumping systems-------------------------------------------------------------------------------------.123 IDENTIFYING RESONANCE IN MECHANICAL SYSTEMS ----------------------.106 Two channel functions -------------------------------------------------------------------------------------------.95 Conclusion ----------------------------------------------------------------------------------------------------------.102 Analysis -----------------------------------------------------------------------------------------------------------.109 Triggering ----------------------------------------------------------------------------------------------------------.118 Air & gases---------------------------------------------------------------------------------------------------------.115 Natural Frequency ----------------------------------------------------------------------------------------------.130 Impact hammer ----------------------------------------------------------------------------------------------------.108 Representation by complex numbers --------------------------------------------------------------------------.122 Critical Speed (Balance Resonance) -------------------------------------------------------------------------.108 Cascade & waterfall plots----------------------------------------------------------------------------------------.96 Measurement Windows ---------------------------------------------------------------------------------------------99 Averaging ----------------------------------------------------------------------------------------------------------.109 Bodé plots ----------------------------------------------------------------------------------------------------------.115 Mechanical ---------------------------------------------------------------------------------------------------------.129 Running machine Bump Test -----------------------------------------------------------------------------------. --------------------------------------------------------.126 Mode Shape ------------------------------------------------------------------------------------------------------.100 Real Time Bandwidth and Overlap Processing --------------------------------------------------------------. --------------------------------------------------------------.93 Stage 1 -----------------------------------------------------------------------------------------------------------------93 Stage 2 -----------------------------------------------------------------------------------------------------------------94 Stage 3 -----------------------------------------------------------------------------------------------------------------94 Stage 4 -----------------------------------------------------------------------------------------------------------------94 Including the Component Failure in the PFA Tree.130 Set up for FFT-type analyzers --------------------------------------------------------------------------------.110 INTRODUCTION TO RESONANCE -----------------------------------------------------.101 Octave Analysis ---------------------------------------------------------------------------------------------------.127 The bump test ----------------------------------------------------------------------------------------------------.131 4 .120 Karman Vortices --------------------------------------------------------------------------------------------------.130 Reverse Bump or “Plucking” the Suspect Part.110 Orbits----------------------------------------------------------------------------------------------------------------.126 Phase---------------------------------------------------------------------------------------------------------------.103 Severity charts -----------------------------------------------------------------------------------------------------.

135 Plates---------------------------------------------------------------------------------------------------------------.139 Springs --------------------------------------------------------------------------------------------------------------.145 Damping ------------------------------------------------------------------------------------------------------------.147 Sources of vibration---------------------------------------------------------------------------------------------.137 DETUNING RESONANT STRUCTURES-----------------------------------------------.147 Frequency ranges -----------------------------------------------------------------------------------------------.145 Changing the Mass------------------------------------------------------------------------------------------------.146 WHOLE BODY VIBRATION ---------------------------------------------------------------.142 Modifying the structure----------------------------------------------------------------------------------------.147 5 .139 Vibration isolators ----------------------------------------------------------------------------------------------.146 Changing the stiffness --------------------------------------------------------------------------------------------.140 Rubber --------------------------------------------------------------------------------------------------------------.135 Uniform Beams --------------------------------------------------------------------------------------------------.Vibration Analysis on Rotating Equipment CALCULATING NATURAL FREQUENCY IN MECHANICAL SYSTEMS-----.

bearing defect _______________________________________ 29 Figure 22 High frequency waterfall ______________________________________________________ 30 Figure 23 Enveloped acceleration spectrum________________________________________________ 30 Figure 24 Comparison .Vibration Analysis on Rotating Equipment List of illustrations Figure 1 Velocity spectrum showing imbalance ______________________________________________ 9 Figure 2 .inner race defect _________________________ 14 Figure 8 Inner race spall_______________________________________________________________ 14 Figure 9 Simple Harmonic Vibration _____________________________________________________ 16 Figure 10 Integration from acceleration to velocity __________________________________________ 17 Figure 11 Integrating to displacement ____________________________________________________ 18 Figure 12 Peak -v.full wave rectified _______________________________________ 43 Figure 37 Same motor .speed lowered by 25% _____________________________________________ 44 Figure 38 Imbalance slide 1 ____________________________________________________________ 46 Figure 39 Imbalance slide 2 ____________________________________________________________ 46 Figure 40 Imbalance slide 3 ____________________________________________________________ 47 Figure 41 Imbalance slide 4 ____________________________________________________________ 47 Figure 42 Imbalance slide 5 ____________________________________________________________ 48 Figure 43 Imbalance slide 6 ____________________________________________________________ 48 Figure 44 Imbalance slide 7 ____________________________________________________________ 49 Figure 45 Imbalance slide 8 ____________________________________________________________ 49 Figure 46 Imbalance slide 9 ____________________________________________________________ 50 Figure 47 Misalignment slide 1__________________________________________________________ 51 Figure 48 Misalignment slide 2__________________________________________________________ 51 Figure 49 Misalignment slide 3__________________________________________________________ 52 Figure 50 Misalignment slide 4__________________________________________________________ 52 Figure 51 Misalignment slide 5__________________________________________________________ 53 Figure 52 Looseness slide 1 ____________________________________________________________ 54 Figure 53 Looseness slide 2 ____________________________________________________________ 54 6 .RMS _______________________________________________________________ 19 Figure 13 Compression mode accelerometer _______________________________________________ 24 Figure 14 Shear mode accelerometer _____________________________________________________ 24 Figure 15 Simple modulation example ____________________________________________________ 25 Figure 16 Bearing modulation example ___________________________________________________ 26 Figure 17 Demodulation process ________________________________________________________ 27 Figure 18 Enveloping process___________________________________________________________ 27 Figure 19 Fast Fourier Transform _______________________________________________________ 28 Figure 20 FFT ._____________________________________ 41 Figure 34 FFT after the comparitor card was replaced _______________________________________ 42 Figure 35 DC motor components ________________________________________________________ 42 Figure 36 FFT from a 5 HP motor .3D view_______________________________________________________________ 28 Figure 21 Two channel time waveform .radial ________________________________________ 11 Figure 4 Velocity spectrum of misaligned fan .axial _________________________________________ 11 Figure 5 Velocity spectrum from a loose fan drive motor______________________________________ 12 Figure 6 Envelope spectrum of a fan drive motor with loose bearing ____________________________ 13 Figure 7 Enveloped acceleration spectrum of bearing .Velocity spectrum showing fan imbalance _________________________________________ 10 Figure 3 Velocity spectrum of misaligned fan .velocity to envelope _______________________________________________ 31 Figure 25 The creation of DC power _____________________________________________________ 35 Figure 26 FFT spectrum of half wave rectification___________________________________________ 36 Figure 27 FFT spectrum of full wave rectification ___________________________________________ 36 Figure 28 Basic DC system circuit _______________________________________________________ 37 Figure 29 FFT spectrum full wave DC firing card frequencies _________________________________ 38 Figure 30 FFT spectrum after repair _____________________________________________________ 38 Figure 31 FFT spectrum showing half wave firing card frequencies _____________________________ 39 Figure 32 FFT spectrum of same motor (no load) ___________________________________________ 40 Figure 33 FFT spectrum showing comparitor card defect.

Roller Bearing Courtesy of the Torrington Company _________________ 64 Figure 65 Waterfall plot from a damaged motor bearing ______________________________________ 65 Figure 66 Early Fatigue .Vibration Analysis on Rotating Equipment Figure 54 Looseness slide 3 ____________________________________________________________ 55 Figure 55 Looseness slide 4 ____________________________________________________________ 55 Figure 56 Looseness slide 5 ____________________________________________________________ 56 Figure 57 Looseness slide 6 ____________________________________________________________ 56 Figure 58 Looseness slide 7 ____________________________________________________________ 57 Figure 59 Looseness slide 8 ____________________________________________________________ 57 Figure 60 Looseness slide 9 ____________________________________________________________ 58 Figure 61 Aerodynamic forces __________________________________________________________ 60 Figure 62 Elasto-hydrodynamic lubrication ________________________________________________ 62 Figure 63 Loss of Lubricant .Ball Bearing Inner Race Courtesy of the Barden Corporation__________ 63 Figure 64 Loss of Lubricant . _______________________________________________ 120 Figure 103 Sonic vibration in a tube_____________________________________________________ 122 Figure 104 Karman vortices ___________________________________________________________ 123 Figure 107 Campbell Diagram _________________________________________________________ 125 Figure 109 Mode shape readings _______________________________________________________ 127 Figure 110 Phase relationships_________________________________________________________ 128 Figure 111 Phase / frequency relationships @ resonance ____________________________________ 129 Figure 112 Impact hammer response ____________________________________________________ 132 Figure 113 Impact hammer specification sheet_____________________________________________ 132 7 .Ball Bearing Courtesy of the Barden Corporation ______________________ 66 Figure 67 Developed Fatigue on Roller Bearing Courtesy of the Torrington Company ______________ 66 Figure 68 Ball Bearing Terminology _____________________________________________________ 68 Figure 69 Waterfall of early damage to a motor bearing collected every 1.___________________________________________ 72 Figure 75 Time waveform from a heavily marked bearing _____________________________________ 73 Figure 76 Velocity spectrum from a spalled bearing _________________________________________ 73 Figure 77 Sources of imbalance _________________________________________________________ 75 Figure 78 Assembly tolerance stack up ____________________________________________________ 76 Figure 79 Heavy spot _________________________________________________________________ 76 Figure 80 Units of measure of imbalance __________________________________________________ 77 Figure 81 Mass centre displacement______________________________________________________ 77 Figure 82 Force due to imbalance _______________________________________________________ 78 Figure 83 The vector diagram___________________________________________________________ 80 Figure 84 Simplified vector diagram______________________________________________________ 81 Figure 85 Additional corrections ________________________________________________________ 82 Figure 86 Direction to shift the weight ____________________________________________________ 83 Figure 87 Sample problem vector diagram_________________________________________________ 84 Figure 88 Determining the flash angle ____________________________________________________ 86 Figure 89 Typical tap block for mounting an accelerometer ___________________________________ 87 Figure 90 Accelerometer mounting techniques a-d___________________________________________ 89 Figure 91 Accelerometer mounting techniques e-g___________________________________________ 90 Figure 92 Overview of accelerometer mounting techniques ____________________________________ 90 Figure 93 PFA development for rolling element bearings _____________________________________ 93 Figure 94 PFA for a main motor_________________________________________________________ 95 Figure 95 General severity chart for vibration _____________________________________________ 105 Figure 96 Vector addition of 2 vibrations _________________________________________________ 108 Figure 97 Cascade of fan over 20mS ____________________________________________________ 109 Figure 98 Bode plots _________________________________________________________________ 110 Figure 99 Orbit showing misalignment___________________________________________________ 111 Figure 102 Harmonic series for the tone C.5 hrs over 14 days _________ 69 Figure 70 Bearing damage severity assessment chart ________________________________________ 70 Figure 71 Demodulated acceleration spectrum from a dry bearing ______________________________ 70 Figure 72 Demodulated acceleration spectrum of a marked bearing_____________________________ 71 Figure 73 Demodulated acceleration spectrum from a slightly more heavily marked bearing _________ 72 Figure 74 Time waveform from a marked bearing.

for the speed of sound in a gas _______________________________ 121 Equation 10 Ratio of specific heats (gamma) ______________________________________________ 121 Equation 11 Speed of sound in a gas ____________________________________________________ 121 Equation 12 Karman vortices __________________________________________________________ 123 Equation 14 Amplitude magnification due to springs ________________________________________ 141 Equation 15 Natural frequency of a spring ________________________________________________ 142 Equation 16 Modulus of elasticity for rubber ______________________________________________ 144 8 . wavelength & frequency ___________________________ 119 Equation 9 Newton-Laplace eq.Vibration Analysis on Rotating Equipment Figure 114 Stress/strain diagram for steel ________________________________________________ 140 Figure 115 Stress/strain diagram for rubber ______________________________________________ 140 Figure 116 Resonance Curve __________________________________________________________ 141 List of Tables Table 1 Speed of sound in liquids _______________________________________________________ 118 Table 3 Natural frequency calculation of uniform beams_____________________________________ 135 Table 4 Standard values for uniform beams _______________________________________________ 136 Table 5 Damping ranges of vibration isolators_____________________________________________ 139 Table 6 Whole body vibration (frequency ranges) __________________________________________ 148 List of Equations Equation 3 Newton's 2nd law __________________________________________________________ 117 Equation 5 Differential equation of motion of a single-degree-of-freedom system _________________ 118 Equation 6 Velocity of sound in materials_________________________________________________ 119 Equation 7 Speed of sound in the ocean __________________________________________________ 119 Equation 8 General formula relating speed.

Vibration Analysis on Rotating Equipment Vibration Analysis . We will carry on that noble tradition and look at some spectra that have been collected from real machines and show typical examples of common defects. So we can see that vibration analysis itself is not new but some of the ways that we take the measurements and apply those measurements as machine health diagnoses are very new. In 1638 Galileo described the vibrations of pendulums. the phenomenon of resonance and the factors influencing the vibration of strings.C. In this section we will briefly take a look at some vibration examples of typical defects suffered by fans and fan drives without delving too deeply into why! Vibration Examples When the novice analyst first carries out vibration analysis he will usually rush out and take a vibration spectrum using the default parameters set up in the analyzer.an introduction The study of noise and vibration phenomena dates back centuries. This spectrum is of velocity vibration so the amplitude units could be in 9 . The fan was driven from the AC motor via a V-belt and rotated at about 720 rpm. The AC drive motor rotates at just under 1200 rpm. The spectrum is a simply a graph of the vibration frequency on the bottom axis with the amplitude at that frequency on the vertical axis. The first recorded incidence of such study was by Pythagoras in the sixth century B. who studied the origin of musical sounds and the vibration of strings. Euler in 1744 and Bernoulli in 1751 developed the equation for the vibrations of beams and developed the normal modes for various boundary conditions. Imbalance Figure 1 Velocity spectrum showing imbalance Figure 1 shows a vibration spectrum that was taken at the sheave end of a centrifugal fan in the vertical direction. In 1882 Hertz developed the first successful theory for impact.

Notice in the spectrum that there is one big spike which is labeled at 716. 10 .Vibration Analysis on Rotating Equipment mm/s or ips (inches/second).59 cpm and there are two much smaller spikes just to the right. Figure 2 . Figure 1 amplitude was displayed with metric units and figure 2 with inch units but the shape of the spectrum is the same in both cases.Velocity spectrum showing fan imbalance Figure 2 shows a similar problem on a different fan but we see that the spectrum looks very similar with one dominant spike at the run speed of the fan. Misalignment Probably 40% of all bearing and shaft failures are caused by misalignment of the components creating an extra axial thrust on the bearings. The frequency is in cpm (cycles/minute) but it could have been displayed in Hz (Hertz or cycles/second) or in orders (multiples of run speed). The first spike to the right is at 1187 cpm which equates to the run speed of the motor and the second spike is at 1433.18 cpm which is exactly twice fan speed. The spectrum was recorded from a vibration transducer which was mounted radial to the shaft (vertical in this case). A check with a stroboscope confirmed that the fan was actually running at 717 rpm so the big spike of vibration is at exactly (within the precision of the strobe) run speed. At this stage we are not concerned about the physics of why a vibration at run speed is usually indicative of imbalance but we will look at our spectral explanation charts (see appendix 1) and have a fair degree of confidence that the fan needs balancing. Because the one spike is so dominant that is the one that we are concerned about.

in figure 4 we are now looking at the vibration taken axial to the shaft. although the amplitudes are relatively low. on a belt driven train. 11 . Again. a high axial velocity vibration relative to the radial vibration is almost always indicative of component misalignment to the belt.Vibration Analysis on Rotating Equipment Figure 3 Velocity spectrum of misaligned fan . Looseness Looseness exists when the component is not directly attached to the structure or rotating element and has a relatively large clearance. Figure 4 Velocity spectrum of misaligned fan . allowing the component to rattle. If the problem was simple imbalance of the fan we would expect all of the forces to be caused by centrifugal force and therefore acting in a direction which was radial to the shaft.radial In figure 3 we see what initially looks like an imbalance condition of the fan.axial However. looking at our spectrum explanation charts we see that.

The motor was running at 590 rpm and immediately we see the large family of harmonics of run speed. Whenever we see multiples or sub-multiples of run speed vibration frequencies we immediately consider the possibility of loose components. In this case the structure of the bedplate was cracked causing parts of the structure to vibrate freely at the excitation frequency of the motor (speed). 12 . The amplitudes do not seem too high but the machine was massive and any vibratory forces have to move the mass before we see a vibration.Vibration Analysis on Rotating Equipment Figure 5 Velocity spectrum from a loose fan drive motor The above spectrum was recorded at the sheave end of the drive motor of an underground colliery main ventilation fan.

13 . As the looseness deteriorates the envelope readings will decrease but then the velocity readings will start to increase. We will study bearing defects in detail later but notice that the main defects are not multiples of run speed. Notice that the spike at about 8. Rolling Element Bearing Defects The primary tool in assessing bearing condition is the use of enveloped acceleration readings.Vibration Analysis on Rotating Equipment Figure 6 Envelope spectrum of a fan drive motor with loose bearing The early stages of looseness can be detected in a similar manner. In other words they are non-synchronous. and using demodulated or enveloped acceleration readings. Figure 7 shows the envelope spectrum from a bearing with a severe spall in the inner race. be looking for harmonics of run speed.772 cpm is marked BPIR which stands for Ball Pass Outer Race. Figure 6 shows the early stages of looseness of a bearing inside the fan drive motor.

Vibration Analysis on Rotating Equipment Figure 7 Enveloped acceleration spectrum of bearing .inner race defect Figure 8 Inner race spall 14 .

The transducer converts the movement into an electrical signal which is passed to the spectrum analyzer which in turn converts that signal from the time domain into the frequency domain. Simple Harmonic Motion Simple harmonic motion can be visualized by many common examples such as a pendulum. its speed is 100 revolutions per minute or 100 RPM. gear meshing forces. and many other factors. 15 .001 inch and one micron (µm) is 0. In vibration work. rotating electrical fields. a mass and spring combination. then the frequency is 100 cycles per minute.001 mm. The time domain waveform is composed of a machine’s response to many individual forces such as imbalance. The purpose of this section is to describe the relationship between each of these and how they are used on rotating machinery. velocity and displacement. However for predictive maintenance techniques where rotational speed is often the key to vibration peaks. like displacement. it is important to realize that the motion of the measured surface varies with time. a rotating mechanism or a diving board. it has a velocity associated with each position and. The frequency of vibration is often expressed in terms of cycles per second or HERTZ after the German physicist Heinrich Hertz.Vibration Analysis on Rotating Equipment Vibration Theory The following section is meant as a primer to help the newcomer to vibration analysis understand some of the terms used and to develop an understanding of the concepts. in the frequency domain it is much easier to separate these elements to determine the importance of each.Figure 9 uses a pendulum. One Mil is equal to 0. In figure 9 the points B and C represent the extreme position of the pendulum and the distance between them is the peak to peak displacementAmplitude meters are often calibrated to give the peak to peak value because it is the displacement extremes that are of interest. misalignment. cycles per minute are used in preference to Hertz. If the pendulum swings back and forth 100 times in one minute. Since the pendulum is continuously moving. the displacement is often expressed in terms of mils or micron. acceleration. When viewing the time domain data it can be quite difficult to separate these components of vibration. Similarly if a machine is rotating 100 times in one minute. Vibration amplitude is measured using three different parameters. In addition to frequency the amplitude is the other necessary quantity that must be known in order to characterize vibration. To understand the concept of vibration analysis. the velocity also varies between a positive peak and a negative peak. However.

called the frequency. The 90 degree phase lead is shown in the diagram on figure 10.. The rate of change of displacement is the velocity. Acceleration Vibration is described by its frequency and amplitude. velocity or acceleration. then to the left.. first to the right. Velocity amplitude is expressed only in terms of zero to peak or zero to RMS. or Hertz. Figure 9 Simple Harmonic Vibration Figure 10 shows that at position B and C. Displacement . The amplitude is expressed in units of either displacement. and at position A the velocity is maximized. one cycle C B A An oscillating system will produce a certain number of cycles per unit time. Frequency is usually expressed in terms of cycles per second. the velocity is zero. Velocity .. therefore if D is expressed in terms of inches. instead of the usual mils. velocity is said to lead displacement by 90°.. not magnitude. 16 .. The negative peak velocity differs only in direction. then the product 2πfD will be the velocity in inches per second which are the units used for velocity in vibration work.Vibration Analysis on Rotating Equipment (48) * FREQUENCY * AMPLITUDE .. Since the peak positive velocity occurs 1/4 cycle before the peak positive displacement.

Figure 11 shows that at position B and C the acceleration is maximum. Velocity is highest where displacement is zero and is zero where displacement is maximum. Therefore a 90 phase shift exists between displacement and velocity. is maximum. The maximum acceleration to the right occurs 1/2 a cycle before the maximum velocity to the right and acceleration is said to lead displacement by 180°. The velocity amplitude is directly proportional to frequency for a given displacement. The unit of acceleration is the “g” which is equal to 9. The diagram in figure 11 shows these phase leads and also the acceleration amplitude relationship. A = (2πf)2D. this acceleration is also associated with the motion. the acceleration. Acceleration is the third way to express vibration amplitude. Therefore the rate of change of velocity (the acceleration) must be zero at A. The amplitude is one half of the peak-to-peak value for a sine wave. Note that acceleration reaches its maximum at Points B and C just as displacement does. but at B acceleration is to the left whereas the displacement is to the right. o Figure 10 Integration from acceleration to velocity As velocity is continuously changing. A B A C A C B A Highest Velocity Figure 4. The distance between the extremes of motion is the peak-to-peak displacement. Displacement meters are often calibrated in peak-to-peak units. velocity is to the right and just after it is to the left. This says that for any given value of displacement. the acceleration is proportional to the square of the frequency.Vibration Analysis on Rotating Equipment (49) A B A C A Disp Peak Peak to Peak C B A Figure 3. Conversely just prior to point A velocity is increasing and just after. Acceleration leads velocity by 90°. Vel Disp 17 . Just prior to point B. an acceleration is also associated with the velocity.81 m/sec2 and is derived from the acceleration due to earth’s gravity. it is decreasing. At B therefore the rate of change of velocity.

Velocity is a quantitative measure of the effective velocity and reflects the power or energy being used to vibrate the machine mass. like vibration signals. At B. RMS and Peak amplitude values for sine waves are not valid for combination or random waveforms. squaring the amplitude value of each point. is an attempt to apply a single quantitative value -which reflects the effective value of this varying function. When using FFT analyzers care should be taken when evaluating peak or rms severity as the peak amplitude in the spectrum is derived from a sine wave.Vibration Analysis on Rotating Equipment (50) A B A C A C B A Acceleration Acceleration Figure 5. During most of the cycle the value of the instantaneous voltage is somewhere between zero and peak.707 x Peak Figure 6. To measure only the peak value may be misleading since the voltage is actually at a peak for only a small portion of the cycle. True peak can be seen in the time waveform. acceleration is maximum to the left and displacement o maximum to the right. Using rms values can be compared to the use of rms in electrical circles i. then.637 x Peak RMS = 0. a 180 phase shift. The simple relationships which exist between average. stereo speakers power values are measured in rms values. Figure 11 Integrating to displacement RMS vs. This same logic applies to vibration. calculating the mean of the squared values and then finding the square root of the mean. Acceleration amplitude varies as the square of frequency for a given value of displacement. PEAK The rms or root mean square value is calculated by breaking the waveform down into a number of points. ranging from zero to a peak value. This. RMS. Peak value is the maximum amplitude seen during the measurement. 18 Accel Vel Disp . Electrical (AC) voltage is also measured in rms. is a continuously varying quantity. Peak Avg Amplitude RMS Peak to Time Peak Avg = 0.e.

The method most analyzers use to transform signals from the time domain to the frequency domain is called :- 19 .Vibration Analysis on Rotating Equipment (51) Peak Avg Amplitude RMS Peak to Time Peak Simple Sine Wave Peak Amplitude Peak to Peak RMS Time Complex Waveform Figure 12 Peak -v. The time domain is a record of what happened to a parameter compared to time. Its height represents its amplitude. analysis of signals in the time domain can be very difficult and is far easier in the frequency domain The Frequency Domain If we now convert a time waveform to the frequency domain we will get a totally different picture. With respect to machinery vibration.Typically the signal would be displayed on an oscilloscope. We now have axes of amplitude v frequency instead of amplitude -v.time. its position represents its frequency. Every sine wave separated out by the FFT appears as a separate line.RMS Time Domain The traditional way of observing signals is to view them in what is called the time domain.

56 No samples / 2. like you would see on an oscilloscope trace. Fourier's theorem states that any waveform in the time domain can be represented by the weighted sum of sines and cosines. which attenuates all frequency components above Fmax by 90 dB to make sure that Nyquist's theorem is satisfied. and compute its frequency spectrum. was measured to determine the signal strength in certain frequency bands. We cannot transform to the frequency domain in a continuous manner.200. The FFT spectrum analyzer samples the input signal. We therefore must sample and digitize the time domain input. Most analyzers have an FFT processor. The FFT Analyzer An FFT spectrum analyzer works in an entirely different way. Most analyzers offer resolutions of 100. The original digital time record comes from discrete samples taken at the sampling rate. The corresponding FFT yields a spectrum with discrete frequency samples.5 Hz (the same as the lowest measurable frequency) 20 .3200 or even 6400 Lines. an analog filter was used to isolate frequencies of interest. computes the magnitude of its sine and cosine components. Many of these measurements were once done using analog spectrum analyzers. the spectrum has less than half as many frequency points as there are time points (remember Nyquist's theorem).56 = = = Fmax / No.800. In simple terms. then the sampled data will accurately represent the input signal. Certain analyzers pass the input signal through an analog filter. The number of samples determines the resolution (number of lines) of frequency. In fact.56) 2. FFT Spectrum Analyzers take a time varying input signal. The signal power.Vibration Analysis on Rotating Equipment FFT (Fast Fourier Transform) What is an FFT? The fast Fourier transform (FFT) is an algorithm for transforming data in the time domain to the frequency domain. The FFT of this record yields 400 frequency points or lines. of lines (2560 / 2. This is the anti-aliasing filter.4 Seconds to take this time record. The lowest frequency is just the Fmax divided by the number of lines: Fmax No. which performs this transformation automatically and then stores the computed spectra into memory. Suppose that you take 1024 samples at 2560 Hz. It takes 0.2. which passed through the filter. Now here's the interesting part. The input signal is digitized at a high sampling rate.400.56 x Fmax usually). The resulting spectrum shows the frequency components of the input signal. a spectrum could be obtained.1600.56. Nyquist's theorem says that as long as the sampling rate is greater than twice the highest frequency component of the signal. By tuning the filters and repeating the measurements. The FFT is simply a clever set of operations which implements Fourier's theorem.56) / (1024 / 2. Of Lines = Bin resolution = data sampling rate / 2. but over what frequency range? The highest frequency will be determined by the in-built ratio of Fmax to data sampling rate . The resulting digital time record is then mathematically transformed into a frequency spectrum using an algorithm known as the Fast Fourier Transform or FFT. (2. and displays the spectrum of these measured frequency components.

5 Hz.1000 Hz.000 values.5 Hz. Spectrum The spectrum is the basic measurement of an FFT analyzer. The magnitude is a real quantity and represents the total signal amplitude in each frequency bin. 21 . For vibration measurements phase is usually considered to be relative to the trigger pulse. (0. all frequencies are examined simultaneously. So our final displayed spectrum contains 400 frequency bins. (Alternately. If we use a synchronous trigger then each frequency component has a well-defined phase.56 kHz is therefore not useable and the actual displayed frequency span stops at 1000 Hz. Measurement Basics An FFT spectrum is a complex quantity.5 . This bin actually covers the range from 0 Hz to 2. providing an enormous speed advantage. There is also a frequency bin labeled 0 Hz (or DC).048. If we keep increasing the time record. then the real (cosine) or imaginary (sine) part or the phase may be displayed. the magnitude of the spectrum is displayed.e.2. The range between 1000 Hz and 2. Normally. Because FFT spectrum analyzers measure all frequency components at the same time the technique offers the possibility of being hundreds of times faster than traditional analog spectrum analyzers. the second 2. Frequency Spans Before we continue. let's clarify a couple of points about our frequency span. In the case of a 1000 Hz span and 400 resolvable frequency bins. The first covers 0 .5 Hz (for this example) is considered to be DC. A typical filter has a flat response from DC to 1000 Hz and then rolls off steeply from 1000 Hz to 2.5 .) If there is no triggering. If there is phase information in the spectrum. the time record is triggered in phase with some component of the signal. and the 400th covers 997.56 kHz. It is simply the complex FFT. independent of phase. then we would need to store more and more points. Advantages of FFT Analyzers The advantage of this technique is its speed. we ought to get 2048 time points (sampling at 2560 Hz) yielding a spectrum from DC to 1000 Hz with 1. the entire spectrum takes only 400 mS to measure. then the phase is random and we generally look at the magnitude of the spectrum. The output spectrum thus represents the frequency range from DC to 1000 Hz with points every 2.5 Hz (the lowest measurable frequency) and contains the signal components whose period is longer than the time record (not only DC). We just described how we arrived at a DC to 1000 Hz frequency span using a 400 mS time record. What happens if we make the time record 800 mS or twice as long? Well. This is because each frequency component has a phase relative to the start of the time record.5 Hz. One is memory.25 Hz resolution containing 800 points.00125 Hz resolution would require 2. The length of the time record determines the frequency span and resolution of our spectrum. (Square root of the sum of the real (sine) part squared and the imaginary (cosine) part squared).) Another limitation is processing time. the entire frequency span is not useable. The magnitude is the square root of the FFT times its complex conjugate. The more points you take. The phase is simply the arc tangent of the ratio of the imaginary and real parts of each frequency component. the longer the processing time. you may wish to think of the input signal being composed of sines and cosines. but again. To measure the signal with higher resolution the time record is increased.Vibration Analysis on Rotating Equipment Everything below 2. i. But the analyzer places some limitations on this. Because the signal passes through an anti-aliasing filter at the input.

at 60 Hz this is rough but at 200 Hz this is very rough and should not be permitted for machines up to the 100HP class. The transducer sees the force once per rev. velocity or acceleration As previously discussed. Therefore when motions are being measured a displacement measurement is in order because frequencies of interest on the shaft are limited to 20 or so orders of rotation. An accelerometer will give an output which is proportional to the acceleration of the transducer in G’s or inches/second/second or meters/second/second. Newton’s second law (F=ma) tells us that the acceleration of a body is directly proportional to the force applied to the body.l mils is observed. displacement amplitude is higher at lower frequencies. Now suppose a velocity of 0. As an example suppose a displacement of 0. For higher frequencies however. Velocity transducers will give an output proportional to the linear speed (velocity) of the transducer in inches/second or millimeters/second.5 in/sec is 0. as near as possible to the bearing and couple the transducer to a spectrum analyzer. In summary. The most prominent advantage of a velocity measurement is that the value of rms velocity is related to the potential for mechanical damage. Is this severe? The answer is “Yes. significant vibration has a displacement value which -is too small to conveniently measure and velocity or acceleration is more appropriate. Velocity measurements are especially good for a number of reasons. How does it work? Consider a rotating machine (a motor) which has. At that frequency 0. for example. this is severe regardless of the frequency”. of the rotor as a “simple harmonic motion”.6 in/sec (15mm/s) is observed. usually the bearing. The output of the transducer will depend on what type of transducer we are using. regardless of the frequency. For predictive maintenance purposes we use accelerometers almost exclusively so we will concentrate on them for now. In applying this to rotating machinery displacement measurements are relative readings of the displacement of the shaft to a reference. We place our transducer on the drive end of the motor in the vertical direction. The many published vibration severity charts are based on this principle. velocity up to 1 kHz and acceleration from 2 Hz to 20 kHz depending on the design of the accelerometer.Vibration Analysis on Rotating Equipment Parameter Selection Selecting displacement. to centrifugal) force. very small but certainly a measurable value. an out of balance condition on the rotor so that for every revolution of that rotor the out of balance mass generates a centripetal (opp. According to Newton’s Second Law F=m⋅x a ⋅ 22 . 20 orders is a frequency of 1200 Hz. For a 3600 rpm machine. Velocity and acceleration measurements are usually made on the bearing cap or on the machine casing in way of a structural web to enhance the transmission of vibration to the pick-up point. That is to say that the machine surface will cause the transducer to move in a downwards direction with the machine as the force itself is acting downwards and will cause the transducer to move upwards when the machine is moving up etc.13 mils pk-pk. In other words the acceleration vibration gives a good indication of impactive forces inside the machine such as bad bearings. • • • Displacement transducers will give an output proportional to the linear displacement of the transducer in thousandths of a inch or micron. is this severe? At 6 Hz this is not severe at all. displacement measurements are good from 0 Hz to 500 Hz.

Vibration Analysis on Rotating Equipment where F = the force m = the mass a = the acceleration So immediately we see that the output from the accelerometer is directly proportional to the internal forces acting on the machine. Either way. For example. then we may safely say that the acceleration is also proportional to the square of the speed. This is the signal that is passed along to the analyzer. a=ϖ2⋅r ϖ The important point here is that the faster the machine goes. the out of balance condition will show itself in the frequency domain as a single spike at a frequency which corresponds to the run speed of the machine. the acceleration amplitudes must increase for a given force even if there is nothing wrong with our machine. We have already said that we have a concern that the acceleration increases with frequency. The analyzer will then convert this time domain signal into a frequency domain signal either as acceleration or as an integration from acceleration into velocity.200 rev/min the spike will have a frequency at 1. we know that acceleration is simply the rate of change of velocity.200 cycles/minute (cpm) or 20 Hertz (Hz). Integrating acceleration will change our value from: inches/second2 to inches/second effectively finding the square root of the acceleration (for time). If we plot the acceleration against time (time domain) we would see a sine wave which is the result of simple harmonic motion. As we know that the acceleration is proportional to the force and we assume that the mass and radius of force of the machine stay constant. Newton also says that for a rotating body F=m⋅ϖ2⋅r ⋅ where ϖ r = the rotational speed in radians/second = the radius at which the force is acting. so if we need a value that is independent of frequency for severity analysis purposes we can use the velocity reading. if the motor is running at 1. So if we integrate our acceleration reading with respect to time we will get a velocity reading. However. or the higher up the frequency range we go. Back to our motor. 23 .

By far the best way to mount an accelerometer is to spot face the subject surface and drill and tap it to accept the stud for the accelerometer. In other words the shear mode accelerometer is not as affected by thermal transients and gives better accuracy for the axis in which it is mounted. hence. the response from the accelerometer is better if it is permanently mounted. Modern accelerometers are available as compression mode or shear mode.Vibration Analysis on Rotating Equipment Accelerometers Figure 13 Compression mode accelerometer Looking at the figure above we see a schematic of an accelerometer. on a motor it is usually not practical to drill into the motor frame for obvious reasons. However. Permanently mounting an accelerometer should be done with care. Figure 14 Shear mode accelerometer Many low cost industrial accelerometers are now shear mode. the useable frequency range. For off-line measurements the accelerometer will probably be connected to a magnet and the magnet positioned at a pre-determined point every time a reading is taken. Generally speaking the shear mode accelerometer offers better axial sensitivity with much better mechanical integrity. However. The way the accelerometer is mounted will affect the resonant frequency and. The best 24 .

This type of modulation is commonly found in maintenance applications but consider the example below. 25 . The run speed signal is being MODULATED by a signal at 4 Hz which is probably a foundation resonance.Vibration Analysis on Rotating Equipment alternative to stud mounting is to have tap blocks made with a tapped hole that will accept the accelerometer stud. The example above was taken from a large steam turbine running at 3600 rpm. Figure 15 Simple modulation example A signal may be said to be amplitude modulated if the amplitude of that signal is changing over a period of time because of the influence of another signal. Acceleration Amplitude Demodulation Theory But before we look at any case histories using DEMODULATION we should be clear in our mind about exactly what is MODULATION.

As can be seen from Figure 2. the modulation is at this frequency. The 2 kHz vibration is the resonance of the bearing which is being excited by the bearing outer race frequency (3. demodulation is a way of extracting the rate of occurrence of high frequency resonances. The demodulation process extracts the modulating frequency to produce a time waveform which can be handled by the F. In other words the bearing outer race frequency is modulating the bearing resonance frequency. low frequency vibration.Vibration Analysis on Rotating Equipment Figure 16 Bearing modulation example Here we see a vibration at 2 kHz which has been modulated slightly more than three times within the time period (50 mS which equates to 1 revolution of the inner race). high frequency pulses sitting on top of the high amplitude. The bearing excitation resonance is shown as small. When we DEMODULATE the above reading we are not interested in the 2 kHz frequency but we are interested in the outer race defect frequency which is: (1000/50*3.07) Hz = 61.07 x run speed). In vibration terms.F. process. 26 .4 Hz. The excitation of the 2 kHz frequency by the bearing defect on the outer race causes the 2 kHz amplitude to be changed like the “roller coaster” example above. The Demodulation Process The time waveform of a machine with a bearing in the early stages of deterioration will look like the top plot below.T.

T.F. conversion would give a single spike in the frequency domain at the resonant frequency which we have earlier said is not what we want. To modify the signal so as to be suitable for F. Figure 18 Enveloping process With the time domain signal in this format the F.T.Vibration Analysis on Rotating Equipment Figure 17 Demodulation process The demodulator circuit now passes the signal through a high pass filter to give the time waveform shown in the lower section of the plot.F. 27 . we must “envelope” (figure above) each parcel of energy by first rectifying and then passing the signal through a smoothing R-C (resistance-capacitive) circuit.

For vibration readings. while this is often the case it is not always so.Vibration Analysis on Rotating Equipment Figure 19 Fast Fourier Transform The signal is NOW passed through the F. high amplitude noise to pass. and we get a spike in the frequency domain at the bearing defect frequency (figures above and below). the accelerometer which we will use to detect the signal will probably be sitting on top of a magnet which will give a structural resonance in the 1. Conventional thinking will tell you that the resonance frequency which we are using as the carrier wave is always the resonant frequency of the bearing.3D view Resonance Sources When taking a demodulated reading we must first decide on which filter setting to use that will allow the carrier signal to pass without allowing the low frequency. The bearing housing will have its own 28 .F.5 to 4 kHz range (typically). Figure 20 FFT .T.

bearing defect The 2 upper plots are the time domain signal in two planes over a period of 640 mS. This leads us to the conclusion that this is the frequency of 2 and/or 3 kHz seen in the spectrum and one or both of these frequencies are the result of impacts and subsequent ring down and they are occurring at the resonant frequency of part of the mechanical structure. Note that the frequency spectrum shows spikes at 2 kHz and 3 kHz while the time domain plots show an “angel fish” pattern which is classic of a bearing defect. a high frequency oscillation. the carrier wave signal resonance could be coming from any part of the mechanical structure. 29 .i. Figure 21 Two channel time waveform . This plot shows the signal from the inboard bearing of a 35 H. A. A.e.C.C.C.C. Motor Example. motor operating a belt-driven fan. Ultra-sound data are heterodyned to the audible range so demodulating at 5 to 8 kHz gives acceptable results while A. If we are taking a reading with a non-vibration parameter we will probably be utilizing a different carrier signal so we may have to use a different high pass or band pass filter. the machine structure will have its own resonance.P. frequency of 60 Hz or 50 Hz.Vibration Analysis on Rotating Equipment resonance. In short. This shows a detail of the one “angel fish” and the amplitude can be seen to be passing from positive to negative and back again many times during the life of a single angel fish . The lower plots show the time domain (left) and frequency domain (right) over a 50 mS period of the lower 640 mS plot. electric current should be demodulated from the A. Note also that the lower left portion of the plot is a zoom of the windowed part of the long time record.

Vibration Analysis on Rotating Equipment Figure 22 High frequency waterfall The plot above shows a time/frequency cascade of the same time interval cropped below 0. The 2 kHz is the resonant frequency and the bearing defect frequency (outer race) is the modulating frequency. Every time one of the bearing balls passes a defect on the outer race. This clearly shows the modulation of the 2 kHz frequency while the 3 kHz frequency is static. The figure below shows the demodulated spectrum on the left with waterfall plot on the right above a trend of the defect frequency. the ball impacts on the defect causing the 2 kHz vibration to suddenly rise and then ring down. The spikes occur at the bearing defect frequency (outer race) with multiple harmonics but there is no sign of the resonant frequency because this high frequency has been removed during its use in the demodulation process. The frequency range of 30 . Figure 23 Enveloped acceleration spectrum Note that the demodulated spectrum is clean and extremely easy to analyze. The modulation has been calculated to be equal to the bearing outer race defect frequency of the motor inboard bearing.001G.

The last spectrum in the waterfall is lower than the previous spectrum due to greasing of the motor bearings which lowered the amplitude at which the impacts caused the bearing to vibrate at resonance. 31 .velocity to envelope This figure shows a similar defect on another machine but here the velocity spectrum (left) is displayed alongside the demodulated spectrum (right). Note that the demodulated spectrum is much cleaner and easier to analyze. Figure 24 Comparison .Vibration Analysis on Rotating Equipment the spectrum is such that the frequency of the impacts is clearly visible but we do not need to see the resonant frequency.

this can help significantly to identify the specific problem as discussed in the following paragraphs. Armature Related Problems Typical problems associated with the rotor or armature of an induction motor which cause electrical vibration include: • Broken rotor bars This type of defect is best discovered by the use of motor current analysis and comparing the height of the slip * No. If bowing of the rotor is suspected then the condition can usually be corrected by “slow rolling” the motor for up to two days to reset the sag. In vibration readings this will look very like imbalance with a small axial component. A bowed rotor This defect usually occurs on very large. If the vibration disappears the instant the power is shut off. If the nature of the pulsating vibration can be determined. The unequal magnetic forces may be due to open or shorted windings. Electrical problems with induction motors will often cause the motor load current ammeter to swing or pulsate in a cyclic manner. Generally. the problem is more likely to be mechanical in nature. defective bearings and looseness. A common way to check for electrical vibration is to observe the change in vibration amplitude the instant electrical power is disconnected from the unit. In other words these frequencies will have sidebands of 1x. broken rotor bars. the rotor will appear to swing back and forth. • • • 32 . This pulsating vibration common with induction motors will either be a single frequency whose amplitude is being modulated or it will be a beat between two frequencies of vibration which are very close together. If the machine has a “beat” at slip frequency this is usually due to a defect in one rotor winding phase such as a broken conductor or bad brush. this will appear similar to imbalance. the largest component frequency of vibration resulting from these electrical problems will be 1 x RPM and. Mechanical problems may include imbalance. horizontal motors where the motor has sat idle for an extended period and the weight of the rotor causes a sag in the middle of the rotor. Wound rotor windings Defects in the rotor will cause a modulation of rotor bar pass or stator slot passing frequencies at run speed. Vibration caused by electrical problems are normally the result of unequal magnetic forces acting on the rotor or stator. the vibration is likely to be due to electrical problems. if the vibration amplitude decreases only gradually after power is disconnected. unequal air gap and other similar problems.Vibration Analysis on Rotating Equipment Failure Modes Induction Motors Mechanical or Electrical Effects Vibration of electrical motors can be either mechanical or electrical in origin. or if the motor has a strobe light flashed at run speed. of poles sidebands around line frequency to previous levels. An eccentric rotor The variable air gap this produces between the rotor and stator give a vibration at 2 x line frequency with sidebands at pole pass frequency as well as sidebands around run speed of pole passing frequency. On the other hand. If phase readings are taken. unbalanced phases. If this is the case conventional electrical testing procedures can be carried out to pinpoint the true cause of vibration. Perhaps an even better indication of the contribution of electrical problems is by observing the time waveform of the vibration as power is disconnected. it will be seen that the phase is erratic and instead of the strobe “freezing” the rotor. misalignment.

If demodulated spectra are used this will show up as very high spikes at 120 Hz (100 Hz in Europe) in the frequency domain. The other vibration will probably be an electrical vibration which occurs at the rotating speed of the magnetic field powering the motor. Since the mechanical and electrical vibrations are relatively close in frequency their amplitudes will alternately add together and subtract at a rate equal to the difference between their frequencies. Care should be taken if a significant 7.Vibration Analysis on Rotating Equipment • Out of magnetic center This is almost always caused by improper fitting of the bearings. the first thing to check is the individual phase power factor. and even 4x torque pulse frequency. There will be a phase difference in the vibration between energized and de-energized Stator Related Problems Electrical problems in the stator of an induction motor can also result in vibration with a pulsating amplitude. Demodulated readings will demodulate the frequencies above the high pass filter or inside the band pass filter.200 cpm) are apparent in your demodulated spectrum. if an AC line frequency is 60 Hz or 3600 cpm. One of these vibration frequencies may be the result of some imbalance or misalignment occurring at the running speed of the armature. Torque pulses have also been known to excite loose rotor bars and loose stator windings at frequencies of 2x. This causes to motor to run out of magnetic center which gives a vibration at run speed. or if the torque pulses should happen to excite a resonance condition in the machine or structure. Thus. Observing the pulsating vibration in time waveform on an oscilloscope or spectrum analyzer can be useful in identifying the beat frequency characteristics of stator related induction motor problems. 33 . However. This vibration is rarely troublesome except where extremely low vibration levels are required. 60 x 1/3 for a 1200 rpm motor) around 2 x line frequency. The result will be noticeable steady pulsation or “beat” of the vibration amplitude. Common stator related problems which can be expected include: • Stator windings Phasing problems such as a loose connector in the stator windings or supply will cause a modulation or sideband of the reciprocal of the number of pairs of poles x line frequency (e. Electric motors have inherent vibration due to “torque pulses”. A defect in the windings themselves will also cause an unequal magnetic flux around the motor although is usually very small except in severe cases. to produce a vibration whose amplitude pulsates in a cyclic fashion it is necessary that two frequencies of vibration be present. the vibration resulting from torque pulses will be two times the line frequency powering the motor. This range will usually also include the rotor bar pass and slot pass frequencies. If any of the above stator problems are encountered a mechanical vibration will occur at the rotating speed of the magnetic field. Windings insulation Gradual degradation of the insulation of the windings will result in a fall off of the insulation to ground resistance which should be at least 1. in this case the pulsation is the result of a beat between two separate frequencies of vibration which are very close together.5 MΩ for a main drive motor. Torque pulses are generated as the rotating magnetic field of the motor energizes the stator poles. Imbalanced phases A difference in the supply voltage or power factor of the three phases will cause a vibration of 2 x line frequency (120 Hz) around rotor bar passing frequency.200 cpm vibration is seen in velocity vibration readings as this can also be caused by stator distortion brought on by a severe misalignment or a “soft foot” condition. 3x. Since each motor pole is essentially energized twice for each cycle of AC current. If frequencies at 120 Hz (7. torque pulse frequency will be 120 Hz or 7200 cpm. The phase relationship between the two individual vibration frequencies is constantly changing producing a resultant vibration whose amplitude increases and decreases in a periodic fashion.g. • • In the case of a stator related problem. It is the author’s experience that inequalities in the power factor at each of the three phases will cause very large changes in this modulation. A modulation of the rotor bar or slot passing frequencies by 2 x line frequency is not uncommon and does not necessarily mean that there is a defect in the motor.

Direct current is created by taking three phase alternating current and converting it through a bank of silicon controlled rectifiers (SCRs) into direct current. thus. Each of these three problems give the same effect to the motor . looseness. a one second snapshot of three phase AC will reveal 180 positive and 180 negative amplitude peaks When AC is rectified to DC these peaks are electronically processed to allow only positive peaks to remain in the wave form. create a pulsing which is detectable through vibration analysis. The quantitative analysis of the number of broken rotor bars relies heavily on the height of the sideband compared to the height of the line frequency spike (generally expressed in dB Amps) at a fixed speed and a steady load of at least 50% during the measurement. However. 34 . When six SCRs are used to convert the AC to DC. The three waveforms are identical in amplitude and duration.variations in current draw. This means there are three individual alternating current waveforms being supplied simultaneously at 60 Hz but 120 degrees out of phase. induction motors. The project concluded that the presence of sidebands at the motor slip multiplied by the number of poles around the line supply frequency indicated not only the presence of broken rotor bars (or equivalent) but also how many bars were affected. Frequency analysis of the motor load current has been consistently proven to be able to detect the presence of broken rotor bars. When three SCRs (half-wave rectified) are used to convert AC to DC then a pulsing or frequency equal to the 180 Hz is created in the DC drive System.Vibration Analysis on Rotating Equipment Broken Rotor Bars AC induction motors experience a wide range of mechanical problems common to most machinery such as misalignment. DC Motors How DC Power Is Created.800 cpm and 360 Hz x 60 sec. AC is supplied in the United States at 60 cycles per seconds or 60 Hz (Hertz). = 21.600 cpm. rather than Hz. To identify the separate spikes clearly the spectrum must have a fine line resolution. These measurements are usually taken with 1600 or 3200 lines with a bandwidth of 80 Hz (for 60 Hz line supplies) or 65 Hz (for 50 Hz line supplies). A 3200 line spectrum with a bandwidth of 80 Hz will take 40 seconds to collect during which time the load and speed of the motor under test must not vary significantly. Thus 180 Hz x 60 sec. (fullwave rectified or High Efficiency System) then a pulsing or frequency of 360 Hz (6 x 60 Hz) is created in the DC drive system. end ring resistances and cast rotor blow holes.Hz format. The driving force for this project was the large number of motor failures suffered at these production facilities.C. The object of the project was to determine a reliable and repeatable method for the evaluation of broken rotor bars on A. bad bearings etc. In the 1960’s Aberdeen University in Scotland carried out a joint research project with Shell Exploration & Production on several off-shore oil and gas production facilities in the North Sea. Industrial power in the United States is supplied in a three phase 60. they also have their own unique set of problems that are often linked to the electro-magnetically generated fields in the stator and the rotor. = 10. Those who are more comfortable using cpm (Cycles Per Minute). These peaks. as the defective bar cuts the lines of flux. need only multiply the frequency in Hz by 60 seconds. although no longer alternating.

800 cpm and also has 180 positive and 180 negative amplitude peaks per second.600 cpm Hz or 360 Hz x 60 A half-wave rectified DC drive system will then have a dominant electrically related frequency of 10.800 cpm Full-wave rectified direct current ( 6 SCRs ) is 6 x 60 Hz = 360 sec.800 cpm and a full-wave rectified system will have a dominant electrically related frequency of 21.600 cpm When these frequencies exist within their respective systems they should be considered normal unless amplitudes greater than 0. Half-wave rectified direct current (3 SCRs) is 3 x 60 Hz @ 180 Hz or 180 Hz x 60 sec. = 10. = 3.600 cpm • • • Three phase alternating current frequency is 180 Hz or 3 x 60 Hz x 60 sec.Vibration Analysis on Rotating Equipment Figure 25 The creation of DC power Single phase alternating current frequency is 60 Hz or 60 Hz x 60 see. This usually means that there is an electrical control problem.1 in. peak are detected. = 21. 35 . @ 10./sec.

This DC source can be varied through system controls to change the running speed of the motor. The self controlled systems or closed loop systems use low voltage control components to fire or open the pathways which allow the full DC power to be supplied to the motor.Vibration Analysis on Rotating Equipment Figure 26 FFT spectrum of half wave rectification Figure 27 FFT spectrum of full wave rectification DC Systems and Controls Direct current drive systems use rectified alternating current to power an electric motor. self adjusting the power source to achieve the desired speed. This can be controlled manually or be adjusted automatically by allowing the control system to monitor the motor speed through the use of a tachometer. This allows the motor to run at the desired speed or 36 .

As the motor turns.600 cpm between existing spectrum peaks or a frequency of 7. 37 . Because there is one firing order per SCR there is a potential for one or more of the SCRs to perform erratically or not at all if the firing card malfunctions. DC Control Firing Cards The firing cards.200 cpm and frequency differences of 7. If one sees a frequency of 3.Vibration Analysis on Rotating Equipment desired rate of speed change to meet the needs of the driven system. low voltage.200 cpm then faulty firing cards or SCRs may be the cause. This is a very basic explanation of a DC motor circuit but should be sufficient to give an understanding of the concept of how the various components interact. The electrical problems associated with these systems are: • • • • AC power supply AC to DC rectification components DC control components and DC motor component failure Figure 28 Basic DC system circuit The system operates by DC power being supplied to the motor which creates a magnetic field and causes the motor armature to rotate at a speed proportional to the DC power supplied. based on the voltage differential between the constant voltage pot and the tach. signals the firing cards. There is a firing card or order for each SCR which controls the power flow through the SCR. The comparitor card. the tachometer (tach.600 cpm on a full-wave system or 10. DC power frequencies will always be constant at 21.800 cpm on a half-wave system. The SCRs fire or open to create and supply DC power to the motor which either speeds or slows the motor to the appropriate speed determined by the pot. low voltage is compared to a constant or predetermined voltage variance by the comparitor card. must perform correctly to allow the system to function normally. Vibration analysis has been used to determine malfunction on this card or SCR’s based on frequencies which have sub multiples of the DC dominant frequency present. which control the opening and closing of the SCRs.) also turns which creates a low voltage proportionate to its speed.600 cpm and frequency separation of 3. The tach.

Vibration Analysis on Rotating Equipment 1 S. The motor was uncoupled from the drive train and a visual inspection performed on the motor when run under a no load condition.R. and a decision to recouple the motor and operate until a vibration analysis could be performed to determine bearing wear or failure seemed the best alternative. The production system was critical to the plant operation. the motor did start and run. problems Figure 29 FFT spectrum full wave DC firing card frequencies Figure 30 FFT spectrum after repair Example of a Firing Card Fault Vibration analysis was requested on a 300 HP DC motor which was critical to a production system. however. A motor bearing then seemed the next logical failure point. as previously mentioned. One day prior to the request for analysis the system would not maintain the expected speed and the motor was pulsing. The pulsing was still apparent even under a no load condition. A mechanical problem within the gearbox or drive train was initially suspected because the electrical system appeared to be functioning normally. The analysis was performed 1 Thanks to Bill Rinehart for his permission to use this data 38 .C.

200 cpm is 0.600 cpm if the system has: • • • One firing card for each SCR.3037 in/sec peak. This would affect one-third of the system power and virtually render one bank of SCRs inactive. Another possible reason for seeing these 1/3 multiples would be if one phase of the AC power source was not present.600 cpm frequency is the normal frequency of a full wave system and should be present. The dominant frequencies are 7.600 cpm are representative of this situation. This DC control system uses three firing cards to control six SCRs which is typical of many DC control systems. When one firing card is not functioning then 1/3 of the power is lost. The FFT spectrum in figure 27 was taken after the repair.Vibration Analysis on Rotating Equipment revealing the FFT spectrum in figure 26. Full-wave rectified AC power sources can also have frequency separation of 3. not bearings or mechanical defects When an FFT appears with dominant 1/3 frequencies of the DC full pulse frequency (21.400 cpm or frequencies equivalent to multiples of 1/3 of 21.400 cpm and 21.800 cpm or frequency separations of 3.600 for a full wave. yet each spectrum represents the same firing card malfunction.600 cpm.200 cpm and 14. The 7. 14.200 cpm. A simple voltage test of the three incoming AC phases should confirm this situation if present. and one card is out A three card system and one card is partially disabled One SCR is not functioning The FFT spectra directly above and below also show the difference between the firing card frequency amplitudes of a motor that is uncoupled and running under no-load (below) and the same motor coupled and running under a load (above). Although the frequencies are the same in each spectrum the amplitudes are considerably load dependent.400cpm frequencies are now gone. then suspect firing card or SCR problems in the DC control. These frequencies are related to DC ELECTRICAL problems.600 cpm. The amplitude at 7. six SCR system).200 cpm on the spectrum below is only 0. Figure 31 FFT spectrum showing half wave firing card frequencies Half-wave rectified AC power sources will tend to have 1/3 multiples of 10. Dominant frequencies of 7. The 21. 39 .00847 in/sec peak while the amplitude of the spectrum above at 7. The firing cards were examined and a loose connection on one card was found and repaired.200 cpm and 14.

Vibration Analysis on Rotating Equipment Figure 32 FFT spectrum of same motor (no load) DC Control Comparitor Card The comparitor card is another low voltage control component which is responsible for determining the difference between the system actual speed and the set or predetermined speed performance. Testing the voltage output of the tach. should confirm this situation. It has not yet been determined if these side-bands are related to the RPM fluctuation or hunting which often accompanies comparitor card problems or if the constant collapsing and regenerating of the magnetic field of a system that is hunting is the cause. however. they will remain equally spaced. When this component malfunctions it has been observed that there are side-bands present around DC frequencies. 40 . The sidebands do. it has also been observed that these side-bands will grow or diminish as the motor RPM is varied. These side-bands are not of a particular set frequency but are always equally spaced from the DC frequencies. To resolve side bands related to comparitor card malfunctions a FFT spectrum of 3200 lines of resolution at Bandwidth of 24. however. Another possible reason to see these FFT characteristics could be a faulty or malfunctioning tachometer which would corrupt the voltage going to the card. These side-bands may occur at small cpm increments as shown below and may require a high resolution FFT to differentiate them from the dominant frequencies. exist regardless of the cause and should be considered a warning of this component failure.000 cpm is suggested.

A RPM check using a digital tach.Vibration Analysis on Rotating Equipment 0. DC motors are different from AC motors because of their power supply which requires different components. also revealed that the motor was fluctuating or hunting approximately 30 cpm at an RPM of 1440. These units usually have small bearings which can be monitored in the same manner as any bearing. usually. There was not a spare card available at this time but there were spare firing cards. Caution: Never Place A Magnetic Accelerometer Mount On A Tachometer These devices use magnets to generate the voltage which the control system monitors to determine the motor speed. The tach. The first set of spectra was collected under these conditions and revealed the above spectrum. The motor had been uncoupled from the belt to see if the motor vibrated when running solo. The most obvious of these is the tachometer which extends. 41 . The speed fluctuations stopped and another set of spectra collected revealing the spectrum below.000 Figure 33 FFT spectrum showing comparitor card defect. Placing a powerful magnet on or near the tach may alter or destroy the voltage output causing the motor to literally speed up until it destroys itself. from the back of the motor. Based on the side-bands a recommendation was made to change the comparitor card (also called the control card).0872 5000 16. A spare comparitor card was eventually located and the replacement accomplished. voltage was then checked but seemed to be consistent with the operating speed and fluctuations. Example of a Comparitor Card Defect Vibration analysis was requested on a 125 HP DC motor which was thought to be vibrating heavily. The firing cards were changed but did not solve the problem.

once again. It has been observed that as these brushes wear. Never assume that the RPM of the system is what the panel display reads. Ensure that your data always has the RPM stored with it to prevent mistakes when looking at your data for analysis. Importance of Exact RPM Knowing the exact RPM of the variable speed system which is being analyzed is critical to determining which system component is creating the vibration frequency of interest. When the brushes arc it has been observed that these one times RPM readings will increase dramatically. the brush wear would. peak or even higher in extreme cases. these displays are seldom accurate. The use of either a good strobe light. Although the significance of this frequency has not been specifically related to a problem.Vibration Analysis on Rotating Equipment Figure 34 FFT after the comparitor card was replaced Figure 35 DC motor components The commutator is the device which transfers the DC power to the motor armature. be suspect./sec. sometimes reaching 0.3 ins. Brushes. Another frequency associated with the commutator is the number of slots on the commutator times the motor RPM. or a digital tachometer will give an accurate RPM. readings at one times the motor RPM will rise in amplitude. usually made of a carbon alloy. 42 . ride against the commutator and supply the DC power to the commutator.

700 cpm. The motor is one of four motors which must work in sequence for the line to run efficiently.Electrical Frequencies Vibration analysis was requested on a 5 HP DC motor which was causing problems on a production line. a 25% difference from 3. What was assumed to be a 3600 cpm separation was due to a bearing outer race defect. A quick reference reading of G/SE (high frequency pulse detection) was taken and registered a higher than normal value. The motor RPM was reduced by 25% from 1004 RPM to 756 RPM and new spectra collected.600 cpm. Figure 36 FFT from a 5 HP motor . be sure that the resolution of the spectrum is enough to see the difference between closely positioned frequencies. mechanically related problems will follow the speed increases or decreases proportionately and electrical frequencies will remain constant.600 cpm separation. The motor had a SKF 6208 bearing which has a 3.full wave rectified 43 .Vibration Analysis on Rotating Equipment When determining a frequency relationship to a machine component. The spectra was collected revealing the FFT in figure 33. This is a powerful analysis tool when determining a mechanical or electrical frequency. Remember. The frequency separation was now 2. Example of Mechanical -v. If necessary take multiple spectra at different resolutions to capture and differentiate all possible frequencies. The following spectra will verify the importance of correct RPM and the ability to determine mechanical frequencies by noticing how the frequencies change related to the RPM. which had what was initially suspected to be DC electrical frequencies with a 3. this is not characteristic of an electrical problem.606 BPFO multiplier (1004 RPM x 3.606 = 3620 cpm). The frequency separation followed the RPM proportionately indicating a mechanical problem.

Vibration Analysis on Rotating Equipment Figure 37 Same motor .speed lowered by 25% 44 .

F m is the mass in kg ω is the rotational speed in radians per second and r is the radius at which the force (the out of balance) is acting then the out of balance forces = [0.81 = [0. = [0.Vibration Analysis on Rotating Equipment Rotating Equipment Imbalance Let us consider a fan impeller of 50 kg weight which is running at 1000 rpm. units F is force in kg. (1476 lb. Newton says that: F = (m ω2 r)/9.71 kgF.F) By doubling the speed to 2000 rpm we have quadrupled our out of balance forces to almost three-quarters of a ton force.722 x 0.5 x 104.5 x (1000/60⋅2π)2 x 0.68 kgF (369 lb.5 x 10.3]/9.2π)2 x 0.000 rpm is 0.920 lb. Let us see what happens if the impeller is running at 2000 rpm.680 lb.7937 = 2.3 meter from the center (we will use S. at 1. units here to make the math easier). By increasing the speed to 2.000 rpm and allowing an imbalance force of about 1 lb at about 1 foot away from center we have halved the effective load carrying capacity of the bearing. 2 3 Data taken from TRW service catalog Form 382-14 Anti Friction Bearing Manufacturers Association 45 .966x 0.I.81 = 670.000 rpm as calculated using the 3AFBMA method of evaluating load ratings.3]/9.680 x 0.3]/9.81 = 167.5 kg at 0. The speed factor for calculating radial load at 2.I.7937 so our load is now 3. which is in the MRC mid range of bearings. This bearing has a static load rating of 23.5 x (2000/60.81 using S.3]/9. Let us imagine that this impeller has an out of balance corresponding to 0. A deep groove Conrad bearing for a shaft of about 3 inches diameter could be a MRC 215-S.F) In other words we have added over an eighth of a ton to the apparent weight of the impeller.81 = [0.

Vibration Analysis on Rotating Equipment Vibration due to imbalance Figure 38 Imbalance slide 1 Figure 39 Imbalance slide 2 46 .

Vibration Analysis on Rotating Equipment Figure 40 Imbalance slide 3 Figure 41 Imbalance slide 4 47 .

Vibration Analysis on Rotating Equipment Figure 42 Imbalance slide 5 Figure 43 Imbalance slide 6 48 .

Vibration Analysis on Rotating Equipment Figure 44 Imbalance slide 7 Figure 45 Imbalance slide 8 49 .

Phase readings will be steady .Vibration Analysis on Rotating Equipment Figure 46 Imbalance slide 9 To confirm a suspected imbalance check the time waveform. Multiple plane balancing (more than two planes) is usually only necessary on complex multiple disk rotors such as turbines which operate above their critical speed. Out of balance may occur in more than one plane. Most spectrum analyzers have a two plane balancing function built into the software. 50 . An rotor imbalance will give a sine wave in the velocity time signal with a period equal to the time it takes for one revolution of the rotor.

Vibration Analysis on Rotating Equipment Misalignment Figure 47 Misalignment slide 1 Figure 48 Misalignment slide 2 51 .

Vibration Analysis on Rotating Equipment Figure 49 Misalignment slide 3 Figure 50 Misalignment slide 4 52 .

Vibration Analysis on Rotating Equipment Figure 51 Misalignment slide 5 53 .

Vibration Analysis on Rotating Equipment Looseness Figure 52 Looseness slide 1 Figure 53 Looseness slide 2 54 .

Vibration Analysis on Rotating Equipment

Figure 54 Looseness slide 3

Figure 55 Looseness slide 4

55

Vibration Analysis on Rotating Equipment

Figure 56 Looseness slide 5

Figure 57 Looseness slide 6

56

Vibration Analysis on Rotating Equipment

Figure 58 Looseness slide 7

Figure 59 Looseness slide 8

57

Vibration Analysis on Rotating Equipment Figure 60 Looseness slide 9 58 .

PM and will appear similar to normal imbalance. coupled to the fan gearbox by means of a long torque tube or drive shaft. On centrifugal fans. If it is confirmed that a condition of excessive aerodynamic vibration is not due to resonance. If the fan operates under a constant aerodynamic load. If a fan must operate smoothly over a broad range of aerodynamic loads it may be necessary to check and correct for significant variations in blade track or pitch before this -can be achieved. After the fan had been satisfactorily balanced. This vibration results from the fan blades striking the air and will occur at a frequency equal to the number of fan blades times fan R. The result is often excessive vibration at the aerodynamic frequency and may require that the distance between the blade path and torque tube be increased to minimize these aerodynamic pulsations. In this case. The torque tube can act as an obstruction to the smooth flow of air through the fan and an aerodynamic pulsation is generated each time a fan blade passes over the torque tube. 59 . the amplitudes of vibration resulting from aerodynamic forces will be low and no cause for concern. This aerodynamic imbalance will result if the fan blades do not have the same track or pitch. the fan should be checked carefully for obstructions that may disturb the smooth flow of air through the fan. Vibration due to aerodynamic forces can also occur at a frequency equal to 1 x fan R. a centrifugal fan was balanced with the access doors in the fan housing removed to simplify the addition and movements of trial weights. it was necessary to balance the fan operating under its normal aerodynamic conditions. Unfortunately. For example.PM. it often occurs that changing the fan load will produce a corresponding change in the vibration at 1 x R. mounted outside the fan venturi.PM. Normally. excessive vibration at the aerodynamic frequency can sometimes result if the fan rotor is positioned eccentrically in -the f an housing. However. this produced a significant change in aerodynamic conditions and the result was a significant increase in vibration. the access doors were replaced. For example. Many of these fans consist of a drive motor. however. the force of aerodynamic imbalance can be compensated by following normal balancing procedures. Therefore this should be checked in the event that the problem cannot be traced to resonance. described near the end of the manual should be carried out to determine the resonant part. high amplitudes of aerodynamic vibration are sometimes encountered on cooling tower fans. When excessive vibration at the aerodynamic frequency is encountered a common cause is resonance of some part of the machine or structure and the checks for resonance.Vibration Analysis on Rotating Equipment Vibration due to aerodynamic forces Fans and blowers will inherently have some vibration due to aerodynamic forces.

The manufacturer of the compressor can supply performance characteristics curves showing the range of stable operation in terms of pressure ratio. The frequency at which this whirl occurs can vary from one machine to the nextThe vibration may have the same frequency characteristics of oil whirl and hysteresis whirl where the lowest natural frequency of the rotor bearing system is excited. Typically a compressor is designed to deliver air or other gases over a specified mass flow range and at a specified pressure ratio. a reversal of gas 60 . similar to the rotating oil whirl in a plane bearing.Vibration Analysis on Rotating Equipment Figure 61 Aerodynamic forces Aerodynamic cross coupling Aerodynamic cross coupling is a problem occasionally encountered on turbines. centrifugal compressors and fans operating above the first rotor critical and generally results from eccentric rotation of the rotor caused by rotor bow or deflection. the layer of rotating gas between the rotor and the machine housing will produce a torque reaction on the rotor causing the rotor to whirl at the rotating speed of the gas layer. Attempting to operate the unit outside the design range can result in excessive vibration and damage to the machine. it seems logical that the condition would be affected by machinery load. In general. In most cases. Surging Surging is a rather common problem encountered on high speed centrifugal and axial flow compressors and occurs when the compressor is operated outside designed limits. If the rotor is bowed slightly. number of compressor stages. for a particular operating speed. Since rotor whirl generated by aerodynamic cross coupling is excited by -the compressed air or gas. Cases have been reported where sub-multiples of 0. blade configurations and other factors. mass flow and rpm. The problem of surge occurs when.25 times the rpm of the rotor have been excited. the delivery pressure to inlet pressure ratio is too high or if mass flow is too low relative to design conditions.5. These requirements are met by selecting rotor speed. the vibration frequency will be less than the rotating frequency.33 or perhaps 0. When this occurs. 0. In the case of a centrifugal compressor the layer of air or other gas being compressed will have a rotating speed less than that of the rotor. the machine will be more likely to experience this condition under heavily loaded conditions and changing the load of the machine to determine its effect on the vibration can be useful in -diagnosing this problem.

Rotating stall shows itself as a vibration at approximately 35% of run speed but is dependent on the physical configuration of the compressor. Choking occurs when discharge pressures are too low. In cases of mild surge. When a full surge condition is encountered the result may be a high amplitude of random.Vibration Analysis on Rotating Equipment flow in the compressor will result. shaft and other components. erratic vibration usually covering a rather broad frequency range. if this condition is allowed to continue. a noticeable increase in the vibration at blade passing frequency can usually be detected. velocities are high and when flow velocity in the diffuser section approaches Mach 1 a turbulent or circulating flow between the blades will occur which has the effect of blocking the flow of gas. temperature. rotor blades. casing. flow etc. extensive damage to the compressor c result. the flow reversal may only occur in the boundary layers of the rotor blades (a rotating stall) however. This frequency is the product of the number of rotor blades times the rpm of the rotor. Choking or Stone Walling The problem of 'choking or 'stone-walling' in a compressor is essentially the opposite of surging but again is the result of attempting to operate the unit outside design parameters. at full surge the gas flow reverses its direction and flows from the discharge to the inlet. In other cases multiples of blade passing frequency may also be detected. In the initial stages of surge. The vibration characteristics resulting from compressor surging can vary depending on the extent of the problem. diffuser blades. The vibration characteristics of choking are essentially the same as those encountered during surging. should be undertaken to distinguish between the two. Of course. When this occurs a noticeable drop in efficiency and pressure ratio can be seen. This is caused by the turbulent flow within the compressor exciting the various natural frequencies of the rotor wheels. 61 . along with an increase in vibration due to the turbulent flow within the compressor. A check of other operating parameters such as pressure. When discharge pressures are low.

an increase in bearing load. So high. THERE IS NO METAL TO METAL CONTACT IN A PROPERLY LUBRICATED BEARING RUNNING AT NORMAL LOAD. Figure 62 Elasto-hydrodynamic lubrication Rolling element bearings have four stages of failure: • • • • Stage 1 Stage 2 Stage 3 Stage 4 Lubrication problems Marking of the raceways and / or rolling elements Spalling of the raceways and / or rolling elements Collapse of the bearing We will discuss each stage of the failure and discover how each stage may be identified with predictive maintenance techniques. is very small but is large enough to keep the roller physically separated from the race by a small distance (in the micron range). The lubricant wedge itself. in fact. As the lubrication mechanism starts to break down the lubricant loses its ability to separate the roller from the race. This may be due to a deterioration of the lubricity. As the roller rotates the pressure point is very small so the pressure loading is very high. 62 . overheating of the bearing or a fatigue failure of an old bearing at the area most prone to elastic deformation. trapping a very small amount of the lubricant into a “wedge”. Elasto Hydrodynamic Lubrication Let us consider the lubrication of a rolling element bearing. Let us consider the lubrication of rolling element bearings.Vibration Analysis on Rotating Equipment Bearing Failures Arguably the most common type of failure on rotating equipment are failures arising from the collapse of the bearings. that the contact point of the roller and the race (the contact ellipse) becomes elastically deformed.

63 . Set at above 20 kHz for industrial applications.just like hitting a bell with a hammer. While this is fairly common on lightly loaded motor bearings. this can be seen to have a very significant effect on bearing health. We know that over packing is the most common cause of raised bearing temperature. may generate high bearing temperatures when running at high speeds. In order to avoid the temperature effect due to over packing. Figure 63 Loss of Lubricant . 4 Ultra sonic . For a rolling element bearing these frequencies will be in the range of 1 to 4 kHz.above the human audible range. Since the life of a grease lubricant is strongly temperature dependent and since grease loses half of its life for every 20°F rise in temperature.Ball Bearing Inner Race Courtesy of the Barden Corporation At this stage of the bearing failure the rolling elements have metal to metal contact onto the raceways because the lubricant is no longer supporting the rolling element via a lubricant wedge.Vibration Analysis on Rotating Equipment First Stage of Bearing Failure Lubrication problems in the case of grease or oil systems is not necessarily the physical loss of lubricant but the loss of the oil or grease’s properties. Both of these frequency ranges do not experience any modulation at this stage as the metal to metal contacts are irregular. Low speed. or bearings which have lost grease due to physical migration. which leads to reduction in grease life and eventual failure. the metal to metal impacts excite the resonant frequencies of the bearing . starved bearings usually wear into a condition of excessive looseness and fail without appreciable temperature increases. Because of this. bearing manufacturers usually suggest that bearings are packed with grease to between 15% to 20% of the bearing's free volume. press main drive motors will usually fail catastrophically. Under-packed bearings. The metal to metal contacts also generate 4ultra sonic frequencies at between 30 and 50 kHz.

This is a bearing resonant frequency and the rise and fall in amplitude coincides with the impacts from the rollers on the outer race defect (in this case). As each roller passes the void it impacts onto the race. Remember that the lubricant wedge is microscopically small so even a tiny crack in the material can cause the roller to impact heavily onto the race. Notice the “roller coaster” in Figure 64 at just below 2 kHz. At this stage of the failure there is no appreciable rise in temperature and the velocity vibration at the defect frequencies is insignificant. it works its way outward and eventually develops into a spall. In other words the resonant frequencies are excited every impact. If a crack or contaminant is found at that location. abrasive contamination. This phenomenon limits the number of cycles a bearing can survive. or rollers. the load is distributed over a different (often smaller) area and therefore greatly increases stress at that point. and wear. 64 .Vibration Analysis on Rotating Equipment Figure 64 Loss of Lubricant . Simply. These material anomalies again provide for a point of crack formation if they fall within a high stress area. As the crack reaches the surface it creates a small void into which the lubricant wedge collapses. and once a crack is formed beneath the surface. spalling. Surface fatigue is usually caused by scratches on races. the impacts generate vibration at resonant and ultra sonic frequencies. giving a vibration like the pattern below. a given load over a given area produces stress. that these frequencies are now modulated by the rate at which the rollers hit the defect. This fatigue may be initiated on the surface or beneath the surface. however.Roller Bearing Courtesy of the Torrington Company Second Stage of Bearing Failure This the early stage of fatigue. These marks produce “stress raisers”. with the difference. a point on the bearing surface that experiences abnormally high stress due to the physical conditions at that spot. As with the first stage. or brinelling. balls. Subsurface fatigue is usually caused by voids. foreign matter or coarse carbides introduced into the material at the time of formation. These raised stress areas provide a start point for micro-crack formation that leads eventually to pitting.

65 . Literally a stress related failure of the material which results from cyclic stresses due to operation at high loads. the velocity vibration becomes apparent at the defect frequencies and harmonics. The resonant vibration has also increased in amplitude at the defect frequencies along with a general rise in floor level. the very small area produces very high point loading so the material starts to flake off around the crack. At this stage of the bearing failure. As the rollers repeatedly impact onto the small crack.Vibration Analysis on Rotating Equipment Figure 65 Waterfall plot from a damaged motor bearing Third Stage of Bearing Failure Fatigue failure or spalling results from mechanical materialogical failure of the bearing. As the crack propagates. particularly not if there is any air movement around the surface of the bearing or bearing housing. Temperature will be elevated above normal but not significantly. possible also with sidebands of run speed and / or cage frequency. more and more material is removed until the crack becomes a visible cavity or spall.

as is common in sealed and grease-packed bearings.Vibration Analysis on Rotating Equipment Figure 66 Early Fatigue . Figure 67 Developed Fatigue on Roller Bearing Courtesy of the Torrington Company 66 . continuous recirculation of the particles causes progressively higher wear to the point where either the bearing becomes excessively loose and fails to support the load suitably or the induced damage leads to failure. or are trapped in the bearing. growth in spalls. and eventual machine failure. in oil-lubricated bearings. In these latter bearings.Ball Bearing Courtesy of the Barden Corporation Fourth Stage of Bearing Failure With time both surface and subsurface fatigue flaws spread over the active bearing surfaces causing bearing wear. The metal contaminants or wear particles removed from the bearing during spalling are either washed out with the oil.

The following figure illustrates the bearing geometry used in the above equations 67 ..45 (rotating inner race) RPM * 0. it is suggested that the approximate equations be used to establish the bearing frequencies of interest. or frequency with which the roller or ball impacts on the defect indicates potential failure and allows us to determine the type of damage that exists.(BD/PD) cos θ] approximated by RPM * O.(BD/PD)2 COS2 θ] (no valid approximation) where PD θ n BD = ball (roller) diameter = pitch diameter = contact angle = number of balls (rollers). A number of characteristic frequencies are generated by a damaged bearing .4* n BPFI = n * RPM/2 [1 + (BD/PD) cos θ] approximated by RPM * O.6 * n BSF = (RPM * PD)/2BD * [1 .Vibration Analysis on Rotating Equipment Bearing Defect Frequency Calculation It is important to understand that as a bearing with a damaged surface rotates. When bearing geometry is not known but the number of balls or rollers can be counted or estimated. the regularity. All the equations listed above show a direct dependence of the calculated frequency on the frequency of rotation.(BD/PD) cos θ] approximated by or (rotating inner race) RPM * 0. and are known as:• • • • • Cage or fault train frequency Ball pass with respect to the outer race = BPFO Ball pass with respect to the inner race = BPFI Ball rolling about its own axes Shaft frequency of rotation = BSF = RPM = FTF The following equations are used to calculate these frequencies FTF = RPM/2 [1 .55 (rotating outer race) BPFO = n * RPM/2 [1 .

Vibration Analysis on Rotating Equipment Figure 68 Ball Bearing Terminology 68 .

For machines that are outside the speed ranges quoted the amplitudes will be lower for slower machines and higher for faster machines. As with any severity chart it is important that the vibration analysts use their own judgment and experience when deciding whether or not to change the bearing.5 hrs over 14 days The plot above shows a waterfall display of acceleration vibration up to 2 kHz over a period of almost two weeks on the drive end bearing of a drive motor.not the derived peak seen in the frequency spectrum. A roller bearing should not be allowed to operate at above about 12 G (5true peak) and a ball bearing should not be allowed to operate at above about 7 G (true peak). 5 69 . Notice also that there appear to be small sidebands around the resonance frequency which are at the bearing outer race defect frequency.001G.Vibration Analysis on Rotating Equipment Figure 69 Waterfall of early damage to a motor bearing collected every 1. Don’t forget that roller bearings will stand more impactive forces than ball bearings. The presence of a bearing defect frequency excited resonance does not give justification by itself to change the motor bearings. The envelope (or demodulation) amplitudes are quoted in dBG re 0. We must also wait until the defect frequencies with sidebands and/or harmonics show up in the velocity spectrum. The severity chart in figure 69 is applicable for motors and fans running between 500 to 3600 rpm. True peak is defined as the peak seen in the waveform . Notice that the vibration at about 1. Velocity readings are heavily affected by the mass of the machine so care should be taken when assessing very small or very large motors.5 kHz has risen steadily. This frequency is typical of the bearing resonance.

Vibration Analysis on Rotating Equipment Figure 70 Bearing damage severity assessment chart Analysis of bearing defects Figure 71 Demodulated acceleration spectrum from a dry bearing The figure above was taken (as was most of these examples) from the drive end bearing of a drive motor.this 70 . Notice in the spectrum that there are no significant spikes but the spectrum is raised up from the floor .

If the bearing is fitted with seals then it is sometimes possible to force some grease pass the seal with a grease gun. the bearing only had a lubrication problem and you have just fixed it. If a plug is fitted opposite the grease fitting make sure that you remove it. In the case of a main drive motor the bearing is almost certainly grease lubricated. Figure 72 Demodulated acceleration spectrum of a marked bearing 71 . Injecting too much grease into the bearing cavity will cause pressurization of the cavity and the grease will force its way past the bearing into the motor windings As you inject the grease into the bearing have a spectrum analyzer attached to an accelerometer on the bearing housing and watch the vibration levels. The grease lubricated bearing may be of shielded (or sealed) construction or non-shielded construction. If the bearing is non-shielded then the bearing should be lubricated following the bearing and motor manufacturers instructions. A simple rule of thumb for bearing condition is that if the vibration goes down and stays down.Vibration Analysis on Rotating Equipment is a raised carpet level. If the vibration level goes down but rises again then the bearing is damaged . the chances are that the bearing is starting to suffer lubrication problems. A certain amount of rise from the floor is normal but when you find that your alarm limits (based on the baseline) have been exceeded in the demodulated spectrum and there are no significant spikes.the sooner the level rises again the worse condition the bearing is in (from a few of days to several minutes for a very bad bearing). but usually you will just have to watch the bearing deteriorate and change the bearings before the damage gets so bad as to cause secondary damage.

1. At this stage of the bearing deterioration it is sometimes still possible to save the bearing with additional lubrication.000 1. 72 . Note that the fundamental frequency has several harmonics but it is the fundamental frequency which will coincide with the generated bearing defect frequency and help us in our analysis.Vibration Analysis on Rotating Equipment 0.872 Figure 73 Demodulated acceleration spectrum from a slightly more heavily marked bearing Figures 71 and 72 show two examples of rolling element bearings which have suffered some marking of the races (both of these examples show marking of the stationary outer race).000 0 Time Waveform mS 160 Figure 74 Time waveform from a marked bearing. At the very least you will extend the life of the bearing.

If these figures are exceeded then the bearing is almost certainly severely damaged. a ball bearing should not be allowed to run with impacts of more than 0-7 G peak in the time domain and a cylindrical roller bearing should be running less than 12 G.500 0 Time Waveform mS 160 Figure 75 Time waveform from a heavily marked bearing The two time waveform spectra above show the individual impacts caused by the rollers impacting onto the damaged raceways. we can correlate this time difference with the frequencies seen in the frequency spectrum. 0.0463 Figure 76 Velocity spectrum from a spalled bearing 73 .Vibration Analysis on Rotating Equipment 2. Looking at the time interval between the peaks in the time domain.500 2. Generally.

if you see the same defect frequencies in the demodulated acceleration and the velocity spectra. 74 . As a general rule of thumb. irrespective of amplitude.Vibration Analysis on Rotating Equipment As the bearing deteriorates the bearing defect frequencies start to show up in the velocity spectrum. then that bearing is spalled. Figure 75 shows a velocity spectrum from a spalled bearing with multiple harmonics of the outer race defect frequency.

transport the part to a balancing machine. Since this is true it would be to our advantage to be able to correct this condition as easily as possible. The faults shown are exaggerated but could easily exist in almost every rotating part manufactured. balance under artificial conditions. operating as it normally does. the eccentric hole location and eccentric machining of the pulley groove all add up to more weight on one side of the rotating centreline than the other. In-place Balancing is a straight forward process which involves following a few simple rules. 75 . The following pages deal directly with this application. In-place Balancing The conditions which must be satisfied so you can balance can almost always be met with the imbalanced part mounted in its own bearings. However. The vibration must be due to imbalance. before we discuss balancing we should first understand imbalance. we must be able to make the weight corrections on the rotor. or other type of key phasor) for phase measurements. In-place Balancing eliminates the need to disassemble the machine. The blow hole in the casting. Figure 77 Sources of imbalance Imbalance is the unequal distribution of the weight of a part about its rotating centreline. Illustrated are some reasons for an unequal weight distribution. the thicker web. and assures smooth operation of the machine when you are done. The process of balancing a part without taking it out of the machine is called In-place Balancing. and we must have a shaft angle reference (such as being able to observe the rotor with the strobe light as used in this paper.Vibration Analysis on Rotating Equipment Balancing Imbalance has been named the most common cause of vibration in our machines. Before we can balance a part with the Vibration Analyzer. where it comes from and what must be done to correct it. certain conditions must be satisfied.

Vibration Analysis on Rotating Equipment Figure 78 Assembly tolerance stack up More important as a source of imbalance is the stack up of tolerances possible when assembling rotating parts. This heavy spot represents the accumulated results of all of the imbalance of the pulley. An equal amount of weight at the same radius but opposite the heavy spot will balance the rotor. The example illustrated is a very common one. Figure 79 Heavy spot All of the combinations of unequal weight distribution can be considered to be concentrated at one point called the heavy spot. 76 . Thus the assembly is out of balance. The hole in the pulley is necessarily larger than the shaft diameter and when a key or set screw is attached the take-up in clearance shifts the weight of the pulley to one side of the rotating centreline of the shaft.

The resulting imbalance is twotenths of an ounce-inch. The mass centre of a part is more commonly known as the centre of gravity or C. The centre of gravity or mass centre is the point about which all of the weight of a body is equally distributed.ounce pulley on a shaft with a two-thousandths clearance between the hole and the shaft. One ounce-inch of imbalance would be a heavy spot of one ounce in weight one inch from the rotating centreline. gram-inches or. Three ounces located three inches from the centre would be nine ounce. 77 . The example illustrated shows a condition of mass centre displacement of two thousandths of an inch. more commonly now.inches. gram-centimetre.Vibration Analysis on Rotating Equipment Figure 80 Units of measure of imbalance The units for measuring imbalance are generally ounce-inches. Figure 81 Mass centre displacement The imbalance of a part may also be given in terms of the distance from the rotating centreline to the mass centre of the part.G. Such a condition could very easily be duplicated by installing a 100.

The amount of force generated by one ounce-inch of imbalance at high speeds is surprising and explains why good balance becomes absolutely necessary for high speed machines. therefore.77 is a constant required to make the formula dimensionally correct. 1.77 (RPM/1000)2 ounce-inch. Vibration Related to Imbalance Balancing is the process by which we learn the amount and position of the heavy spot so we can either add an equal amount to the opposite side of the rotor or remove weight at the heavy spot. For instance. RPM. We will use the position of the reference mark as seen by the strobe light to tell us the location of the imbalance. In the illustration the part has an imbalance represented by a heavy spot located some distance from the centre of rotation.Vibration Analysis on Rotating Equipment Figure 82 Force due to imbalance The effect of imbalance is that it generates a force which causes the part and the supporting bearing and structure to vibrate. at 3600 RPM nearly 23 pounds of force is generated for each ounce-inch of imbalance. 78 . “F'' is the force in pounds'. The size of the force generated depends on the speed of rotation and the amount of imbalance. the force generated can be computed by the formula F = 1. When the imbalance is given in terms of gram-inches the formula becomes F = 1/16(RPM/1000)2 gram-inches. If the heavy spot and distance are in ounces and inches. the rotating speed of the part and ounce-inch the amount of imbalance. So we will use the amount of vibration to indicate how much imbalance we have. the greater the vibration. We have learned that the more imbalance we have the greater the force.

If we double the amount of the imbalance weight without changing its position the vibration will increase to 10 mils and the reference mark will still appear at 12: 00 o'clock when the strobe light flashes. If the trial weight were smaller than the imbalance. or where the trial weight must be placed to be opposite the original imbalance and how large the trial weight must be to be equal but opposite the original heavy spot. If we move the weight 30 degrees counterclockwise from its first position we will still see the same amount of vibration. The experiment tells us two things: 1. The change caused by the trial weight can be used to learn the size and location of the original imbalance. The imbalance in the part at the start is called the original imbalance and the vibration amount and phase which represents that imbalance is called our original reading. Note that the reference mark appeared at the same place each time. Let's start by adding a trial weight. The reference mark shifts in the opposite direction to a shift of the heavy spot. First. The angle the reference mark shifts is equal to the angle of shift of the heavy spot. The amount of vibration is 5 mils and the reference mark appears at 12: 00 o'clock. but the reference mark will now appear at 11: 00 o'clock or 30 degrees counterclockwise from where it was before. in which direction.creased or decreased to be equal but opposite the original imbalance? 79 . and should its amount be in. At the start of a balancing operation we do not have the least idea how large the heavy spot is nor do we know where on the part it is located. the vibration will in.Vibration Analysis on Rotating Equipment If an imbalance weight is added to a perfectly balanced part it will have a vibration at a frequency equal to the rotating speed of the part.Single Plane We said earlier that balancing is the process by which we learn the size and position of the heavy spot so we can either add an equal amount to the opposite side of the rotor or remove weight at the heavy spot. By what angle should the trial weight be shifted. When this happens the reference mark shifts to a new position and the vibration displacement changes. The amount of vibration is proportional to the amount of the imbalance. To balance the part all we have to do is move the trial weight directly opposite its first position and adjust the amount until we reach a satisfactory balance.8 mils of vibration and a reference mark position at 12:40 o'clock. 10 mils. The second thing that could happen is we could add the trial weight in exactly the right place opposite the heavy spot. but the reference mark will appear at 1: 00 o'clock or 30 degrees clockwise from it’s first position of 12: 00 o'clock. The part will vibrate a certain number of mils and the reference mark will appear at some definite position when the strobe flashes. This experiment shows . We change the original imbalance by adding a trial weight to the part.it is correct to use the amount of vibration to tell us how much imbalance is in the part. To balance we would reduce the trial weight amount until we reached a satisfactory vibration level. 10 mils. The new total imbalance in the part will be represented by a new amount and phase of vibration. Starting with the part out of balance. If the trial weight were larger than the imbalance then it’s position would now be the heavy spot and the reference mark would shift 180° or directly opposite where it was originally. What we must do is find a weight and position for that weight which would oppose the original imbalance. Three things can happen. If we do. we see 1. To balance all we would have to do would be to add more weight until we reached a satisfactory vibration level. then the heavy spot would still be the heavy spot but we would see a decrease in vibration and the reference mark would be in the same place. How can we use this information to balance? How to Balance .crease and the reference mark will appear in the same position it did on the original run. What happens when the position of the heavy spot or imbalance weight is changed? If we move the heavy spot 30 degrees clockwise from where it was before we will see the same amount of vibration. if we are lucky we might add the trial weight right on the heavy spot. The third thing that can happen is the usual one where the trial weight is added neither at the heavy spot or opposite it. 2.

O + T. until it crosses line B. We see this as a new vibration displacement and reference mark position or phase. is not clockwise as shown here but is counterclockwise. Remember our experiment where we learned that the reference mark moves in the opposite direction to that of the heavy spot. A vector is simply a line whose length represents the size of the imbalance and whose direction represents the angle of the imbalance. Parallel to line A we draw a line outward from the centre of the graph. After adding a trial weight the imbalance due to both the original plus the trial weight is represented by 2. 12: 40 o'clock.30 mils and a phase of 2. line B. Using polar coordinate graph paper we plot our original imbalance vector.30 o'clock. Line B then represents the vector for the trial weight. Then parallel to the original imbalance vector and from the end of the original plus the trial weight vector. In order to balance the part we want the trial weight to be equal and opposite to the original imbalance as shown by the dotted line. however. we should adopt this rule. We can see that the trial weight must be moved by an angle equal to the angle between the trial weight vector and the place where we want it to be. When the reference mark after adding a trial weight moves clockwise from its 80 . The direction in which the trial weight acts with respect to the original imbalance is represented by the direction of line B. Next we draw the original plus the trial weight imbalance vector “O + T'' to the same scale.8 mils and a phase of 12:40 o'clock. When a trial weight is added to the part we actually add to the original imbalance in the part and change the position of the total imbalance to some new position between the two. The direction we must move the weight. It is drawn at the new angle for the original plus the trial weight imbalance phase 2: 30 o'clock. Our original imbalance was represented by 1. We select a convenient scale for the length of the line such as 1 mil equals 2 divisions. Therefore. We draw a line from the centre at the same angle we saw our reference mark. line A. we draw another. We'll call this vector “O” for the original imbalance. This is all the information needed to solve for the effect of the trial weight alone.Vibration Analysis on Rotating Equipment Single Plane Vector Method of Balancing This question can best be answered by making what we call a vector diagram. line C. Each hour is equal to 30°. Figure 83 The vector diagram To solve for the trial weight vector we must complete the vector parallelogram. To complete the vector parallelogram we first connect the end of the original imbalance vector to the end of the original plus trial weight imbalance vector. The length of line B represents the amount of the trial weight.

is the same as the angle between the trial weight vector. To determine the correct amount of the trial weight we simply multiply the trial weight by the length of the original imbalance vector and then divide by the length of the trial weight vector. By following these instructions carefully the part should now be balanced by moving the trial weight 75 degrees counterclockwise and reduce its size. measure the angle between “O'' and “T''. and this becomes the “T'' vector or Trial Weight Vector. line B. Therefore. and A. O. connect the ends of the two. or line A in the more complex vector diagram shown above. The correct balance weight is equal to the trial weight x O/T. is longer than the “O'' vector. Figure 84 Simplified vector diagram Our vector method of balancing then becomes quite simple. to solve the vector problem we actually need only the three lines. We know this because the trial vector. 81 . and adjusting the weight to the proper amount can result in some remaining vibration still due to imbalance. The direction of this angle is opposite from the shift of the reference mark on the trial run. The balance weight should be added at an angle equal to the angle between “0'' and “T'' away from where the trial weight was added. O + T. Furthermore. In the example shown the trial weight is larger than it needs to be. very small errors in measuring the phase angle. However. line B. To solve this problem measure the lengths of line “O'' and line ”T''. the angle between the original imbalance vector and line A. the dotted line. We draw the “O'' vector then the O + T vector. in shifting the weight. and its position to get the desired effect to balance the part.Vibration Analysis on Rotating Equipment original position then 'we must more the trial weight counterclockwise or shift the trial weight opposite the shift of the reference mark. By inspecting our vector parallelogram we can see that the line A connecting the ends of the original imbalance vector and the original plus trial weight imbalance vector is exactly the same length as the trial weight vector line B.

Vibration Analysis on Rotating Equipment

Figure 85 Additional corrections
To make further corrections we simply observe the new amount and phase of vibration, .5 mils at 9:00 o'clock. We must plot this reading as a new “O + T'' (shown above). We then connect the ends of “O'' and the new “O'' + “T'' vector to find the new “T'' We can see that we shifted the trial weight too far and must move it back by the angle between “O'' and the new “T'' This may be repeated as many times as necessary but always using the original “O'' line. This then is the vector method for single plane balancing. It is simple to use and provides accurate information to balance a part in the fewest number of starts and stops of the machine as possible.

Four-step Method of Balancing Single Plane
Another method for balancing which is not as precise as the vector method follows the same basic procedure except we do not construct a vector diagram. The four- step method follows a few simple rules to find the proper location for the correction weight after which the amount is adjusted to balance the part. Since it generally requires many starts and stops of the machine it is not too popular except when the number of balancing runs is not important. First we observe the amount of vibration and the phase or position of the reference mark when the strobe light flashes. In the illustration these readings are 2.4 mils at 5: 00 o'clock. This is the original run. AII future data will be referred to these readings just as we did in the vector method. Next we add a trial weight to the part at any convenient location and again observe the amount and phase; 1.8 mils at 7:00 o'clock.

82

Vibration Analysis on Rotating Equipment

Figure 86 Direction to shift the weight
Our first goal is to shift the trial weight to a position where the reference mark returns to its original position or 180° away. We do this by shifting the trial weight in a direction opposite the shift of the reference mark. Do not be afraid to move the trial weight by a large angle because if you move it too far, the phase will direct you to move it back. The reference mark shifted from 5: 00 o'clock to 7: 00 o'clock after we added the trial. This is a clockwise shift of the reference mark. Therefore, we move the trial weight in the counterclockwise direction. A new reading might be 2.4 mils at 3:û0 o'clock (not illustrated) . This would indicate we moved the weight too far because the mark is counterclockwise from the first position of 5: 00 o'clock. Shifting the weight again we see the reference mark has been returned to its original position, and the vibration has been reduced to 1 mil. This means the weight is in the proper position and we need only to increase the size of the trial weight. If the reference mark had appeared 180° away from its original position the weight would have been in the proper place but too large. If the reference mark had returned to its original position but with an increased vibration then the trial weight would have been on the heavy spot. It should be apparent that the vector method and the four-step method both follow the same basic rules. Know and understand both of these methods well. Sample Problems Following are some examples of balancing problems. You are given the original readings and the readings after a trial weight has been added. See if you can solve each problem correctly. Your answer should indicate first the direction to move the trial weight, second the angle, and third, the adjustment required to get the correct size of the trial weight

Amount 1. Original Reading First Trial Reading 5.0 mils 3.0 mils @ @

Phase 3: 00 o'clock 4: 00 o'clock

Answer. The reference mark moved from 3: 00 to 4: 00, a clock- wise shift. The trial weight must be moved counterclockwise from its position for the trial run.

83

Vibration Analysis on Rotating Equipment

Figure 87 Sample problem vector diagram
Discussion: The vector diagram was constructed as shown above. The “O'' vector was scaled off to five units long representing five mils vibration. The “O'' + “T'' vector was scaled off to the same scale as the “O'' vector. The ends of the “O” and “O” + ''T'' vector were connected. This is the ''T'' vector. The angle between the “O'' vector and the “T'' vector is the angle the trial weight must be moved from its position for the trial run. The length of the “T'' vector was measured using the same scale used for the “O'' and “O'' + ''T'' vector The effect of the trial weight is smaller than the original imbalance because the “T'' vector is shorter than the “O'' vector. The amount the trial weight must be increased was calculated using formula Corr. Wt. = T.W. x O/T.

The solution is to move the trial weight 32° counterclockwise from its position on the first trial run. At the same time increase the size of the trial weight 1.75 times its original size. Now, you work the rest. Amount 2. Original Reading First Trial Reading 3.0 mils 4.0 mils @ @ Phase 7: 00 o'clock 7: 00 o'clock

3. Original Reading First Trial Reading

9.0 mils 4.5 mils

@ @

10: 30 o'clock 4: 30 o'clock

4. Original Reading

18.0 mils

@

9: 00 o'clock

84

This information can be learned for any system but applies only to the individual system. once a unit has been balanced by either of the methods described before it is possible to learn how much and where weight must be added to balance the unit in one run. lf the weight is added at two inches 12 grams will be needed. Before a trial weight was added to the pulley the phase or position of the reference mark was 2:00 o'clock. This is true of course only for the machine mounted the same way. 40 gram-inches of balance correction was required for g 5. After the corrections indicated by the readings above have been made a new reading was taken indicating additional correction is required.1 mils with the reference mark stopped at 2: 00 o'clock. For example if we measure 3 mils. Original Reading First Trial Reading 9. To show how this is done we will use the motor and pulley system illustrated before. operating at the same speed. 12: 00 o'clock Determine the corrections required. for the example is 5. Second Trial Reading 3.0 mils @ 9: 00 o'clock 5. With the reference mark at 2:00 o'clock we can see in the illustration that the weight was added at 5:00 o'clock.0 mils of vibration or 8 gram-inches per mil. Original Reading First Trial Reading 9.0 mils 1. With the pickup mounted in the vertical direction and the filter tuned to rotating speed of the pulley we measure the vibration caused by the imbalance. Balancing in One Run Do not let the title of this section fool you. The imbalance corrected was therefore 40 gram-inches. If the system is changed in any way.0 mils 9. Then try some real problems on a simple system like the one pictured here in the text. We will assume that it is the pulley that is out of balance. However. The heavy spot or point of imbalance is 180° away or at 11: 00 o'clock when the strobe light flashed. of course.Vibration Analysis on Rotating Equipment First Trial Reading 3. 85 .0 inches was required.0 mils @ @ 2: 00 o'clock . After the part is balanced to a vibration of 0.5 mils @ @ 2: 00 o'clock 9: 00 o'clock 6.1 mil we can learn how much weight was added to correct for 5. The next time it is necessary to balance this particular unit or one just like it we will know 8 graminches of correction is required for each mil of vibration. 24 gram-inches is required.0 mils of vibration. how much and where weight must be added will also change. and with the vibration pickup in the same place. Knowing how much weight to add is important. but where is even more important. We will assume that 20 grams of weight added at a radius of 2. To date no method has been devised to permit balancing in only one run the first time a unit is balanced. The pulley's position with the reference mark at 2: 00 o'clock is the position of the pulley when the strobe light flashed. Therefore. This is the original reading which.0 mils @ 4: 45 o'clock What additional correction is required? Check your answers to these problems with the answers supplied further in the manual.

These principles are also useful in two plane balancing. Remove the reference mark from the pulley to avoid any confusion regarding it’s position and the heavy spot. So that you can be sure you understand we suggest the following experiment.Vibration Analysis on Rotating Equipment Figure 88 Determining the flash angle The angle between the point where the pickup is applied and the position of the heavy spot when the strobe light flashes is called the Flash angle for the system. The location of the heavy spot with respect to the position of the pickup when the strobe light flashes should be apparent. 86 . Balance a pulley mounted on a motor similar to the one we have used here. The reference mark is simply a way to see the position of the pulley when the strobe light flashes. for this system the heavy spot will always appear at 11: 00 o'clock or 30 degrees counterclockwise from the pickup when the strobe light flashes. Now put the pickup in a new location and repeat the experiment. Therefore. Do the same thing with the heavy spot in other locations. The reference mark has nothing to do with the relationship since it can be placed anywhere on the pulley. Next put a heavy spot on the pulley and observe its position when the strobe light flashes. The heavy spot can be at any angle around the pulley but it will always appear at 11: 00 o'clock when the strobe light flashes. After these relationships have been learned about a part then many parts may be balanced on a production basis.

follow the manufacturers torque settings. The block is designed for mounting to the surface with an epoxy adhesive such as Loctite DEPEND or similar. for this reason the material quoted is bright steel or austenitic stainless. Before mounting the accelerometer it is worth considering the purchase of accelerometers with local connectors rather than those with integral cables so that if the motor has to be dismantled it is easier to disconnect the accelerometer. Be very careful not to over-tighten the accelerometer onto the tap block . Do not mount the accelerometer directly onto the surface of the motor with the adhesive unless you want to lose your accelerometers every time your motor goes away for repair. The tap block should be mounted without the accelerometer and allowed to set firmly. 87 . 1/4” Figure 89 Typical tap block for mounting an accelerometer When it comes time to mount the accelerometer. Width across flats should be just wider than the base of the accelerometer. This will help the transmission of vibration through the tap block and allow for easy disassembly.Vibration Analysis on Rotating Equipment Single Channel Analysis Taking measurements The figure below shows a typical tap block for mounting an accelerometer to a motor. it must set hard in order to transmit the vibration. This block should be machined with a “gramophone” finish on the reverse side to give good adhesion to the mounting surface. Whichever adhesive is used. The motor mounting surface should be prepared with a hand grinder or similar to remove all paint and get down to bare metal. Drilled & tapped 1/4” 28 Material: Bright Steel or Austenitic Stainless Note: One side to be machined smooth and the other side to be machined to a rough “gramaphone” finish. The material used for the tap block should be capable of being magnetized so the tap blocks can also be used for off line measurements with magnet and accelerometer prior to an accelerometer being permanently fitted. put a thin layer of grease on the accelerometer and a thin layer of epoxy adhesive on the top of the tap block. holding the tap block in place while the adhesive is setting with duct tape or similar. Notice that the tap block is made from hex. stock so that if it is necessary to remove the block then it can be wrenched off with a suitable spanner.

L. When running your accelerometer cable back to the junction box. 6 V = I. Failure to do this will result in high amplitude multiples of electrical line frequency (60 Hz). Case isolated accelerometers are usually less susceptible to picking up electrical frequencies. pulse. which gives a 4-20 mA output.L. 88 .. care should be taken to ground the shield wire at one end only .T.the end furthest away from the motor . keep the cable as far away as possible from the motor power cable to avoid “cross-talk”.L. once per rev from an installed proximity switch. triggered by a key or from the installed tachometer channel on the P.C. Before installing this resistor check with the engineers responsible for the drive system to make sure the drive system or the P.C.Vibration Analysis on Rotating Equipment Particularly on motors. is not adversely affected by the resistor. To use this output the 4-20 mA must be dropped across a resistor of about 6250 Ω to give a 2. I is amps. A speed output should be installed for variable speed motors. Once installed the speed input will have to be calibrated in volts / rpm and fed to a channel input as opposed to a trigger input as would be required for the T.5 volt drop for a 10 mA signal and 5 volt at full speed (20 mA).T. R is ohms. Do not connect both ends of the shield to ground as you will get beautiful ground loops that will really dominate your signal.R where V is volts.on to a good electrical ground. This could be in the form of a T.L. pulse.

Vibration Analysis on Rotating Equipment Thin film of silicone grease Steel stud Max temp 1000°C (1800°F) Thin film of silicone grease Mica washer Steel stud Max temp 250°C (482°F) Thin layer of bees wax Max temp 40°C (100°F) Methyl cyanoacrylate cement (super glue) Methyl cyanoacrylate cement (super glue) soft glue Steel stud Max temp 80°C (178F) Figure 90 Accelerometer mounting techniques a-d 89 .

Vibration Analysis on Rotating Equipment Thick double sided adhesive disk Double sided adhesive disk Thin double sided adhesive disk Max temp 95°C (200°F) Max temp 150°C (300°F) Hand held probe Figure 91 Accelerometer mounting techniques e-g H a n d P ro b e D u a l R a il M a g n e t F la t M a g n e t M o u n t in g P a d A d h e s iv e M o u n t S tu d M o u n t Sensitivity Deviation (dB) ~ Ref.0 1 0 1 0 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0 0 L o g F re q u e n c y (H z ) Figure 92 Overview of accelerometer mounting techniques 90 . 100 Hz + 4 0 + 3 0 + 2 0 + 1 0 0 -1 0 -2 0 1 .

Vibration Analysis on Rotating Equipment Potential Failure Analysis A methodology for objective set up by Ron Frend Introduction The purpose of monitoring equipment in predictive maintenance (PdM) is to be able to assess the health and condition of the machine relative to any potential failures. 91 . Failure type This is the failure that we could expect from the base cause. Base cause As the name suggests this is the root cause of any potential failure. In order for us to carry out this assessment we must be absolutely certain that we are taking the correct measurements at the right place and that we are taking them often enough so that we do not miss a developing failure. for example misalignment could cause a bearing failure or a shaft breakage. The methodology that we will employ to ensure that we are taking the right measurements with the correct parameters is the potential failure analysis (PFA) tree. In other words this is a very short description of the actual failure that the machine would suffer should the base cause carry on without remedial work being carried out. misalignment. The PFA Tree The tree is structured in the following way: Setup Interval Analysis Parameter Technology External Manifestation Failure Type Base Cause Let us take each of these components and look at them in detail from the roots up. Examples could be: lubrication problems. The base cause often branches out to more than one failure type. manufacturing defects and so on.

Setup To detect the misalignment we have said that we will use vibration technology with a velocity parameter and that we are looking for two or three multiples of run speed in the frequency spectrum. For example if we are looking for rolling element bearing defects we should take the readings at least once a month but if we are looking for a misalignment then once every three months would be ample. Technology So now we know how the problem will show it’s ugly head but what is the best technology to detect it. 92 . At this stage you need to have detailed understanding of the failure modes and predictive maintenance applications. Parameter So the technology of choice for detection misalignment is vibration. For the heat exchanger we may decide that the most suitable parameter is a radiometric thermal image or maybe we will decide to use a simple point and shoot infra-red thermometer. whereas a plugged heat exchanger would show itself as a rise in differential pressure and with a change in temperature from optimum of the cooled fluid. An example would be that a misalignment would show itself as an increase in vibration at the machine. We have already said that misalignment shows itself as a rise in vibration so the technology to use would be a vibration based technology.Vibration Analysis on Rotating Equipment External manifestation This is where we start getting technical. If that is the only defect that we are looking for then we can safely set the maximum frequency in the spectrum at about ten time run speed with 400 lines of resolution as we are not looking for small changes in frequency such as slip sidebands. How will the problem show itself at the various stages of failure? For illustration we will discuss the four stages of bearing failures and how they show themselves. The heat exchanger temperature change may be detected by a mercury-in-glass thermometer or alternatively by the use of infra-red technology. Interval How often do we need to take the reading to ensure that we do not miss a developing problem. Our misalignment will show as an increase in velocity vibration in the axial direction for a belt driven machine at run speed with two or three harmonics and our thermal image will show a high temperature gradient across the cooler. Analysis At this point we are giving the analyst an idea of what detail he should expect to see in the chosen parameter when the subject machine has a developing defect. but what type of vibration is best? In this case the best vibration parameter is velocity.

Vibration Analysis on Rotating Equipment Developing a Potential Failure Analysis for Rolling Element Bearings Figure 93 PFA development for rolling element bearings Consider figure 1. Before you can develop the PFA tree for the rolling element bearing you must have a thorough understanding of the possible failure modes and extrapolate these modes to their logical conclusions. 93 . To detect a defect at this stage with vibration we need to use a high frequency technique such as envelope signal processing (ESP) with the following parameters: Fmax Envelope filter No lines Window Averages = BPFI x 8 (or thereabouts) = 2. It shows the four stages of a rolling element bearing failure.5 . At this stage we can detect any problems using high frequency vibration parameters such as demodulation or ultra sonics.5 kHz (for electric motor bearings up to about 250 HP) = 400 or 800 = Hanning. Stage 1 Notice that the first stage of the defect is linked to lubrication problems. above. Hamming or Kaiser = typically 4 with maybe 50% overlap If your data collector does not support a high frequency function such as ESP or HFB then take an overall acceleration reading.

Stage 4 The final stage of failure is when the bearing physically collapses. To calculate the Fmax in the point setup screen we use the following formula: Fmax = No. The onset of stage 2 can be also be detected by searching for a rise in amplitude in resonance frequencies so if you do not have ESP you can take a velocity or acceleration spectrum with an Fmax above resonance. The velocity reading should be set up as follows: Fmax or No lines Window Averages = BPFI x 8 (if also using ESP or similar reading) = 2.Vibration Analysis on Rotating Equipment Stage 2 The next stage of failure involves light marking of the bearing. If we expect resonance to occur at about 2 kHz then the Fmax should be about 3 kHz. If we require a resolution of 1. Stage 3 The third stage of the bearing failure is when the bearing starts to spall. We may also see sidebands of run speed or cage frequency around the defect frequencies or their harmonics. We should be taking an ESP and a velocity reading.56 Example. Note that if you are using seismic velocity transducers you will not be able to look this high in the frequency range. At this stage we start to see the defect in velocity usually at the third or fifth harmonic of the bearing defect frequency.5 kHz (do not confuse AC motor frequencies with a brg defect) = 400 or 800 = Hanning. So the choice here would almost certainly be ESP. Temperature measurement using embedded thermo-couples have been used 94 . At this stage the bearing overheats dramatically and very quickly. Usually 1.5 kHz or the nearest available frequency. the appropriate Fmax would be: Fmax = 1024 0. of samples time for 3 revs x 2.024 samples is adequate resolution.2 milliseconds (mS). For a bearing on a shaft running at 1195 rpm the time for on rev is: 60/1195 = 0. Again the parameter of choice is a high frequency technique but experience tells us that a spectral parameter is most suited to detecting marked races.0502 x 3 x 2. We also need enough resolution in the time domain to be able to differentiate the impacts and calculate the frequencies of occurrence of the impacts.0502 seconds or 50. Hamming or Kaiser = typically 4 with maybe 50% overlap We should also take a time waveform reading in G acceleration to look for the bearing impacts at all stages of the failure.656 Hz So we would set the Fmax to 2.024 points. The time duration should be calculated to be enough to show about 3 revs of the shaft. The setup for the ESP reading would be the same as in stage 1.56 = 2. The ESP reading should be the same as that set up in stage 1.

The procedure for building the PFA tree is the same for any equipment or component: • • • • Identify the possible failure mechanisms Determine how the failure mechanism will show itself Quantify the best contemporary method for detecting the failure mechanism Define set up parameters to best use the detection method for this failure mechanism Press Main Drive Motor PFA Tree Base Cause Dry/Marked bearings Failure Type Bearing Failure External Manifestation High frequency resonance @ bearing defect frequency intervals Technology Vibration Parameter Demodulation (Ultra Sound) Time Waveform Analysis Spikes @ Brg defect frequencies (20 dBG) See sect. & harmonics and/or sidebands.5 Peaks @ 12 G (roller) 7 G (ball brg) Interval On Line Set Up Fmax = BPFI x 8 Radial G acceleration = 3 revs Fmax = BPFI x 8 (higher if not using resonance parameter) Radial = 3 revs G acceleration Fmax = 10 x rpm Axial Fmax = 10 x rpm Vertical Fmax = 120 kcpm Drive End Compare SCR to SCR Compare phase to phase 1 baseline On Line On Line Spalled bearings Bearing Failure Harmonics of bearing defect frequencies Vibration Velocity Time Waveform On Line Misalignment Bearing Failure/ Shaft Failure Bearing Failure/ Shaft Failure/ Structural Failure Reduced Power Motor Burnout High axial vibration Vibration Velocity Multiples of run speed axial > radial Multiples of run speed up to 15x.PreDiCon Figure 94 PFA for a main motor 95 . We now include all of the information from our failure study of the rolling element bearings into our potential failure analysis tree (figure 2). s/band on DC(Sect.s Vibration Thermography Velocity Thermal Image Amps Thermal image (spot temp) Megger On Line 6 month 6 month 6 month Winding defects Motor burnout Different current / Current Draw input phase Stator temp. Possible sub-harmonics 240 Hz.2.C. rise/ Thermography Uneven temp. Not only have we considered the possibility of bearing failure but we have included insulation breakdown. loose components and even drive problems on D. motors. which is operating under these particular conditions. 2. of poles around 60 Hz (Sect. See sct. 1/3 DC pulse.5) Bad SCR is colder Unbalanced supply Hot spot on stator On Line Looseness Run speed harmonics & subharmonics Vibration @ SCR Frqys Changes in SCR temps.2) 6 month 6 month Motor de-energized Press idling Rotor defects Rotor failure Line frequency + sidebands Vibration @ rotor bar pass frequency Motor load Motor current current analysis frequency Vibration Velocity Motor burnout On Line Fmax = RBPF x 3 Radial c. may eventually suffer.5 MΩ Sidebands of slip x No. Vibration Velocity On Line Bad S. 2. Ron Frend . distribution Earth current leakage Insulation test Should be > 1. Notice that the bearing failure is only one of the potential failures that we now know of.2) Sidebands of 1x (Sect 2.R. rotor bar defects.5 Peaks in waveform Spikes @ Brg defect freqys.Vibration Analysis on Rotating Equipment successfully for protection systems but usually only for thrust bearings which develop the raised temperature before standard radial support bearings.C. 2. Including the Component Failure in the PFA Tree. We have carried out potential failure analyses on all of the possible defects that we think a motor.

The engineer or technician defining the measurements must be aware of the limitations of the technology which is used for the measurement. DISSOLVER FEED CHUTE FAILURE MODES CHUTE ASSEMBLIES LINER WEAR CLAMP/SEAL FAILURE CAM SIEZURE INSPECTION . repeatable and useful measurements if these factors are taken into consideration: • • • • The engineer or technician defining the measurements must be familiar with previous failures encountered on machines similar to the machine under scrutiny The engineer or technician defining the measurements should be cognizant of all available predictive maintenance technologies.CCTV FORIEGN BODY INGRESS CORROSION FRETTING/CLEARANCE INCREASE INSPECTION .Vibration Analysis on Rotating Equipment Conclusion The method outlined in this paper for setting up the type of readings in a predictive maintenance system will result in quantifiable. don’t limit yourself to only taking vibration measurements .CCTV (wieght ?) INSPECTION .there are a host of other technologies out there which complement vibration in a predictive maintenance application.CCTV (reference marks req'd) HEAD 1 96 . Finally.CCTV (reference marks req'd) DRIVE ASSEMBLY GREASE SEAL FAILURE MATERIAL CHOICE DESIGN ROTARY SIEZURE BEARING COLLAPSE BEARING CLEARANCE INCREASE OVER TEMPERATURE LOADING MATERIALS / LOADING LUBRICATION / LOADING DESIGN DESIGN DESIGN DRIVE SIEZURE FORIEGN BODY INGRESS CORROSION BRINNELLING BEARING DEFECTS SLIDE BUSH SIEZURE LUBRICATION / CLEARANCE MATERIAL COMPATIBILITY / CORROSION DESIGN DESIGN GEAR WEAR / DEFECTS DIRT/MATERIALS CHOICE NOISE / VIBRATION INSPECTION .

Vibration Analysis on Rotating Equipment BEARING DEFECT DETECTION BEARING COLLAPSE OVER TEMPERATURE FORIEGN BODY INGRESS CORROSION PROBLEMS BEARING DEFECTS ULTRA-LOW SPEED PARTIAL ROTATION INNER BEARING MOUNTS BRINNELLING FREQUENCY RESOLUTION NOT POSSIBLE TIME-BASE READINGS REQ'D IN : ACCELERATION STRESS WAVE PROBLEM RADIATION HEAD 2 97 .

See sct.5 Peaks @ 12 G (roller) 7 G (ball brg) Interval On Line Set Up Fmax = BPFI x 8 Radial G acceleration = 3 revs Fmax = BPFI x 8 (higher if not using resonance parameter) Radial = 3 revs G acceleration On Line On Line Spalled bearings Bearing Failure Harmonics of bearing defect frequencies Vibration Velocity Time Waveform On Line Looseness Bearing Failure/ Shaft Failure/ Structural Failure Run speed harmonics & subharmonics Vibration Velocity Multiples ofBPFO or BPFI in velocity Sect. 3.5 Peaks in waveform Spikes @ Brg defect freqys.PreDiCon 98 .5 On Line Fmax = 10 x rpm Vertical 1 Baseline c. Ron Frend . & harmonics and/or sidebands. 3.Vibration Analysis on Rotating Equipment Press Flywheel PFA Tree Base Cause Dry/Marked bearings Failure Type Bearing Failure External Manifestation High frequency resonance @ bearing defect frequency intervals Technology Vibration Parameter Demodulation (Ultra Sound) Time Waveform Analysis Spikes @ Brg defect frequencies (20 dBG) See sect. 3.

its points are multiplied by the window function. To make up for this reduction in power. In the frequency domain. and Kaiser. What happens if a signal is not exactly periodic within the time record? We said that its amplitude is divided into multiple adjacent frequency bins. Several types of window functions are available including Uniform (none). In order to reduce the effect of leakage. which can have several different shapes. and noise floor. What is windowing? Let's go back to the time record. This is called "leakage". If the filter is wide. but this is impossible to achieve in practice. since the filter is narrow. Hanning. and the resulting time record is by definition periodic. windowing algorithms give extra weight to the values near the middle of the sequence. it is as if the spectral lines are wider. The shape of the "leaky" spectrum depends on the amount of signal truncation. A signal will appear as narrow as a single bin if its frequency is exactly equal to a frequency bin. resulting in a discontinuity in the sampled signal. Uniform The uniform window is actually no window at all. the signal level is not at zero at one or both ends of the time series record. They start and stop at zero and are smooth functions in between. This smearing will also change wildly between records because the amount of mismatch between the starting value and ending value changes with each record. The difference between each smoothing window is the way in which they transition from the low weights near the edges to the higher weights near the middle of the sequence. it is necessary that the signal level is forced zero at the beginning and end of the time series. This is true but it's actually a bit worse than that. These sidebands. time bin by time bin. If the filter is narrow. However. If there is no windowing function used. If a sine wave is passing through zero at the beginning and end of the time series. A narrow filter is called a selective window . energy in the signal "leaks" from its proper location into the adjacent lines. When the time record is windowed.it selects a small range of frequencies around each bin. The time record is used with no weighting. The measured amplitude of the weighted signal is also incorrect because a portion of the signal level is removed by the weighting process. it also adds distortion to the time series which results in sidebands in the spectrum. It may not be identical from record to record. This discontinuity causes problems with the FFT process. BlackmanHarris. it falls off from center rapidly.Vibration Analysis on Rotating Equipment Measurement Windows Many people get confused with this topic. If. The amplitude of each frequency bin is determined by centering this filter on each bin and measuring how much of the signal falls within the filter. Flattop. amplitude accuracy. on the other hand. then frequencies far from the bin will contribute to the bin amplitude but those close by will not be attenuated significantly. The net result of windowing is to reduce the amount of smearing in the spectrum from signals not exactly periodic with the time record. This is done by multiplying the data samples by a "smoothing window" function. and the result is a smearing of the spectrum from a single line into adjacent lines. (It is exactly periodic 99 . truncation of the waveform will occur. and is generally unpredictable for real signals. This means that even frequencies close to the bin may be attenuated somewhat. "Flat". this is called "Rectangular". or "Uniform" windowing. but it will be periodic (zero at each end). or side lobes. While the smoothing window does a good job of forcing the ends to zero. the resulting FFT spectrum will consist of a single line with the correct amplitude and at the correct frequency. Windows are functions defined across the time record which are periodic in the time record. a window acts like a filter. The different types of windows trade off selectivity. Leakage could be avoided if the time series zero crossings were synchronized with the sampling times. If the time record does not start and stop with the same data value. the signal can actually smear across the entire spectrum. then only frequencies near the bin will contribute to the bin. effectively reduce the frequency resolution of the analyzer.

a very good approximation of the actual random noise floor can be displayed. The RMS average has no phase information. In general. It has an amplitude variation of about 1. As a result.5 dB (for signals between bins) and provides reasonable selectivity. the Flattop window has a wide pass band and very steep rolloff on either side. Blackman-Harris The Blackman-Harris window is a very good window to use with the spectrum analyzer. RMS averaging reduces fluctuations in the data but does not reduce the actual noise floor. If its frequency is between bins.02 dB. Flattop The Flattop window improves on the amplitude accuracy of the Hanning window. averaging many spectra together improves the accuracy and repeatability of measurements. Kaiser The Kaiser window. signals appear wide but do not leak across the whole spectrum. Unlike the Hanning. Thus.7 dB) than the Hanning. it will affect every bin of the spectrum. (The real part is averaged separately from the imaginary part. Its between-bin amplitude variation is about 0.Vibration Analysis on Rotating Equipment within the time record). displaying the real or imaginary part or phase of an RMS average has no meaning. 100 . Its filter roll off is not particularly steep. the selectivity is a little worse. which is available on IRD analyzers. Because of these properties. even when their amplitudes are very different.8 dB for signals between exact bins). this window is only useful when looking at transients which do not fill the entire time record. However. the Hanning window can limit the performance of the analyzer when looking at signals close together in frequency and very different in amplitude. With a sufficient number of averages. This allows signals close together in frequency to be distinguished. it is the best window to use for measurements requiring a large dynamic range. The Kaiser window has the lowest side-lobes and the least broadening for non-bin frequencies. Averaging In general. The filter is steep and narrow and reaches a lower attenuation than the other windows.) This can reduce the noise floor for random signals since they are not phase coherent from time record to time record. Hanning The Hanning window is the most commonly used window. RMS Averaging RMS averaging computes the weighted mean of the sum of the squared magnitudes (FFT times its complex conjugate). These two cases also have a great deal of amplitude variation between them (up to 4 dB). Since RMS averaging involves magnitudes only. It has better amplitude accuracy (about 0. The weighting is either linear or exponential. combines excellent selectivity and reasonable accuracy (about 0. Vector (Synchronous Time) Averaging Vector averaging averages the complex FFT spectrum. very good selectivity and the fastest filter rolloff.

it is possible to compute the spectra for every time record with no loss of data. Averaging takes place according to the formula. the analyzer does not wait for the next complete time record before computing the next FFT. The signal of interest must be both periodic and phase synchronous with the trigger. The resulting display shows the peak magnitudes which occurred in the previous group of spectra. most window functions are zero at the start and end of the time record. Linear Averaging Linear averaging combines N (number of averages) spectra with equal weighting in either RMS. Once in steady state. the display shows the real or imaginary part or phase of the complex peak value. This type of averaging is useful for eliminating transients. Instead it uses data from the previous time record as well as data from the current time record to compute the next FFT. and Octave Analysis measurements. Thus. The update rate would be no faster than one spectra per time record. This is because the complex information is preserved. What about narrow spans where the time record is long compared to the processing time which is what we normally see when taking vibration measurements? The analyzer computes one FFT per time record and can wait until the next time record is complete before computing the next FFT. the points at the ends of the time record do not contribute much to the FFT. New Average = (New Spectrum * 1/N) +(Old Average) * (N-l)/N where N is the number of averages. Remember. some data samples will be lost while the FFT computations are in progress. Real Time Bandwidth and Overlap Processing What is real time bandwidth? Simply stated. The spectra are computed in "real time". the new spectral magnitudes are compared to the previous data. Vector or Peak Hold fashion. With vector averaging. the peak magnitude values are stored in the original complex form. and if the new data is larger. With narrow spans. Exponential averages "grow" for approximately the first 5N spectra until the steady state values are reached. the real and imaginary parts of the signal will not add in phase and instead will cancel randomly. Peak Hold Peak Hold is not really averaging. At this span and below. instead. this could be quite slow. PSD. 101 . the real and imaginary parts as well as phase displays are correctly averaged and may be displayed. At larger spans.Vibration Analysis on Rotating Equipment Vector averaging requires a trigger. Otherwise. This speeds up the processing rate. And what is the processor doing while it waits? Nothing. Make sure that the number of averages is not so large as to eliminate the changes in the data that might be important. Peak Hold detects the peaks in the spectral magnitudes and only applies to Spectrum. This is done on a frequency bin by bin basis. these points are “re-used" and appear as middle points in other time records. If the real or imaginary part or phase is being displayed for spectrum measurements. then the new data is stored. Exponential Averaging Exponential averaging weights new data more than old data. This is why overlap effectively speeds up averaging and smoothes out window variations. With overlap. With overlap processing. further changes in the spectra are detected only if they last sufficiently long. it is the frequency span whose corresponding time record exceeds the time it takes to compute the spectrum. However.

= 1 kHz x 2(n-30/3) 102 . The full octave bands have band centers at: Center Freq. Typically the shape of each band is a third-order Butterworth filter whose bandwidth is either a full or 1/3 octave. Octave analysis measures spectral power closer to the way people perceive sound. yet sufficient samples are gathered for a valid analysis. 0% is no overlap and 99. For vibration analysis of rotating machinery a good overlap is 50% as this ensures that no data is zeroed out by the smoothing windows. Octave analysis computes the spectral amplitude in logarithmic frequency bands whose widths are proportional to their center frequencies. For vibration measurements.Vibration Analysis on Rotating Equipment Overlap Percentage The amount of overlap is specified as a percentage of the time record. in octaves. that is. The bands are arranged in octaves with either 1 or 3 bands per octave (1/1 or 1/3 octave analysis). The maximum overlap is determined by the amount of time it takes to calculate an FFT and the length of the time record and thus varies according to the span.8% is the typical maximum. Octave Analysis The magnitude of the normal spectrum measures the amplitudes within equally divided frequency bins.11 (1986). octave analysis is now only used for sound analysis or for quality acceptance checks at the end of an assembly line but is included here for the sake of completeness. = 1 kHz x 2n The 1/3 octave bands have center frequencies given by: Center Freq. Octave Band Center Frequencies The center frequency of each band should be calculated according to ANSI standard S1.

800 cpm (for a 3600 rpm motor) . Once a spike at a bearing frequency has been identified you should check the baseline for this type of machine for the trend.Vibration Analysis on Rotating Equipment Analysis The following procedure gives an overview of the procedure to take in the analysis of vibration frequency spectra.4 Ball Pass Frequency Inner Race Elements x 0. One common bearing frequency is just over 3 x run speed for BPOR on a 8 element bearing. If the trend is deteriorating then further checks are necessary. Of Rolling Run Speed x No. Of Rolling Run Speed x 0. • • • • • 3) As a rule of thumb: Ball Pass Frequency Outer Race Elements x 0. Activities: 1) Select the first plot of the machine which will be at the driver outboard and will be an enveloped acceleration reading. The run speed of an electric motor cannot exceed the electrical speed so harmonics of run speed cannot have a frequency even slightly more than 3600 cpm (for a 1200 rpm motor) or 5400 cpm (for a 1800 rpm motor) or 10. Predictive Maintenance Procedure Date:July 11. Identify the run speed accurately.4 2) Note that this vibration is not necessarily direction specific.if the spike is even at a slightly higher frequency then it is likely caused by a bearing defect otherwise it is likely a run speed harmonic or an electrical frequency. 2003 Procedure Name: Signature Analysis Purpose: To provide a routine procedure for analysis of vibration in order to promote understanding of the relationship between vibration frequencies and their causes.if not you must change the speed reference before continuing. If you have frequency information for the machine ensure that the reference speed is accurate . 103 . Remember that the bearing frequencies may not be completely accurate if the bearing which has been nominated in the frequency setup has been replaced with an equivalent.6 Cage Frequency ≅ ≅ ≅ Run Speed x No. In the envelope spectrum see if any of the generated frequencies coincide with (or are close to) any significant spikes. Be careful that you do not confuse a run speed harmonic or an electrical frequency with a bearing defect frequency. Check the time domain signal for “angel fish” patterns.

Vibration Analysis on Rotating Equipment

4)

In the single spectrum plot double click on the convert the display to dB(G) (referenced to 0.001 G). Identify the carpet amplitude in dB(G) - this is the average amplitude (excluding spikes). Identify the amplitude of the spikes above the carpet level. The following rules of thumb apply to enveloped acceleration levels in dB(G): • • • > 20 dB(G) rise in carpet level 10-15 dB(G) spike above carpet > 20 dB(G) spike above carpet under lubrication minor marking marked race.

5)

If the spectrum shows multiples of run speed then there is an impact every rev of the rotor, with possible looseness if there are many multiples. If the bearing defect frequencies have sidebands of cage frequency then there is a FALSE BRINELLING problem. If the bearing defect frequency and harmonics have sidebands of run speed then there is probably a defect on the inner race. As the defect deteriorates then the carpet level will rise and the sidebands and harmonics will increase in amplitude up to a certain amplitude then stop. The carpet level will continue to rise as random marking occurs around the bearing and may rise to mask the spikes completely. Move to the velocity spectrum for the same point but with amplitude set to linear. Check to see if there are any spikes in velocity at the bearing defect, harmonics of the defect and/or sidebands of the defect - look particularly for the third and fifth harmonics. If any spike at these frequencies exist then there is physical spalling of the race. If the amplitude of the spike reaches 1 mm/s then the spalling is severe. In the velocity spectrum the following patterns indicate the associated defects:
Dominant Frequency Secondary Frequency Harmonics of Dominant Frequency Sidebands Dominant Direction Defect Suggested Maximum Amplitude @ Dominant Frequency 6 mm/s 3 mm/s 4 mm/s 0.5 mm/s 5 mm/s 7 mm/s 1 mm/s 5 mm/s

6)

7)

1x 1x 1x BPFO / BPFI 7,200 cpm Any Gear Mesh 2x Belt

Nil 1/2 or 1/3 x 3x 1x Rotor Bar Any 1x 1x

Nil Multiple 2 or 3 Multiple 2 or 3 Nil 3 or 4 2 or 3

Nil Nil Nil 1x / Cage 2xLF of RBF Nil 1x N/A

Radial Radial Axial Radial Any Any Radial Radial

Imbalance Looseness Misalignment Bearings Electrical Resonance Gearing Belts

If there is a significant defect raise a work request.

104

Vibration Analysis on Rotating Equipment

Severity charts

Figure 95 General severity chart for vibration
Figure 95 shows a general severity chart for vibration which is in widespread use. There are many versions of these charts. The best use for these charts is for new or rebuilt equipment acceptance limits. Dependency on these charts can be confusing, especially if the bedplate is flexible or on resilient mounts.

105

Vibration Analysis on Rotating Equipment

Two Channel Analysis
Two channel functions
Two-channel analyzers offer additional measurements such as transfer function, cross-spectrum, coherence and orbit. These measurements are discussed below.

Transfer Function
The transfer function is the ratio of the spectrum of channel 2 to the spectrum of channel 1. For the transfer function to be valid, the input spectrum must have amplitude at all frequencies over which the transfer function is to be measured.

Cross Spectrum
The cross spectrum is defined as: cross spectrum = FFT2 conj(FFT1) The cross spectrum is a complex quantity which contains magnitude and phase information. The phase is the relative phase between the two channels. The magnitude is simply the product of the magnitudes of the two spectra. Frequencies where signals are present in both spectra will have large components in the cross spectrum.

Orbit
The orbit is simply a two dimensional display of the time record of channel 1 vs. the time record of channel 2. The orbit display is similar to an oscilloscope displaying a "Lissajous" figure.

Coherence
Coherence measures the percentage of power in channel 2 which is caused by (phase coherent with) power in the input channel. Coherence is a unit-less quantity which varies from 0 to l. If the coherence is 1, all the power of the output signal is due to the input signal. If the coherence is 0, the input and output are completely random with respect to one another. Coherence is related to signal to noise ratio (S/N) by the formula: S/N = γ2/(l -γ2) where γ2 is the traditional notation for coherence.

Correlation
The two channel analyzer may also compute auto and cross correlation. Correlation is a time domain measurement which is defined as follows: Auto Correlation(τ) = ∫x*(t)x(t-τ)dt Cross Correlation(τ) = ∫x*(t)y(t-τ)dt where x and y are the channel 1 and channel 2 input signals and the integrals are over all time. It is clear that the auto correlation at a time “t” is a measure of how much overlap a signal has with a “delayed-by-t” version of itself, and the cross-correlation is a measure of how much overlap a signal has with a “delayedby-t” version of the other channel. Although correlation is a time domain measurement some analyzers use frequency domain techniques to compute it in order to make the calculation faster.

106

Imagine what something 0.Vibration Analysis on Rotating Equipment Spectrum Although the linear magnitude scale is used most often for displaying spectra.) 107 . The phase is relative to the pulse of the trigger. The phase is displayed in degrees or radians on a linear scale. This avoids the messy phase display associated with the noise floor. phase measurements are only used when the analyzer is triggered. usually from -180 to +180 degrees. its phase extends over the full 360 degrees.s. This avoids the problem of taking the log of negative voltages. The phase of a particular frequency bin is set to zero in most analyzers if neither the real nor imaginary part of the FFT is greater than 0. Phase In general. If we wanted it to be 1 inch high on the graph. (Remember.01% of full scale would look like on a linear scale. The real and imaginary parts are always displayed on a linear scale. Why is the Log Mag display useful? Remember that the 16 bit analyzer has a dynamic range of about 90 dB.). even if a signal is small.It turns out that the log display is both easy to understand and shows features which have very different amplitudes clearly. The Log Mag display graphs the magnitude of the spectrum on a logarithmic scale. below full scale. another way of displaying amplitude is the Log Magnitude. the top of the graph would be 833 feet above the bottom .012% of full scale (-78 dB below f.

the horizontal projection of which is the harmonic motion.2. For complex signals this is often much more convenient. For numerical addition. There exists a simpler method of handling the vectors numerically by employing imaginary numbers. We do not necessarily need to make much use of this number but we do need to know where it comes from.3 etc.Vibration Analysis on Rotating Equipment Advanced functions Representation by complex numbers Sometimes the representation of the spectrum is carried out by the use of rotating vectors instead of sine waves. Consider the vector below. however the vectors must be resolved geometrically which loses almost all of the advantage of ease. With the notation j = √-1 these imaginary numbers are j. A complex number can represented graphically by a point in a plane where the real numbers 1. Harmonic motions are represented by rotating vectors. written in the form “a(cos ϖt + j sin ϖt)” represents that motion. But this horizontal projection is also the real part of the vector. are plotted horizontally and the imaginary numbers are plotted vertically. The first use that we will make of the imaginary number is the Nyquist chart which is useful for identifying resonances. So if we say that a “vector represents harmonic motion” what we really mean is that the horizontal projection of the rotating vector represent that motion. a ϖt 0 b φ a+b Figure 96 Vector addition of 2 vibrations The parallelogram rotates at the vibration frequency ϖt so the two vibrations must be at the same frequency for this representation to work. Almost all of the algorithms in the analyzer which involve phase make reference to the imaginary number. Similarly if we say that “a complex number represents harmonic motion” we imply that the real part of such a number. etc. j3. A substitution of the variable angle ϖt for the fixed angle of the vector (α) leads to a(cos ϖt + j sin ϖt) representing a rotating vector. j2. 108 .

Once per rev pulse trigger As the name implies this is a simple pulse. Encoder trigger The encoder will give many pulses per rev and must be input to the external clock input. or for Bodé plots. once per rev of the shaft.Vibration Analysis on Rotating Equipment Cascade & waterfall plots Figure 97 Cascade of fan over 20mS There is quite a lot of confusion over the terms cascade and waterfall plots in the vibration world. Ensure that the data collection rate is valid for the necessary analysis. 109 . A many pulse per rev trigger requires an external clock input to the collector which is programmed to the number of pulses per rev. This could be a very high speed machine over a short time duration or a slow speed machine over a very long time period. which triggers data collection at a specific time. Triggering As mentioned previously a trigger may be set up on a machine to control the acquisition of data. This trigger is often a 5 volt. the collection of data must be considered very carefully before data acquisition takes place. A third type of trigger is a single pulse or voltage change and is known as an event trigger. The OR25 series analyzer is limited to displaying data at the acquired Fmax or 1/10th or 1/100th. orbits. Whichever is the particular application. The advantage of these plots is that we can record data over a certain period of time and see how all of the relevant frequencies are affected. once per rev pulse or it could be from an encoder giving many pulses per rev. This function is useful in cascade plots. A cascade plot is a 3-D representation of the amplitude against speed with respect to time and the waterfall is the 3-D representation of the amplitude against frequency with respect to time such as the example shown above. This input is necessary for torsional vibration and should ideally be used for time synchronous averaging where the speed may be expected to change from moment to moment.

110 . The display is set so that the output of one transducer is set to one (X) axis while the output of the other sensor is on the other (Y) axis. The displacement sensors are mounted orthogonally which is so that each sensor is 45° from the vertical. the total motion of the shaft's orbit within the bearing. A post-trigger will collect data only after the event. then some data collected before the trigger event will be collected as well as some data after the trigger. If the trigger is set up as a pre-trigger. When the trigger is activated the data storage will commence. Figure 98 Bode plots Orbits Displacement sensors measure the amount of shaft displacement. The trigger level is set and the analyzer just sits there waiting for the trigger amplitude to reach the predefined level. In the first example we see an elliptical orbit which is typical of preload caused by misalignment. The analyzer is actually collecting data all of the time but is only storing as much as the memory will allow. they can be displayed as amplitude or phase against speed (Bodé) plots or real-imaginary (Nyquist) plots. The Bodé plot is useful when assessing the resonant condition of a machine or even bearing deterioration as in the example below. This can be very useful if the subject machine is suffering from a transient fault.Vibration Analysis on Rotating Equipment Event trigger This is probably the simplest trigger and may be setup as a channel reading or an external trigger. Bodé plots When the cross channel properties are calculated by the analyzer. or rather.

basically anything that can influence the "system" mass. and sometimes the shape. Imagine a 111 . synchronous orbit analysis.. CRACKED SHAFT A crack in a rotor. synchronous orbits with the phase angle superimposed on the orbit allows rapid identification of this condition. This fault is diagnosed easiest using filtered. a change in this frequency may indicate the "possibility" of a crack existing. A similar diagnosis can be made by viewing the filtered signals from two orthogonally mounted Eddy Probes sensors as orbits. also. By superimposing the phase angle input signal onto the orbit a shift in this parameter can be easily determined.peak amplitude exceeds acceptable limits. As the balance condition deteriorates the size. and/or a change in the resonance frequency as the machine starts or stops. The "possibility" must be emphasized and carefully analyzed because many other causes can produce these changes. of the orbit will grow larger until the peak-to. This can be accomplished by viewing the spectra from any single Eddy Probe sensor. LOOSE ROTATING PART A loose rotating part can generate unusual vibration signals. They may fluctuate in amplitude and the phase angle may shift. damping. foundation problems. By noting the operating speed at which the resonance frequencies occur. and/or stiffness.. a change in the operating phase angle. a damaged or loose bearing support. or shaft.Vibration Analysis on Rotating Equipment Figure 99 Orbit showing misalignment BALANCE Diagnosis of a degrading balance condition is performed by concentrating on the synchronous amplitude which coincides with the rotor speed. such as. loose rotating parts. can generate several different effects on how the machine behaves: a change in the vibration level. but observing filtered. Changes in the filtered amplitude can be determined using orbits analysis. Spectral analysis can be used to identify this fault.

the orbit will appear to have two phase marks on it. As the machine accelerates the whirl frequency will increase as machine speed increases. The orbit will appear to vibrate slightly as this condition is created. or critical speed. This is observable using a spectrum analyzer.Vibration Analysis on Rotating Equipment mass. This display will be round in shape and the amplitude will greater that the amplitude noted during oil whirl. distinct signal displays for each condition. such as an impeller. OIL WHIRL Oil whirl and oil whip are sometimes listed as a single machine fault. The orbit will be more or less round in shape with an amplitude that nearly approximates the bearing clearance. Generally. appears to have several phase marks. The preload may be internally or 112 . The size of the orbit be will larger because the shaft uses up the entire bearing clearance as an oil wedge can no longer be established by the rotor and the shaft is in direct metal-to-metal contact with the bearing. An improperly designed bearing is the usual source for oil whirl conditions. The orbit display will no longer rotate because the oil whirl frequency has coincided with the first natural resonance. synchronous orbit. also. EXCESSIVE PRELOAD All journal bearing machines have some amount of preload so that a stable oil wedge can be established. and when the phase angle is superimposed upon the display. The inception of a loose part condition will produce a "nervous" filtered. but a change in the fluid viscosity or machine alignment state are other possibilities. synchronous orbit produces a distinctive display. it can rotate freely on the shaft independently. but the changes may be too rapid for the sampling rate of the instrument. depending upon the machine/bearing design or construction. The phase shifting can. the part may be slipping and then sticking on the shaft just prior to becoming a full fledged loose rotating part. As the loose part rotates it influences the balance condition of the rotor which appears as a cyclical increase and decrease in the synchronous amplitude. This characteristic is due to filtering at shaft speed and the fault being generated at a sub-synchronous frequency. be observed using an oscilloscope. which has come loose. The display. with the phase input superimposed on the display. This phenomenon creates an individual sub-synchronous frequency which can occur within a frequency range from 35% to 48% of rotor speed. Oil whip is a dangerous condition because the rotor uses up the entire bearing clearance and is in direct metal-to-metal contact that will wear away the bearing rapidly and destroy the rotor if not corrected. Observing oil whirl as a filtered. and has "locked" onto this frequency. but closer observation of the vibration signals and the machine conditions causing these signals will produce different. OIL WHIP Oil whip occurs during the later stages of an oil whirl condition and it has a distinctive orbital display. an oil whirl condition precedes an oil whip condition. The two phase marks will not be displayed symmetrically on the orbit because the whirl frequency is not at exactly ½ machine speed. Spectral and orbit analysis can be used to identify either condition. An oscilloscope set up to observe a filtered orbit will sample continuously so that the changes can be seen. This fault is caused by a condition which prevents the rotor from creating a stable oil wedge on which ride.

RUB A common problem in newly rebuilt or modified rotors is a slight rubbing condition as the rotor is initially operated. the fundamental rub frequency will be ½ shaft speed with multiples at 1X. etc. Spectra displays of rub conditions are characterized by distinct frequencies that occur at multiples of a fundamental frequency. As the rub gets heavier the orbit will flattened and may appear as an excessive preload. etc. External preloads may be from coupling misalignment or piping and support system thermal changes. with the point of the tear drop coinciding with the impact spot. 5/2X. After the orbit has been flattened into the "banana" shape a 2X frequency is present on spectra displays. if not corrected. The shape of the orbit display will differ depending upon the relationship of the shaft speed to the first natural frequency and the severity of the rub. 2X. the fundamental rub frequency will coincide with the shaft speed with multiples at 2X. they usually increase the clearances until the rub has been cleared or. Internal sources of preloads are from gear meshing or hydraulic loading during pumping actions. The severity of the rub will affect the shape of the orbit. 4/3X. 2X. the fundamental rub frequency will be shaft speed with multiples at 2/3X. etc. As the preload increases the orbit is further flattened. As excessive preload increases further the orbit begins to collapse to form a "banana" shape as the shaft tries to continue its normal rotation pattern and direction. 5/3X. 7/3X. Between three and four time the natural resonance frequency. At higher machine speeds (above twice the first natural frequency) the unfiltered orbits will begin to have internal loops with the fundamental rub frequency inversely proportional to the number of internal loops. These internal loops will have their own phase marks displayed and the loops will be located symmetrically on the display. 3X. 3/2X. they will wear away the internal clearances until the machine cannot be operated. 113 . Heavy preloads further distort the orbit into a figure eight shape. These sources of preload create an elliptical orbit that is flattened in the direction of the preload vector. Between twice and three times the first natural resonance frequency. Rotor rubs are not a phenomenon which continues over an extended period. As preload increases the shaft centreline will shift in the direction of the preload vector. The fundamental frequency will depend upon the relationship of the shaft speed to the first natural resonance frequency. At shaft speeds up to twice the natural resonance frequency. A light rub will produce a "tear drop" shaped orbit. 1X.Vibration Analysis on Rotating Equipment externally produced.

This is not the case. however. The vibration analyst running the average predictive maintenance system knows that most of his machinery failures are from the “bread and butter” causes listed above. a vibration is encountered which does not fall into any of the above categories and the vibration analyst is left scratching his head. The purpose of this course is to help the in-house vibration analyst identify and cure vibration resonance problems without having to call in outside assistance. the number of teeth on the gears. Eventually. In many cases the big spike is related to a resonance of a natural frequency or perhaps running a machine near a critical speed. Most failures are due to material fatigue from normal operation leading to mechanical failure. The vibration consultant. He knows the bearing numbers. is usually only called in when the in-house analyst is having trouble identifying a troublesome vibration and these problems usually are caused by resonance. on the other hand. Many consultants in the field of vibration analysis will tell you that 90% of all failures are directly caused by vibration resonance.Vibration Analysis on Rotating Equipment Introduction to resonance The majority of vibration analysis programs concentrate on detecting dynamic machinery faults such as: • • • • Bearing defects Gearing defects Misalignment Out-of-balance. the belt and pulley dimensions and still none of the pre-calculated defect frequencies line up with the big spike in the frequency spectrum. He knows the complete structure of the machine. the speeds. By the end of this course you will be able to: • • use several techniques to identify resonant conditions recommend appropriate measures to cure the resonance 114 .

Each small push (as long as it is large enough to overcome friction or damping) will add to the amplitude of the swing. You are the adult who pulls the child back and releases the swing so that the child swings back and forth. Natural Frequency The natural frequency of a system is that frequency at which the system will oscillate or vibrate when excited by a single external exciting pulse. misalignment. The natural period of oscillation of a pendulum is calculated using the following formula: 115 . repeatedly applied force causes the amplitude of an oscillating system to become very large is called resonance. If the exciting force is not large enough to overcome the damping. Fortunately for us. which is the frequency an oscillating object tends to settle into if it is not disturbed. Mechanical For example. the natural frequency of a pendulum 1 m (39 in) long is 0. gear mesh. The rate at which the child swings is the NATURAL FREQUENCY of the child/swing system. there is little damping then the interaction of an exciting force with a natural frequency can give rise to very large amplitudes of vibration resulting in heavy load cycling and eventual failure of the material. Let us consider the pendulum mentioned above. Such forces include imbalance. When we say a system. In fact.Vibration Analysis on Rotating Equipment What is resonance? Imagine a child on a swing. Every time that the child swings back to your position. Any oscillating object has a natural frequency. NATURAL FREQUENCY + EXCITING FORCE = RESONANCE In the world of maintenance the exciting forces are from a myriad of sources with the strongest usually being related to run speed. we can see a host of interconnectivity which can give rise to a large number of natural frequencies. The damping acts like the friction in the child/swing system. then the amplitude at that frequency will diminish. anything which can be measured as a vibration may be considered to be an exciting force. The phenomenon in which a relatively small. we mean any part or component (or combination of either) of the machine and/or the structure. If. however. The problem we have is that we can easily determine the exciting force frequencies but cannot always tell which system has it’s natural frequency being excited. If the pendulum is struck lightly once every 2 seconds. the amplitude of the swing increases gradually until the amplitude of oscillation is very large. bearing frequencies etc. In this case the system is not in resonance with the exciting force. most of the systems have inherent damping. you give the swing a small push – you are the EXTERNAL EXCITING FORCE. even though there may many natural frequencies. In this case the child/swing system is in RESONANCE with the exciting force. which means the pendulum swings back and forth once every 2 seconds. Looking at the average machine and associated structure. You time the small pushes to coincide with the rate of swing of the child – in other words the exciting force is at the same frequency as the natural frequency of the system. Sometimes the pushes would add to the amplitude of the swing but sometimes the pushes would reduce the amplitude of the swing. Imagine what would happen if you pushed at a different frequency.5 Hz.

The main influences on the natural time period are the length of the pendulum and gravity. g 2 1 0 Equation 1 Simple pendulum time period Figure 100 Simple pendulum 1 2 2 x .Vibration Analysis on Rotating Equipment π L 4. . we have to consider all aspects of the system that affect the natural frequency. such as the one shown below. To calculate the natural frequency from the time period we simply divide the time period into one second. In this case the time period is: f Hz 1 Ts = 1 Ts Equation 2 Calculating frequency from time period In a conventional mechanical system. 116 . sin( φ ) dφ = Notice that the time period of the swing is affected slightly by the angle of the pendulum but the effect is only slight.

s u .Vibration Analysis on Rotating Equipment The main three components in the system shown are: Figure 101 Mass/Spring system • • • Mass (Spring) Stiffness Damping Let us look at the spring/mass system in a little more detail. and this could be as simple as someone pushing and pulling the mass with their hand. if the spring force has magnitude kx.t 1. Newton's second law states that: Force = mass x acceleration Equation 3 Newton's 2nd law So at any instant we can determine the acceleration of the mass. the spring/mass system will move vertically up and down at a rate which is determined by the spring rate. Think of a spring with a mass of a certain weight hanging off the end. because the damping is always in the opposite direction to the velocity. the weight and gravity. 2 (a t) 2 Equation 4 Newton's law of motion (x) (x2) (x3) Where s = displacement u = velocity a = acceleration t = time In the single degree of freedom system such as the mass/spring system shown. being "-cx2". If we pull the mass down and then release it. and the extension "x" is zero then there is no force. Note: Po sin ωt is a combination of direct force "Po" and angular position "sin ωt" where ω = rotational velocity & t = time 117 . The sign of the spring force is negative as it is always acting in the opposite direction to the extension. and by applying Newton's laws of motion we can determine the instantaneous velocity and displacement. So: Spring force = "-kx" The damping force is also negative. The forcing mechanism is defined harmonically as Po sin ωt.

2. For example a spring is a good example of a low damped system. Speed of sound in various liquids (m/s) Table 1 Speed of sound in liquids Liquid water methyl alcohol Benzene castor oil Glycerin Velocity m/s 1497 1123 1326 1500 1923 118 . Liquids & pumping systems Pumping systems in general and hydraulic systems in particular are often associated with hydraulic resonances. In order to have a resonant condition we need to have the right combination of stiffness and mass to create the natural frequency and a source vibration that matches the natural frequency. The damping qualities of a structure control the resonant response. Sonic vibration and Pulse vibration Sonic vibration To think of a hydraulic resonance. The damping controls the amplitude response. You can easily see the three components of mass. acceleration velocity and displacement respectively. There are two different types of vibration in liquids: 1. consider a pipe filled with liquid. a vibration will travel through the liquid in the pipe and the speed at which the vibration travels will be at the speed of sound. damping and stiffness directly affect the amplitudes of oscillation in. If we impact one end of the pipe. Sound travels at different speeds in different liquids but in fresh water it travels at roughly five times the speed of sound in air.Vibration Analysis on Rotating Equipment The forces acting on the mass are: -kx –cx2 + Po sin ωt Newton's law gives: d2x m -----. A shock absorber is a good example of a highly damped system where a large input results in a small response. a relatively small amplitude of input results in a large amplitude response. The higher the damping effect the lower the amplitude of vibration will be as a result of a resonant condition.= mx3 = -kx –cx2 + Po sin ωt dt2 Equation 5 Differential equation of motion of a single-degree-of-freedom system or mx3 + cx2 + kx = Po sin ωt This equation is known as the "differential equation of motion of a single-degree-of-freedom system".

The exact relation between the velocity of sound. V(water) = √(bulk modulus/ mass density) The speed of sound in water is approximately 1500 m/s. S salinity in parts per thousand z is depth in meters Now. The speed of sound decreases as the mass of the molecules increases because the mass of the molecules is related to the force per acceleration of the molecule. V(solid) = √(Young's modulus/ mass density). wavelength & frequency where c = speed of sound in the material f = resonant frequency λ = length of pipe (or direct sub-multiples) If this frequency coincides with a run speed or vane pass frequency then a serious resonant situation could occur.055T^2 + 0.00029T^3 + (1.34-0.6T . V. However. The speed of sound in an ocean is approximately: c = 1449. if we know how fast the impulses travel through the liquid.01T)(S-35) + 0. The speed with which the sound wave or sound energy travels depends on the interaction between neighboring molecules of the substance. For sound waves in water.016z Equation 7 Speed of sound in the ocean T temp in degrees Celsius. V(air) = √ (bulk modulus/mass density) For longitudinal waves in a solid. the speed of sound increases as the strength of the interaction between the molecules increases. we can determine the resonant frequency of oscillation for a certain length of pipe: c = f ⋅λ Equation 8 General formula relating speed. It is possible to measure changes in ocean temperature by observing the resultant change in speed of sound over long distances. Equation 6 Velocity of sound in materials For sound in air.Vibration Analysis on Rotating Equipment Mercury air (reference) liquid helium 1440 320 240 The sound waves created in a substance travel with a certain speed through the substance.2 + 4. and the properties of the substance through which the sound travels is given by: V = √(elastic property/inertial property).0. 119 .

vibrates simultaneously as a whole and in equal parts (halves. it pushes liquid in front of it because most liquids are virtually incompressible. quarters. 4:1. The impulsive frequency can also excite any mechanical resonances in the system from pipework or supporting structures. giving the player a new fundamental wavelength with a new harmonic series. for example: • • • • Vane pass Piston pulse Hydraulic valve operation Impeller eccentricity As each hydraulic pulse enters the piping system. The harmonic series for the tone C is given in the accompanying notation. The usual outcome is a very noisy operating environment. and the musical pitches closer together. Figure 102 Harmonic series for the tone C. such as the bugle.Vibration Analysis on Rotating Equipment Impulsive vibration The other type of vibration in liquids is from the pulses generated by the system itself. the player isolates and makes predominant one of the higher harmonics. others. 3:1. producing wave frequencies that are in simple ratios with the fundamental frequency (2:1. we must first know the speed of sound in air. which is usually perceived as the basic pitch of the musical sound. particularly if they are both at the same frequencies. on a clarinet. and so on). In musical instruments harmonics are a series of subsidiary vibrations that accompany a primary. and so on). In “over-blowing” a wind instrument. the valves of the trumpet and the slide of a trombone add extra tubing. Harmonics contribute to the ear's perception of the quality. In musical sound the full-length vibration produces the fundamental tone (or first harmonic or first partial). The shock wave from this pulse then travels right through the system giving system-wide excitation at that impulsive frequency. to calculate the resonant frequency in a pipe. The subsidiary vibrations produce faint overtones (second and higher harmonics or partials). the vibrating segments become smaller. let us look at the historical use of air resonance – music. The harmonics of string players are flutelike tones produced when they cause the string to vibrate solely in halves (or thirds. thus extending the range of the instrument upward. the frequencies higher. It has been the authors experience that heavy vibrations on pumping systems are usually caused by an interreaction of these impulsive and sonic vibrations. The subtle timbre of a particular instrument is caused by the particular frequency in resonance and several harmonics and/or sub-harmonics of that frequency. thirds. of a sound: On a flute. 120 . certain harmonics of the series are most prominent. but mechanical failure often results. for example. or fundamental. As the series progresses. They result when the vibrating body. Air & gases Before we get too deep into the vibratory aspects of air-borne sound for vibration analysis. Unvalved brass instruments. produce only the tones of the harmonic series. and so on). a stretched string or an enclosed air column. for example. the instrument can thus produce more tones. wave-motion vibration. black notes show pitches that do not correspond exactly with the Western tuning system. As in liquids (described above). or timbre.

Just as in the liquid pipe. R is the ideal gas constant. 1. will be reflected back from the closed end.2 to 1.40 for diatomic gases. a condition of constructive interference exists which we call resonance and the loudness of the sound will be greatly increased. 1.Vibration Analysis on Rotating Equipment The Newton-Laplace equation for the speed of sound in a gas is v = √( (γ P) / ρ) γ Equation 9 Newton-Laplace eq. The speed of sound in air at a temperature of 0°C (32°F) and 50% relative humidity is 331. m is the mass of one molecule. and vice versa. Higher modes of resonance exist when the wavelength is 4/3(third harmonic).. The lowest frequency where a condition of resonance exists in the tube occurs when the wavelength is four times the length of the tube.. of the length of the tube. The speed is proportional to the square root of absolute temperature and it is therefore about 12 m/s greater at 20°C (64°F). and in the range of Applying the Ideal Gas Law. γ is the specific heat ratio. The first four resonance modes for tubes open at both ends.6 m/s. the equation for the speed of sound in a gas becomes v = √ (( γ k T) / m) = √ (( γ R T) / M) Equation 11 Speed of sound in a gas where v represents speed. The displacement of the air will be greatest at the open end of the tube. k is Boltzman's constant. whereas the pressure difference will be greatest at the closed end. γ = ratio of the specific heat of the gas at a constant pressure over the specific heat at a constant volume.30 for triatomic gases.67 for monatomic gases. sound waves which travel down the tube. γ = cp over cv Equation 10 Ratio of specific heats (gamma) ≈ 1. molecular weight. but not on the pressure of the gas. The speed is nearly independent of atmospheric pressure but the resultant sound velocity may be substantially altered by wind velocity. This is called the fundamental mode of vibration. and M is the molecular weight of the gas. T is the temperature in Kelvin. 1. for the speed of sound in a gas where and P =pressure ρ = density. 4/5(fifth harmonic). and molecular structure.. and closed at one end are shown below: 121 .1 for polyatomic gases.8 If a tuning fork is set in vibration and held over the open end of a tube closed at one end. If the returning wave is exactly in phase. we can calculate the natural frequency of the air inside a pipe by using the formula: c = f ⋅λ Eq. The speed of sound in a gas depends on the temperature.

5v/4f and so on for a tube closed at one end. and the end correction may be calculated from 1/2(L2 . The acoustic length of the tube is equal to its physical length plus the end correction. and tuning forks of frequency about 500 Hz. If the positions of resonance from the open end of the tube are L1. From these relations the wavelength may be calculated from 2(L2 . If the frequency is fixed. It is actually a small distance beyond it. The bridge began to sway violently when the wind blew steadily at a certain velocity. and L3. as in the case of the tuning fork held over the open end of a tube. Karman Vortices Everyone knows the story of the Tacoma Narrows suspension bridge.3L1) or 1/2(3L3 . 2v/2f. 3v/2f for a tube open at both ends. 122 . 3v/4f. and the value of the end ⋅ correction may be determined from the relations above.L2) or (L3 . where e is the end correction.L1). Since the frequency of the tuning fork is known. This extra distance beyond the end of the tube is called the end correction . The mechanism for the resonance-destruction of the bridge was the same mechanism that causes a flag to flutter – Karman Vortices. the displacement is a minimum (node) at the closed end. When resonance exists. but the antinode is not exactly at the open end. and v/2f. In a typical laboratory experiment with a water reservoir-resonance tube apparatus about 1m long.Vibration Analysis on Rotating Equipment Figure 103 Sonic vibration in a tube The antinodes (A) are displacement antinodes (pressure nodes) and the nodes (N) are displacement nodes (pressure antinodes). L2. the wavelength of the sound wave is equal to 4(L1 + e) or 4/3(L2 + e) or 4/5(L3 + e).5L2) or 1/4(L3 .5L1). three positions of resonance can usually be found.L1) or 2(L3 . the resonance lengths are v/4f. both the velocity of sound may be determined by using the relation c = f⋅λ.

perpendicular to the air flow. any shaft rotational speed which is associated with high vibration amplitudes.Vibration Analysis on Rotating Equipment Figure 104 Karman vortices As the air passes over the object a vortex will form on one side of the object and down wind. Critical Speed (Balance Resonance) Critical speed is. There is a distinct relationship between the frequency (f) of the force. The vortices for the wind at this speed developed an alternating force which was exactly at one of the natural modes for the complex structure of the bridge. f * D = 0. The vortices are alternately clockwise and anti-clockwise and are shed from the object in a regular manner resulting in an alternating force on the object. the speed which corresponds to a rotor lateral mode resonance frequency excited by rotor unbalance. 123 . in general. a steady wind blew along the valley onto the side of the bridge. As the wind hit the side structure of the bridge it developed Karman vortices. The increased pressure will cause a deflection of the air flow so that another vortex is formed on the other side of the object just as the first vortex dissipates.22 V Equation 12 Karman vortices Figure 105 Tacoma Narrows Bridge In the case of the Tacoma Narrows bridge. the velocity of the wind (V) and the width or diameter (D) of the object. in which case it is more correctly called the balance resonance speed. Often.

The resonance speed is at the point which is 90° phase away from the angle of the heavy spot for that resonance mode.Vibration Analysis on Rotating Equipment Figure 106 Graphical identification of critical speed Rotor balance resonance speed is the shaft rotational speed (or speed range) which is equal to a lateral natural frequency of the rotor system. damping.48X). A diagram used in rotating machinery design. oil whirl (. The term is sometimes used incorrectly to describe the cascade plot and waterfall plot. This may be slightly different than the rotational speed peak amplitude point.. blade or vane passing frequencies. The X axis represents the various possible excitation frequencies. the observed vibration characteristics due to rotor unbalance are: 1) a peak in the 1X amplitude and 2) a rapid change in the 1X vibration phase angle. A tool for selecting and checking shaft operational rotational speeds and other possible forcing function frequencies against the spectrum of natural frequencies to avoid resonances. 124 .. and/or asymmetry in the system stiffness. gear mesh frequencies. These changes may not happen at the same frequency due to non-linearity. rotational speed (1X). When the speed increases or decreases. Campbell Diagram.40X to . The Y axis represents the rotor lateral and torsional natural frequencies.e. etc. i.

Vibration Analysis on Rotating Equipment Figure 107 Campbell Diagram 125 .

usually at its point of connection. Cracking almost always occurs at one of the mode shape nodes. lost lubricant can wreck an expensive machine or a flammable product can be released. This could account for so many familiar situations whereby pump shafts “mysteriously” break within a few weeks to a few years.” The break usually shows a crystalline appearance and sharp edges. gear driven shafts broken with the characteristic 45° torsion break. pipe weld breakage (often repetitively). radial lines). Premature foundation or floor cracking. there is no problem. is resonant to a specific frequency of machine vibration. However. The phenomenon is similar to bending a wire back and forth until it breaks. resonance should be suspected. But sometimes the pipe. It may take many months or even years for the tube or pipe to crack at one of the nodes. it can be determined if there are tell tale large “loops” (antinodes) and nodes. even though the calculated stresses indicate the shaft should last a lifetime without breaking. Sometimes rotating machinery does not show excessive vibration at the usual points of measurement. When hand feeling a component or structure reveals a possible resonant situation further confirmation should be obtained by plotting out the mode shape on graph paper. The break has the characteristics of a pure fatigue failure which is often mistakenly called “crystallization. but when it does. cooling water or instrumentation piping. characteristic of a resonant condition. have all occurred as the result of metal or concrete fatigue at one of the nodes due to resonance. by feeling with fingers along the small diameter. lubrication. Most often the vibration frequency is high enough and the tube or pipe rigidity low enough to allow a higher resonance with several nodes. Remember to 126 . If non-resonant. The material does not break due to excessive stresses but due to the many millions of reversed stresses concentrated about the nodes. This simple procedure can be performed with the most rudimentary of instruments and is very effective in identifying the mode shapes. causing fatigue. such as at the bearings. gear cracking or failure (usually in equal-spaced. Eventually it will break. Identifying resonance in mechanical systems Mode Shape Figure 108 Mode shape node points Whenever cracking of the machine or structure is reported. The same phenomenon is very common for pressure gauges cantilevered on a pipe. with a gauge acting as a weight. While it is unlikely that the vibration on the tubing will cause vibration on the machine the problem should still be addressed to prevent failure of the tube.Vibration Analysis on Rotating Equipment .

6. The 180° relationship will remain the same for all frequencies above the resonant frequency range. This phenomenon is shown in the 127 . On a horizontal zero line drawn on a sheet of graph paper mark the measurement locations spaced 1” or 25mm apart. Divide the component to be tested into a number of roughly equal parts and identify each location. Displacement units are preferred but velocity or acceleration can also be used. On piping sections the direction indicating the greatest “curl” is the resonant direction. Measure the vibration amplitude at the frequency of interest at each location. the force and its resulting deflection will move in the same direction at the same time.” and if the frequency of that vibrating force is below the critical speed range or resonance range. 4. then the force and resulting deflection will remain in phase. If a vibrating force is applied to the same object or “spring system. When the vibration frequency reaches the actual critical speed or resonance frequency. The procedure for performing the mode shape plot is as follows: 1. Join the dots “curve fit” if required. the deflection will time-lag the force. However. Divide each reading by 2 (for displacement).Vibration Analysis on Rotating Equipment plot both directions (horizontal and vertical). Figure 109 Mode shape readings Phase If a force is slowly applied to a spring system. 1 cm = 25 microns. Then the process starts all over again with another phase angle reversal. until the second resonance frequency range is approached. 5. the force increases its lead until the force finally precedes the resultant deflection by 180° . 3. and so on. usually with a letter to avoid confusion. or “in phase” with each other. 2. the force will precede the deflection by 90° . Plot the results above and below a horizontal zero line on a sheet of graph paper using a scale of ½ “ = 1 mil of vibration. when the vibration frequency enters the resonance frequency or the critical speed range. As the frequency increases further. Overall readings can be used but filtered data is preferred.

bases. Parts may be temporarily braced to add rigidity and. moved to a higher resonance frequency that is at least 20 to 25 percent away from the source frequency.Vibration Analysis on Rotating Equipment resonance diagram of a rotor’s vibration amplitude and phase relative to resonance of some part. These parts can usually have their resonant frequencies changed while the machine is still running. beams. resonance of the rotor itself. such as pipes. therefore. such as the rotor support system. part of a steel base or beam. yet the phase change method to determine whether a resonant condition exists can still be used. or more rarely. 128 . Remember that almost all troublesome resonances occur in nonrotating parts. pedestals. Typically the running machine cannot be shut down or have its speed varied for resonance tests. decks and so on. Figure 110 Phase relationships This phenomenon can be used to help us determine resonance.

Since an object will undergo free vibration at its natural frequency when bumped or struck.Vibration Analysis on Rotating Equipment Figure 111 Phase / frequency relationships @ resonance The bump test Another simple yet effective way to confirm whether or not a part is vibrating in resonance is the “bump” test. if the part tested is connected to another part that is relatively flexible. however. Bump tests. One method of determining the natural frequency of a structure is to calculate it. the natural frequency generated in this way will be indicated on the analyzer. The calculations are fairly straightforward for simple structures. Although care should be taken that the frequency at which the "bumps” are struck is not read as a resonant frequency. For the more complex structures found in the plant the calculations are far more difficult and therefore not usually as accurate as those obtained through simple bump or impact tests. simply bump the machine or structure with a force sufficient to cause it to vibrate. With the machine shut down and a vibration pickup held or attached to the machine. Also. are almost always limited to determining the first critical frequency of the part bumped. If the vibration diminishes very quickly it may be necessary to bump the machine several times in succession in order to sustain free vibration long enough to register on the analyzer frequency meter. then it is sometimes difficult to be sure that the resulting vibrations are originating from the part being bumped rather than from the part to which it is connected. 129 . The Real Time Spectrum analyzer providing instantaneous display of the vibration amplitude versus frequency data is an ideal instrument for determining a natural frequencies this way.

The rope will need to be retied. rolls. The peak’s width will also increase if the peak is affected by resonance. Then with the machine still running. giving its resonant frequency in a manner similar to regular bump tests. a “come-along” or even a lift truck or tractor. cut. beams and platform. fan blades. only the frequency that is generated between bumps. 2) Set the Fmax (maximum frequency) to a suitable value higher than the suspected frequency. The structure will spring back. The area where the rope could whip must be cordoned off and all bystanders moved out of harms way. For a very large structure. The low frequency of the repeated bumps is not measured. but will not necessarily work well for the total structure that includes all the vertical columns. Ensure that the transducer weight is less than 10 percent of the system weight. including cast iron or cast steel bearing supports and brackets. columns. A regular impact bump test can be used for the horizontal beam. a horizontal force can be applied in one direction at the top of the structure. tension applied. If the structure is small. but its frequency remains the same. the structure can be deflected as much as a mm or two without harming the structure. Its amplitude decreases with each cycle. The two spectra can then be compared to determine if there is a resonance. pedestals and welded steel bases. Running machine Bump Test Normally. 130 . bump the structure repeatedly.Vibration Analysis on Rotating Equipment Bump tests are also ideal for rotors. Assume that the force is applied to the horizontal beam supporting the platform. Usually a clamp is attached to the top of the structure. Care should be exercised during this type of test as the rope will whip when cut and could cause injury to anyone in its path. Position the pickup and set the vibration instrument for the expected frequency range. If the bumps generate natural frequency peaks that correspond to a peak produced by the running machine (before bumping). tension on the rope deflects the structure by several mils. in certain cases a bump test can be performed while the machine is running. A strong rope is tied to the clamp. A structure such as a platform supported by columns will have individual resonant frequencies for each column and the platform. It takes more time and patience than regular impact bump tests. then the amplitude on the analyzer screen should increase.. also has several flexibilities depending on the direction being considered. However. etc. If peakhold averaging with running display capability is available. covers. Another is taken while bumping the part. take data with the machine running. A running machine spectrum is obtained. but often reveals that the magnification due to resonance is not that of an individual part but instead that of the total structure. such as one supporting a fan or the total frame of a vertical pump. The force will deflect the beam but will also cause the whole structure (including columns) to deflect as well. Repeat the procedure several times to make sure repeat readings are obtained. etc. such as main steel support structures of buildings. They are also practical for relatively rigid parts. With the instrument set up accordingly. such as a tower of several stories. then its use is preferred. Very large columns and beams. This test is based on the principle that when a “spring system” or part is bumped or deflected in some way. a “reverse” bump or pluck test works very well. The procedure for setting up the instrument varies slightly with each instrument but the general procedure is as follows: 1) Mount the transducer on the structure to be tested in the direction that the excessive vibration was recorded. However. pulley. beams. To determine the resonant frequency of the total structure. For example. Now cut the rope. Set up for FFT-type analyzers FFT-type analyzers have various settings which determine the type of spectra obtained. bump tests are performed with the machine shut down. Reverse Bump or “Plucking” the Suspect Part. do not as easily respond to the regular bump test. Tension is applied to the rope using any practical means such as pulling with a turnbuckle. the total structure as a combination of columns and platform. it will vibrate for several cycles at its natural frequency. floors.

7) The structure should now be impacted in the direction of transducer orientation using a soft hammer or piece of wood. If “peak-hold” is not available. This is simply a hammer with a force transducer mounted which will send a trigger signal to start collecting data the instant the hammer hits the structure. However. Instead. When the hammer impacts one part. the peaks will relate to the resonant frequencies of various machine parts that were displaced by the bumping action. If you have an analyzer which can perform negative averaging. then that part may cause the largest amplitude peak. except that the various peaks will not necessarily be related to running speed.Vibration Analysis on Rotating Equipment 3) Set the instrument up for manual amplitude scaling (disable the autorange feature) and select an appropriate amplitude scale. A single blow should cause a response on the instrument and the strength of the blow should be adjusted to give an appropriate response on the instrument. harder hammer heads such as plastic are usually recommended. 131 . For higher expected frequencies. the longer it takes for the instrument to acquire the data. then the number of averages should be minimized. then its use is recommended for the best results. Special windows and trigger settings are available on most data collectors today. using an FFT. although not necessarily in all instances. then a setting of 100 lines may be more appropriate. 4) Set the resolution to an appropriate number such as 400. may not always determine with certainty which peak represents a specific part. representing vibration coming from other machines will have to be mentally subtracted from the bump test results. 5) If the analyzer has the capability to display continuously updated (running) spectra using “peak-holdtype” averaging. Start with a sensitive scale and adjust the scale as required to suit the strength of the impact and the size of the structure. Impact hammer A better way of determining natural frequencies is to use the impact hammer. The amplitude scale and strength of impact should be set to respond to the “ring down” rather than the impact. An accelerometer is mounted on the structure and the resultant output is generated by the natural frequencies of the structure. A single bump test. Typically. If autorange remains active the instrument will tend to range to the initial bump rather than the “ring down” afterwards. then use this capability to remove the background vibration after the bump test. For expected resonant frequencies in the range of normal speed machines. it causes a deflection and release. 6) Take a spectrum and observe the background vibration data (if any). the results are usually acceptable for most situations. If multiple blows are required then random impacts are preferred. A typical spectrum obtained from a bump test will look similar to regular spectra obtained from vibrating running machines. bumping with a rubber mallet is preferred over harder materials such as steel or plastic. Despite the loss of accuracy due to the lower resolution. there is one peak that has a much higher amplitude than the rest. Some condition monitoring software programs will subtract spectra as well. such as 4 to prevent excessive data collection time. If multiple peaks appear then multiple tests should be performed on adjacent structures and in different directions in order to determine which peak relates to which structure. Most modern data collectors have the capability to generate a “live” time display. They can be used but are not necessary for this simplified test. The largest peak usually represents the resonant frequency of the part that receives the bump test hammer’s impact. The more the lines of resolution. If the other part is more flexible or has much less damping. the impact also causes some deflection and release on parts that are connected to the part receiving the hammer blow. Take care that the data collection rate will be high enough to differentiate between the oscillations in the ring-down and not cause them to be smeared. If this is used. These peaks. such as under 5000 rpm. Soft wood (such as a “2x4”) can also be used.

Vibration Analysis on Rotating Equipment Figure 112 Impact hammer response The response can be analyzed with spectrum. For this course we will use medium sledge from IMI. Bode and Nyquist plots to identify resonances. Figure 113 Impact hammer specification sheet For specific information on the use of the impact hammer refer to the hammer operating manual. model 086C41. 132 .

above: Note that if the individual oscillations are smeared into one pulse in the time domain.Vibration Analysis on Rotating Equipment Table 2 Fourier spectra of pulses Care should be taken when defining the spectrum that sufficient resolution is used to stop the oscillations from smearing into one another. then the FFT cannot have relevant frequency domain data. In some cases it is possible to get a quasi-triangular pulse which. If the sampling rate is not high enough to separate out the individual oscillations then the time waveform will take on the form of a “pulse”. when 133 . Consider table 2.

totally meaningless. in fact. 134 .Vibration Analysis on Rotating Equipment passed through the Fourier transform. will give a frequency response which looks good but is.

This section gives guidelines for the calculation of natural frequencies for simple structural elements. The structure can usually be broken down into a number of components which can be calculated quite easily.Vibration Analysis on Rotating Equipment Calculating Natural Frequency in Mechanical Systems At certain times it is very useful to be able to calculate the natural frequencies of certain components of the system – especially the supporting structure. Table 3 Natural frequency calculation of uniform beams 135 . Uniform Beams One of the most common sources of resonant vibration is the structural supporting beam.

This is µ in our calculation above. • • For low carbon steels E (Young’s Modulus) can be estimated at 30.000. calculating a2 will return the frequency for the second mode etc.000 psi. E.natural frequency of beams The difficulty in making this type of calculation is getting the values for E.Vibration Analysis on Rotating Equipment The notation a1 through a5 is for each vibration mode. standard text books such as the “Machinery’s Handbook” have all of the required information for different types of beams..l Equation 13 General formula . calculating a1 for a clamped-clamped beam will return the natural frequency of the first vibration mode. The moment of inertia I may be calculated but it is far easier to lift the value from tables as in the example below: Table 4 Standard values for uniform beams • 136 . I and µ. For example. I 4 µ . The weight per unit length is a part of the beam identifier. Luckily. ωn a 1 . For example a wide-flange section beam designated “W 18 x 64” has a nominal depth of 18” and nominal weight of 64 pounds per foot length. however.

you must convert to the units for which your formula is designed. / in = 22.000. ω n a 1 .Vibration Analysis on Rotating Equipment Notice that there are two values if I.l 4 E µ a1 l = 30. all edges clamped. The way that the example above is drawn./ft = 8. Plates The procedure for calculating the natural frequency of any component is similar to that of uniform beams except that the formula varies from case to case. the section X-X will give us the moment of inertia horizontally and the section YY will give the moment of inertia vertically. designated W18x106. “clamped-clamped” beam. (DOVER 0-486-64785-4) 137 . length of side t.5 = 29. 12 feet long.000 * 220/(22 * 1444))0. the fundamental mode is 36. ωn l D µ1 2 Where And µ1 = mass per unit length D = the plate constant The plate constant (in this case) is defined by 7 Available in Borders bookstore. $12.0 (from table 2) = 12 feet = 144 inches ωn = 22 * (30.000.616 Hz I (section Y-Y) (calculator notation) Notice that we have been consistent with our units – we have only used pounds and inches in the equation.00117 rad/s = 4. For plates we should consider • circular or rectangular plates • point load or uniform load • simply supported edges or clamped edges In 7Hartog’s “Mechanical Vibrations” we have a number of formulae for different applications in the appendix.000 = 220 in4 = 106 lb.833 lb. There is a value for each of the two non-axial directions (vertical and horizontal) so we can calculate the natural frequencies for each of these directions.95. Whenever you are given properties in other units. For a square plate. Let us calculate the natural frequency for a low carbon steel. I µ . E.

138 . the plate constant D will be 4.221 Hz If your plate is clamped at each end but free on the other two sides. t 12 1 3 D µ 2 Where t = thickness (inches) µ = Poisson’s ratio (≈ 0.293 x 104 So that for a square plate. of side 6 feet we would have ωn = 0.3) For a low carbon.Vibration Analysis on Rotating Equipment E.611 Hz If the same 3 foot plate were ½” thick: ωn =1. ¼” thick.108 Hz (Don’t forget to divide radians/second by 2π to convert to Hz) For a plate of equal side = 3 feet ωn = 0. then treat the plate as a clamped/clamped beam. clamped at the edges. steel plate.

thus impairing the isolation result and causing it to have no effect at all after a period of use.5 Hz 240 to 630 cpm Cork 8. Steel springs are normally used in the form of coil springs or leaf springs. The types of spring material which are used most often are rubber and steel. The same applies to felt. The benefit of these is that they permit relatively high deflections. and more often than not special devices have to be installed in order to limit the deflections. which is why it is no longer used in more demanding applications.8 Hz 72 to 228 cpm Rubber 4. The following 8table suggests some appropriate damping ranges for popular vibration isolation materials: Table 5 Damping ranges of vibration isolators Material Steel Springs Effective Damping Range 1.058 mils p-p 0. Vibration isolation is based on installing machinery on springs or resilient material of uniform stiffness. which has more or less the same kind of properties. Because of this.045 mils p-p 0. or Change the natural frequency of the structure or component In the case where it is not feasible to change the natural frequency we must isolate.6 to 30 Hz 1176 to 1800 cpm 0.66 mils p-p Lowest Amplitude Expected 3. Another alternative is air springs. However. the locked-in air is eventually pressed out. excessive movement occurs when running the machine through the resonance range. but their disadvantage is that they provide very little damping.0 to 12.0 to 10. Cork was formerly used as a "spring" material and its elastic properties were based on compression and expansion of the air locked inside the cork.2 to 3.5 mils p-p It should be noted that these ranges are for virgin materials and many manufacturers offer specialist materials which cover much wider ranges than shown on this list.Vibration Analysis on Rotating Equipment Detuning Resonant Structures To take any structure out of resonance we must do one of two things: • • Isolate the structure or component from the exciting force.6 Hz 480 to 756 cpm Felt 19. 8 Taken from information prepared by Larry Riley (3/6/96) 139 . Cork does not permit a large spring travel. Vibration isolators This section explores some of the isolation materials that we can use.

Figure 115 Stress/strain diagram for rubber The tensile sequence for rubber is quite different. The stress-strain diagram illustrates clearly the difference in the strain properties of rubber and steel. the modulus of elasticity is not constant. has good elasticity and chemical resistance.Vibration Analysis on Rotating Equipment Figure 114 Stress/strain diagram for steel Compared with steel. Consequently.1 %). rubber does not "stretch" proportionally to the load. Springs Springs have long been a favorite with fan manufacturers for separating fan vibration from the floor. The maximum tensile force is reached at the ultimate strength. and these are properties which make rubber particularly suitable as a spring material. Steel is elastic up to the elastic limit and restores to its original shape (this value is less than 0. and in this range Hooke's law applies. but the stress increases continuously until there is abrupt failure. and the proportionality constant is equal to the modulus of elasticity E. and Hooke's law does not apply in this case apart from approximately in a small strain range of the magnitude of 5-10 %. The principle relating to vibration-isolating with springs is that they are placed between the machine and the base or plinth. it also absorbs noise. otherwise the result could be impaired performance. There is no yield point. Rubber achieves high strain levels even at very low tensile stress. the material is extended without the tensile force increasing. Figure 1 illustrates in diagram form a tensile test on steel. The stress strain curve for steel is linear up to the proportionality limit. which is just above the proportionality limit. or yield point. In such cases. and figure 2 illustrates the corresponding test on rubber. the springs must be calculated very accurately. The increase in load is neither linear. To ensure effective isolation. you could say that the vibrations are practically 140 . At the tensile yield limit. in other words. This states that stress is proportional to strain. the force can be reduced to only 2 or 3% of the force of a rigidly mounted machine. In favorable cases. rubber has high internal damping.

and thereby influence Z. 3) The force transmission percentage is reduced substantially by correct calculation and suitable mountings being installed between the machine and base. B drops rapidly. which has minor internal damping and in which the amplitude. Z should be between 3 and 5.Vibration Analysis on Rotating Equipment eliminated. In rubber D has the value 0. If the system's natural vibration coefficient can be altered in any way. the force transmission percentage will be small. in theory. that is. conforms to the formula: B= Equation 14 Amplitude magnification due to springs Where D = damping Z = ratio of actual frequency to resonance frequency B = magnification factor The factor D depends on the internal damping of the spring material. The magnification factor B.0. depending on the hardness of the rubber. Here. B (at Z=D√2) has dropped to 100% and when Z is further increased.e. through many interacting movements. it is said that there is resonance. The low elasticity and shear moduli of rubber are used to achieve a low natural frequency. but in favorable circumstances can be as low as 2 %. All machines have more than one resonance point as. Typical reductions can be from 100 down to 10 %. then. The resonance points can be determined. 2) In the case of an unsuitable spring system. a rubber spring has a distinct advantage over a steel spring. and the vibrations will be infinitely large if there is no damping. This is exactly what happens when vibration isolation is achieved i. they can vibrate in many different ways. Vibration isolation is therefore of significance first when the operating frequency considerably exceeds the natural frequency. that is. by means of springs. If Z=1. The term “4D2 Z2” can generally be neglected completely except in the resonance range. the change in force acting on the base of a machine supported on springs compared with a machine which is rigidly mounted. To sum up. which means that 88 and 96 % of interference forces are eliminated. but the methods of calculation are often 141 . As a rule you usually know the operating speed of a machine (interference frequency). grows to a very high value in the resonance point. which is therefore forced to be a part of the movement of the machine.04 . The relative magnitude of the transmission of force depends entirely on the tuning ratio Z. that is. when Z=1. the machine speed (rpm) = the natural vibrations of the system. the magnification factor can increase considerably and may amount to several hundred per cent. then it is possible to change the force transmitted.1. The magnification factor can be regarded as being 100%. For practical applications. transmission of vibration forces can be affected in three ways: 1) Rigidly mounted machines transmit vibration forces in unchanged form to the base. Figure 116 Resonance Curve As can be seen in figure 10. If Z is high.

192” Number of turns (spring) = 6 so: k = (11. If a material is subjected to a load below the elastic limit. however. To calculate the natural frequency of a spring: Equation 15 Natural frequency of a spring Where k = the spring constant k (linear spring) = Gd4 8nD2 (refer to Hartog p.000 psi for ASTM A229 oil tempered spring steel) d = spring wire diameter n = number of turns D = coil diameter Let us calculate the response of a motor/fan system sitting on a frame which.1924)/(8*6*12) = 317. Calculate the natural frequencies of the unit using the following assumptions: Weight of motor and fan = 500 lbs. the strain may be higher than 1000 %. Metals will normally be softer towards the end of a tensile test. the tensile strength of rubber is rather low. The desired level of isolation and the interference frequency determine where the resonance frequency shall be. Rubber does not have any yield point. and the modulus is increased until there is abrupt failure. have very small strains below the elastic limit. rubber is very ductile.200. However. while the opposite is often the case with rubber. This does not apply to rubber under tension or compression. be proportional to the load. It is usually quite sufficient to calculate the most significant ones which can be determined easily.000 * 0. In some cases. Compared with metals.429) where G = torsional modulus (typically 11. To make the calculation easy we will assume that the weight of the unit is equally distributed over the four springs. on the other hand.2535 Hz Rubber Compared with other engineering materials. The maximum level that can be achieved with rubber is 25-30 MPa. according to Hooke's law.Vibration Analysis on Rotating Equipment very difficult. Experience has shown that all resonance frequencies that can arise do not need to be clarified. is supported by by four springs.200./inch ωn = (317.089 lbs.089*4/500)0. This means that rubber does not have any constant tensile or compression module of elasticity. and by far the highest proportion of this strain is elastic. Spring coil diameter = 1” Wire diameter = 0. rubber has a very large work absorption capacity compared with the best grade of steel. in turn. Metals. 142 .5/2π = 0. the deformation will. because of the high strain.

Vibration Analysis on Rotating Equipment
High elastic ductility is therefore the most pronounced feature of rubber. Just how easy it is to deform rubber is shown by the fact that the modulus of elasticity of compression for rubber within the normal hardness range is between 2 and 12 MPa; while the modulus of elasticity of steel is 210,000 MPa. This means that soft rubber is about 100 000 times softer than steel. Damping capacity is another important feature of compounded rubber. This is of particular importance when operating a machine that is supported on springs through the resonance range. The resonance deflection with rubber springs is only 1/5 to 1/50 compared with the deflection when using steel springs with the same stiffness, see figure 17.

Figure 117 Resonance curve for spring material with different internal damping

With a spring made of natural rubber working under compression or shear load, the direct loss of energy is between 6 and 30% depending on the hardness of the rubber. This energy loss is such that it is possible in many cases to use rubber springs as dampers. But care must be taken when it comes to damping in a rubber element. If the element works with high amplitudes, a substantial amount of energy is converted into heat, and the heat which is generated may lead to the rubber element being destroyed.

Figure 118 Schematic representation of the internal damping properties of rubber. The elliptical area indicates the loss off energy
In the case of single impact, the vibrations sequence will be as shown in figure 19. The left-hand curve represents a steel spring, while the right-hand curve represents a rubber spring. These two curves clearly show just how quickly the vibrations degenerate in the rubber, while in steel springs they diminish slowly. As sound-insulating material, rubber is one of the very best. The effect of sound isolation increases with the

Figure 119 Vibrations sequence with single impact for steel and rubber springs
thickness of the rubber. Rubber is an excellent absorber of impact sound, which occurs in foundations, floors, buildings, etc. High chemical resistance is yet another valuable feature when using rubber. Steel, on the other hand, can rust easily when exposed to air and acids leading to fatigue problems. Rubber is fully resistant to moisture and common acids, and at normal temperatures, does not give rise to ageing problems. The highest temperature to which natural rubber should be continuously exposed is about +75°C (167°F). At extremely

143

Vibration Analysis on Rotating Equipment
low temperatures, -30°C (-22°F) to -40°C (-40°F), rubber stiffens and becomes rigid. At even lower temperatures it becomes hard, brittle and non elastic (though this is reversible). Rubber mountings may sometimes have to be installed at places where they can be exposed to the risk of coming into contact with oil. The harmful effects of oil can often be avoided through the design of the mounting or by fitting a mechanical shield. An alternative would be to use oil-resistant rubber material, but compared to natural rubber it has inferior damping properties. When calculating compression characteristics of rubber, it should be noted that the deflection is not directly proportional to the load, as the modulus of elasticity in compression increases with the degree of stress. The modulus of shear, however, remains constant for normal stresses. The factor with the most effect on stiffness calculation is the ratio between loaded and free surface area of rubber - this is the so-called shape factor (often designated S). With thin rubber sections, a very high modulus of elasticity can be achieved. Equation 16 Modulus of elasticity for rubber E8 = E0 (1+2.const.S2) In other respects, the stiffness of a rubber spring is determined by the dimensions and the hardness of the rubber.

Figure 120 Relationship between hardness and shear modulus

rubber

Figure 20 illustrates the relationship between rubber hardness and shear modulus. Figure 21 shows the dependence of the bulk modulus on the shape factor. The latter curve applies at 10% deformation.

Figure 121 The dependence of the compression modulus upon the shape factor

Rubber at a shape factor of 0.25 is about 6-8 times softer when in shear than when in compression (for the same rubber hardness). Since only 3-4 times the stress value in compression can be considered, it may be said that rubber is best used in shear to achieve large deflections and good isolation properties, particularly at low interference frequencies. The given modulus values apply at a static compression test where the end surfaces can not slide, but when rubber is subjected to rapid vibrations it appears to be stiffer. Thus a rubber mounting displays static and dynamic spring constants. The dynamic spring constant is dependent

144

Vibration Analysis on Rotating Equipment
on the frequency and amplitude. The effect of frequency, however, is small within the ranges in which modern vibration mountings are most effective: from about 13 to 400 Hz. The effect of amplitude is such that rubber will be stiffer the lower the amplitude. The ratio between dynamic and static stiffness is dependent on the material and the geometrical shape of the rubber body in the mounting. One important question is how large a continuous stress can rubber be subjected to? When it comes to shear, normally 0.25 MPa is considered to be a normal stress factor, but in some instances it can go as high as 0.5 MPa without problems. Higher shear stresses are not suitable as they change to tensile stress. Experience has shown that tensile loads are detrimental to rubber bonded designs. If subjected to high shear stresses, rubber should be pre-compressed. When it comes to compression, 1 MPa is a normal stress for the rubber bonded to metal with the restriction that the deformation will not exceed 20-25% of the rubber thickness. With hard rubber, we can except the stress to increase slightly. The fact that the element is made of rubber and bonded to metal does not represent any restriction in the use of the rubber's strength properties; quite the reverse, in fact. The bond between the rubber and the metal is often stronger than the rubber material itself. If the area under load is large in relation to the free area (high shape factor), the modulus of elasticity will be large and the deformation small. Higher compression stresses can then be utilized. Examples of applications are bearings for bridges and elements for vibration-isolation of buildings. The surface pressure here may be as high as 15-20 MPa. One of the reasons why the aforementioned relatively low stresses for normal rubber mountings must be taken into consideration is that the rubber element will be permanently deformed if it is subjected to too high a load. This is called "compression set". The most important factor for reducing compression set is that the rubber is of a high quality and correctly vulcanized. Moreover, it is essential that the rubber spring is not exposed to high temperatures.

Modifying the structure
As we mentioned at the beginning of this section, if it is not feasible to isolate the source from the structure, we must de-tune the structure so that the natural frequency is removed from the forcing frequency or somehow reducing the amplitude. This can be achieved by changing the mass, the stiffness or the damping. Consider the generalized formulae: ωn = k √(Stiffness/Mass) and Where Ra = k (Forcing_Function/Damping) ωn = natural frequency Ra = amplitude of vibration k = a constant

Let us consider what effect each of these three factors will have on the resonant condition of the machine/structure system.

Damping
Looking at the formulae above we see that changing the damping will have zero effect on the natural frequency of the system. In the real world, damping is usually associated with friction, either at the molecular level or because of friction between machine components. The damping qualities of a structure control the resonant response. The higher the damping effect the lower the amplitude of vibration will be as a result of a resonant condition. For example a spring is a good example of a low damped system, a relatively small amplitude of input results in a large amplitude response. A shock absorber is a good example of a highly damped system where a large input results in a small response. In order to have a resonant condition we need to have the right combination of stiffness and mass to create the natural frequency and a source vibration that matches the natural frequency. The damping controls the

145

If stiffness is added to the machine. great care should be exercised. I µ . In other words. By increasing the value for I we can increase the natural frequency of the system. increasing the stiffness by 10% will increase the natural frequency by only 3. Additional stiffness can often be added by using turn-buckles at a very small mass penalty. you could end up with a totally new set of resonance problems.l 4 Remember that µ is the mass per unit length and l is the length. but usually we cannot change the damping of the system significantly. Adding mass to a structure is often a very easy way of checking your analysis of the problem. So increasing the mass of the structure will lower the natural frequency of the structure. I µ . If you think that adding stiffness is the correct solution for the resonance problem.16% (√10). E. only stiffen in the direction which has a problem with the natural frequency being co-incident with the exciting force. you will add mass. 146 . E. But the ratio of increase is not linear. • • If you stiffen in places where it is not required. As you stiffen the structure. Changing the Mass Remember the formula for natural frequency? ωn a 1 . Just by adding sand bags to a structure we can check how much weight we need to add to the structure to remove the natural frequency away from the exciting force (without adding stiffness).Vibration Analysis on Rotating Equipment amplitude response. In small systems it is sometimes feasible to add a damper such as a shock absorber.l 4 The combination of E and I may be considered to be the stiffness of the beam. It is very possible that the additional stiffeners will increase the natural frequency because of their stiffening properties but reduce the natural frequency due to the extra mass. Examining the formula it is obvious that increasing the stiffness will increase the natural frequency of the beam. Changing the stiffness ωn a 1 . be very careful about the mass that is added. But additional mass reduces the natural frequency. Take care when adding the finished mass (often in the form of steel) that you do not increase the stiffness of the structure. In the first place. In other words you have done a lot of work for no change in the vibration characteristics of the machine.

sea vessels (hovercraft. In many cases. Very low frequency vibration (less than one Hz or cycle per second) can result in motion sickness. work environment.) The ACGIH guidelines are based on symptoms of white finger or Raynaud's phenomenon. and gasoline powered tools such as chain saws (70 m/sec). low back pain and spinal degeneration. for example). skill level. aircraft (helicopters.g. Even a building's vibration may reach levels of concern (such as in process control rooms or at machine operators' control platforms). The initial effects of whole body vibration are discomfort and irritation. frequencies between four and eight Hz are likely to affect the gastrointestinal and spinal systems. •Holding vibrating work pieces such as when grinding or polishing. and buildings (generators. neural disorders. posture) and non-occupational factors (age. higher frequency vibration (15 to 40 Hz) can interfere with vision. health. pneumatic and electric rotary tools such as grinders [380 m/sec]). heavy machinery. tractors. vibration-related syndromes are aggravated by other occupational factors (work history.Vibration Analysis on Rotating Equipment Whole body vibration Sources of vibration Exposure to localized vibration occurs when a part of the body comes in contact with a vibrating object. •Holding vibrating controls (e. Whole body vibration usually originates from a moving vehicle or from the operation of heavy machinery. dozers and construction. pumps. trucks. ships). ACGIH and other agencies that recommend vibration guidelines update their guidelines when new information becomes available. Prolonged exposure to whole-body vibration has been associated with back and neck musculoskeletal disorders. Exposure can occur while holding: • • • vibrating tools (e. fitness). fixed-wing aircraft). Much can be gained by routine maintenance of the 147 . Guidelines established by the American Conference of Government Industrial Hygienist (ACGIH. Localized vibration from hand-held power tools can increase grip force and symptoms of hand-arm vibration syndrome (HAV) and carpal tunnel syndrome. ventilation systems). lawn mowers). (The tool manufacturer should be contacted to obtain tool operation characteristics. Frequency ranges The specific effects also depend on the vibration frequencies to which the worker is exposed. forestry or mining equipment). This can contribute to fatigue and a reduced ability to perform certain tasks. Sources of whole body vibration include virtually all modes of transportation (the jarring and jolting associated with off-road vehicles. such as trucks or heavy machinery.. such as power hand tools. Exposure to whole-body vibration can also occur while standing or sitting in vibrating environments or objects. Some examples of whole body vibration sources include the following: • • • • • on-road and off-road vehicles (buses. rail vehicles. impact tools such as chippers (425 m/sec) and riveters.. 1993) suggest that hand vibration in excess of 12 mm/sec should have exposures of less than 1 hour. Prolonged or extreme exposure to whole body vibration has been associated with gastrointestinal disorders.g.

This vibration may be addressed in a number of ways: 1. isolate the system from the vibration source (mount the system on anti-vibration mounts).1. improved seating selection may provide further protection for the worker.Vibration Analysis on Rotating Equipment vibration source and existing isolation systems. Control room vibration may result from machinery operation or heavy vehicles traveling in the vicinity. 148 . Generally. or • 2.1. There are large differences between seats and the performance of an individual seat may change during its lifetime. track and suspension systems. Tire pressure. 2. Anti-vibration suspension seats include a damper and a spring in the seat system. vibration measurements will likely be necessary to make an educated selection of appropriate vibration control strategies. 2. this may be accomplished by • 2. increase the damping in the system (damping materials may be coated on the system).2. Conventional foam seats tend to amplify vibration frequencies in the range of greatest susceptibility for the human body. Vehicle cab isolation is available in some vehicles and can effectively reduce vibration exposure. reduce vibration at the source (in the case of heavy machinery this may be accomplished by balancing moving parts) modify a vibrating system to reduce the likelihood of excessive vibration due to resonance (in the case of a cover panel. as well as seating and engine mounts should be routinely checked and serviced when required. vehicle maintenance and control of vehicle speed. If vibration control available through vehicle and roadway maintenance is insufficient. stiffening the system by welding a steel section to it). Table 6 Whole body vibration (frequency ranges) Frequency Range < 1Hz 4 – 8 Hz 15 – 40 Hz Physiological Result Motion sickness Gastrointestinal & spinal problems Vision difficulties Whole body vibration transmitted to vehicle operators may be reduced by proper maintenance of roadways. The spring may be constructed of steel or it may be a column of air. two types of seating systems are considered when addressing a vibration problem: conventional foam seats or suspension seats. It is important to know the vibration levels and frequencies in the vehicle in order to select an appropriate suspension seat. For more persistent problems.1.1.

This is the antialiasing filter. Averaging In general. ADC Acronym for Analog to Digital conversion Alarm Alarms are used to identity specific operating conditions or to define the boundaries between safe and unsafe conditions. Acceleration The rate of change of velocity in inches/second/second or G’s (acceleration due to gravity is 1 G). AC Literally Alternating Current but often used to imply a time-fluctuating signal. averaging many spectra together improves the accuracy and repeatability of measurements. When an amplitude reaches or exceeds the alarm amplitude then the software automatically generates an exception report. Anti Aliasing Nyquist's theorem says that as long as the sampling rate is greater than twice the highest frequency component of the signal. ASCII American Standards Code for Information Interchange. Attenuation The weakening of a signal by distance from the source or a mechanical interface. It is one of the most rugged transducers for vibration measurements and also has the widest frequency range.a (Newton’s 2nd law). Bin See “resolution” Blackman-Harris Window 149 . then the sampled data will accurately represent the input signal. Certain analyzers pass the input signal passes through an analog filter which attenuates all frequency components above Fmax by 90 dB to make sure that Nyquist's theorem is satisfied. B Band Pass Filter A measurement filter that removes data below the low cutoff frequency and above the high cutoff frequency.Vibration Analysis on Rotating Equipment Glossary A Accelerometer Transducer for measuring vibration in the form of acceleration. The band pass filter only passes the data between the cutoff frequencies. This is a good indicator of the forces inside a machine since F = m.

If the coherence is 1.Vibration Analysis on Rotating Equipment The Blackman-Harris window is a very good window to use with the spectrum analyzer. Coherence Coherence measures the percentage of power in channel 2 which is caused by (phase coherent with) power in the input channel. Cross Spectrum The cross spectrum is defined as: cross spectrum = FFT2 conj(FFT1) The cross spectrum is a complex quantity which contains magnitude and phase information. Frequencies where signals are present in both spectra will have large components in the cross spectrum. 150 . The magnitude is simply the product of the magnitudes of the two spectra. CPM Abbreviation for Cycles per Minute . very good selectivity and the fastest filter rolloff. e. even when their amplitudes are very different. all the power of the output signal is due to the input signal. the input and output are completely random with respect to one another. This allows signals close together in frequency to be distinguished. It has better amplitude accuracy (about 0.g.7 dB) than the Hanning. and the cross-correlation is a measure of how much overlap a signal has with a delayed-by-t version of the other channel. C Calibration The process of multiplying or dividing the voltage signal from a transducer by a factor that represents a specific engineering quantity. Coherence is related to signal to noise ratio (S/N) by the formula: S/N = γ2/(l -γ2) where γ2 is the traditional notation for coherence. The filter is steep and narrow and reaches a lower attenuation than the other windows. It is clear that the auto correlation at a time t is a measure of how much overlap a signal has with a delayed-by-t version of itself. Correlation The two channel analyzer may also compute auto and cross correlation. If the coherence is 0.the most common format for displaying frequency in vibration analysis. Correlation is a time domain measurement which is defined as follows: Auto Correlation(τ) = ∫x*(t)x(t-τ)dt Cross Correlation(τ) = ∫x*(t)y(t-τ)dt where x and y are the channel 1 and channel 2 input signals and the integrals are over all time. Although correlation is a time domain measurement the some analyzers use frequency domain techniques to compute it in order to make the calculation faster. Coherence is a unitless quantity which varies from 0 to l. 100 mV/G for an accelerometer. The phase is the relative phase between the two channels.

dB = 10 log (X/Xο) where X or Xο is a power or squared quantity or where Y or Yο is a linear quantity. Exponential Averaging Exponential averaging weights new data more than old data.1 GdB re 0.806 m/S2 E Envelope Measurements A.f) = Displacement /(2. Decibels are usually referenced to a standard. 0.k. Data is outside the defined “safe” area.Vibration Analysis on Rotating Equipment D Database A collection of information files that are ties together by a common topic. Averaging takes place according to the formula.a. Decibel (dB) A convenient unit for displaying logarithmic data.g. DC Literally Direct Current.I/N) +(Old Average) . 151 .b.π. demodulation (see appendix) Exception To have an alarm condition. remember 1G = 386 in/s2 or 9.f) = Displacement /(2.π.001G. Data Collector Hardware device for collecting vibration data off line.π. New Average = (New Spectrum . In PdM terms this may be used to check the accelerometer or for a process measurement. Differentiate Differentiation changes displacement to velocity and velocity to acceleration relative to time using the formulae: Acceleration Velocity or Acceleration = Velocity /(2. A database allows rapid access of the database files.f)2 dB = 20 log (Y/Yο) where f = frequency n. e.(N-l)/N where N is the number of averages.

However. the Flattop window has a wide pass band and very steep rolloff on either side. Cycles per Minute (cpm) or orders (multiples of run speed). As a result.Vibration Analysis on Rotating Equipment Exponential averages "grow" for approximately the first 5N spectra until the steady state values are reached. The unit of frequency that is measured in cycles per second was renamed the hertz. Make sure that the number of averages is not so large as to eliminate the changes in the data that might be important. German physicist born in Hamburg and educated at the University of Berlin. Fundamental Frequency Primary frequency. Hertz proved that electricity can be transmitted in electromagnetic waves. Typical units are Hertz (Hz). Its filter rolloff is not particularly steep. Hertz (Hz) Common frequency units in cycles per second. Unlike the Hanning. Heinrich Rudolf Hertz (1857-94). signals appear wide but do not leak across the whole spectrum. Once in steady state. H Hanning Window The Hanning window is the most commonly used window. such as operating speed. High Pass Filter 152 . His experiments with these electromagnetic waves led to the development of the wireless telegraph and the radio. From 1885 to 1889 he was a professor of physics at the technical school in Karlsruhe and after 1889 a professor of physics at the university in Bonn. FFT Spectrum Analyzer FFT Spectrum Analyzers take a time varying input signal. it is commonly abbreviated Hz. to which other frequencies may be referred back. Thus. Named after. Frequency The rate at which periodic events happen.5 dB (for signals between bins) and provides reasonable selectivity. and displays the spectrum of these measured frequency components.02 dB. Its between-bin amplitude variation is about 0. It has an amplitude variation of about 1. and compute its frequency spectrum. the Hanning window can limit the performance of the analyzer when looking at signals close together in frequency and very different in amplitude. which travel at the speed of light and which possess many other properties of light. F Flattop Window The Flattop window improves on the amplitude accuracy of the Hanning window. like you would see on an oscilloscope trace. Fourier or FFT Fourier's theorem states that any waveform in the time domain can be represented by the weighted sum of sines and cosines. the selectivity is a little worse. Harmonic Frequencies at direct multiples of a fundamental frequency. The fundamental is not necessarily the run speed. further changes in the spectra are detected only if they last sufficiently long. computes the magnitude of its sine and cosine components. Hertz clarified and expanded the electromagnetic theory of light that had been put forth by the British physicist James Clerk Maxwell in 1884. The FFT spectrum analyzer samples the input signal.

8 dB for signals between exact bins).It turns out that the log display is both easy to understand and shows features which have very different amplitudes clearly. it is the best window to use for measurements requiring a large dynamic range. which is available on IRD analyzers.01% of full scale would look like on a linear scale. Linear Linear scaling displays all data on an equal basis. L Leakage Errors resulting from Fourier transforming nonperiodic time domain data. This avoids the problem of taking the log of negative voltages. Imagine what something 0. Because of these properties. The real and imaginary parts are always displayed on a linear scale. Line of Resolution see “resolution” Logarithmic Although the linear magnitude scale is used most often for displaying spectra. Low Pass Filter A measurement filter that removes data above the cutoff frequency. DC coupling includes both DC and AC signals whereas AC coupling does not include any DC offsets. Linear Averaging Linear averaging combines N (number of averages) spectra with equal weighting in either RMS. The Kaiser window has the lowest side-lobes and the least broadening for non-bin frequencies. This type of averaging is useful for eliminating transients. the top of the graph would be 833 feet above the bottom . another way of displaying amplitude is the Log Magnitude. combines excellent selectivity and reasonable accuracy (about 0. below full scale. This effect is reduced by using windows such as Hanning or Flat Top.π.Vibration Analysis on Rotating Equipment A measurement filter that removes data below its low cutoff frequency. 153 . I Input Couple Measurements may be AC or DC coupled. Integrate Integration is the opposite to differentiation and changes acceleration to velocity and velocity to displacement where A = V x (2. The 16 bit analyzer has a dynamic range of about 90 dB.f) K Kaiser The Kaiser window. If we wanted it to be 1 inch high on the graph. The Log Mag display graphs the magnitude of the spectrum on a logarithmic scale using dBEU (Engineering Units) as units. Vector or Peak Hold fashion. Use DC coupling for process measurements and use AC for vibration measurements. Small signals may be hard to detect with linear scaling but become visible with logarithmic scaling.

With overlap processing. It loads programs. 0% is no overlap and 99. The amount of overlap is specified as a percentage of the time record. the second order is two times the operating speed and so on. When RMS averaging narrow spans. the analyzer does not wait for the next complete time record before computing the next FFT.8% is the maximum (511 out of 512 samples re-used). or 1/12 octave. Orbit The orbit is simply a two dimensional display of the time record of channel 1 vs. And what is the processor doing while it waits? Nothing. The first order corresponds to the operating speed. Orders A frequency axis scale which is useful for viewing data as a function of the operating speed. The full octave bands have band centers at: Center Freq: = 1 kHz x 2n The 1/3 octave bands have center frequencies given by: Center Freq: = 1 kHz x 2(n-30/3) Operating System A form of software that controls and supervises how the computer operates.11 (1986). this could be quite slow. This is why overlap effectively speeds up averaging and smoothes out window variations. the time record of channel 2. Typically. With overlap. Overlap Processing What about narrow spans where the time record is long compared to the processing time which is what we normally see when taking vibration measurements? The analyzer computes one FFT per time record and can wait until the next time record is complete before computing the next FFT. The maximum overlap is determined by the amount of time it takes to calculate an FFT and the length of the time record and thus varies according to the span. This speeds up the processing rate. this can reduce the measurement time by a factor of two. the points at the ends of the time record do not contribute much to the FFT. time records with 50% overlap provide almost as much noise reduction as non-overlapping time records when RMS averaging is used. handles input and output operations and accepts and executes commands issued by the user. Octave analysis measures spectral power closer to the way people perceive sound. in octaves. The orbit display is similar to an oscilloscope displaying a "Lissajous" figure. The bands are arranged in octaves with either 1 or 3 bands per octave (1/1 or 1/3 octave analysis). that is. Remember. 1/3. Typically the shape of each band is a third-order Butterworth filter whose bandwidth is either a full. measurements this value includes vibration at all frequencies. The update rate would be no faster than one spectra per time record. Octave analysis computes the spectral amplitude in logarithmic frequency bands whose widths are proportional to their center frequencies. With narrow spans. The center frequency of each band should be calculated according to ANSI standard S1. For vibration P 154 . these points are “re-used" and appear as middle points in other time records. most window functions are zero at the start and end of the time record. Overall A single value representing the vibration or some other measurement parameter. Instead it uses data from the previous time record as well as data from the current time record to compute the next FFT.Vibration Analysis on Rotating Equipment O Octave Analysis The magnitude of the normal spectrum measures the amplitudes within equally divided frequency bins. Thus.

Peak Hold detects the peaks in the spectral magnitudes and only applies to Spectrum. it is the frequency span whose corresponding time record exceeds the time it takes to compute the spectrum. This allows measurements with different spans to be compared. This measurement approximates what the spectrum would look like if each frequency component were really a 1 Hz wide piece of the spectrum at each frequency bin. PSD.s. Since the PSD uses the magnitude of the spectrum. Phase In general.Vibration Analysis on Rotating Equipment Peak The maximum value as seen in the time domain data. The phase is relative to the pulse of the trigger. The phase is displayed in degrees or radians on a linear scale.). normalizes all measurements to a 1 Hz bandwidth and the noise spectrum becomes independent of the span. If the real or imaginary part or phase is being displayed for spectrum measurements. and if the new data is larger. the display shows the real or imaginary part or phase of the complex peak value. the new spectral magnitudes are compared to the previous data. If the noise is Gaussian in nature. phase measurements are only used when the analyzer is triggered. usually from -180 to +180 degrees. and Octave Analysis measurements. The resulting display shows the peak magnitudes which occurred in the previous group of spectra. R Rayleigh’s Principle This principle states that ∆f is the lowest measurable frequency for a time record length T: ∆f = 1/T Real Time Bandwidth What is real time bandwidth? Simply stated. However. some data samples will be lost while the FFT computations are in progress. The spectra are computed in "real time". the amplitude of the spectrum changes with the frequency span.012% of full scale (-78 dB below f. At larger spans. even if a signal is small. instead.414 Peak Hold Averaging Peak Hold is not really averaging. This avoids the messy phase display associated with the noise floor. the PSD is a real quantity. The phase of a particular frequency bin is set to zero in most analyzers if neither the real nor imaginary part of the FFT is greater than about 0. the peak magnitude values are stored in the original complex form. This is because the line width changes so the frequency bins have a different noise bandwidth.) Power Spectral Density (PSD) The PSD is simply the magnitude of the spectrum normalized to a 1 Hz bandwidth. This is done on a frequency bin by bin basis. its phase extends over the full 360 degrees. At this span and below. it is possible to compute the spectra for every time record with no loss of data. Thus the PSD is displayed in units of V/√Hz or dBV/√Hz. then the new data is stored. Rectangular Window (Uniform or No Window) 155 . There is no real or imaginary part or phase. (Remember. then the amount of noise amplitude in other bandwidths may be approximated by scaling the PSD measurement by the square root of the bandwidth. The PSD. on the other hand. For a sine wave peak = RMS x 1. What good is this? When measuring broad band signals such as noise.

S Sampling Rate The original digital time record comes from discrete samples taken at the sampling rate. Of Lines Bin resolution = data sampling rate / 2. It takes 0. a very good approximation of the actual random noise floor can be displayed.5 Hz (the same as the lowest measurable frequency) Everything below 2. Sidebands occur because of a modulation of the fundamental by another frequency. The RMS average has no phase information. The weighting is either linear or exponential.56. For example a spectrum of 500 Hz Fmax with 400 bins or lines would have a line resolution of 1. but over what frequency range? The highest frequency will be determined by the in-built ratio of F-max to data sampling rate . Sideband A frequency which occurs either side of a fundamental frequency.Vibration Analysis on Rotating Equipment The rectangular windows applies an even weighting (no window) over the time period. Route For off-line data collectors this is a an ordered list of points containing the sequence for collecting data. In fact. The lowest frequency is just the F-max divided by the number of lines: F-max No.56) / (1024 / 2. Resolution The accuracy of a reading based on the number of discrete values used to define it. RMS averaging reduces fluctuations in the data but does not reduce the actual noise floor. The corresponding FFT yields a spectrum with discrete frequency samples.56 = No samples / 2. the spectrum has less than half as many frequency points as there are time points. RMS (Root Mean Square) The square root of the average of a set of squared values. Spectrum 156 . Since RMS averaging involves magnitudes only. It does not reduce leakage errors and should only be used for impulsive or transient data that dies out within the time sample period. of lines = (2560 / 2.5 Hz.4 Seconds to take this time record.56 = F-max / No.7071 RMS Averaging RMS averaging computes the weighted mean of the sum of the squared magnitudes (FFT times its complex conjugate). The FFT of this record yields 400 frequency points or lines. For a frequency domain spectrum this refers to the number of lines (or bins) of resolution that are combined to display the spectral data.56) = 2.2. The output spectrum thus represents the frequency range from DC to 1000 Hz with points every 2. displaying the real or imaginary part or phase of an RMS average has no meaning.25 Hz. With a sufficient number of averages. Suppose that you take 1024 samples at 2560 Hz.5 Hz is considered to be DC. For a sine wave RMS = Peak x 0.

The magnitude is a real quantity and represents the total signal amplitude in each frequency bin. coherence and orbit. U Uniform Window The uniform window is actually no window at all. it will affect every bin of the spectrum. Time Record The time record measurement displays the filtered data points before the FFT is taken For baseband spans (spans that start at DC). (It is exactly. this window is only useful when looking at transients which do not fill the entire time record. (Square root of the sum of the real (sine) part squared and the imaginary (cosine) part squared). For non-baseband spans (zooms) the heterodyning discussed earlier transforms the time record into a complex quantity which can be somewhat difficult to interpret. If its frequency is between bins. The time record is used with no weighting.e. Normally. the time record is triggered in phase with some component of the signal. For the transfer function to be valid. Two-Channel Measurements Two-channel analyzers offer additional measurements such as transfer function. the input spectrum must have amplitude at all frequencies over which the transfer function is to be measured.periodic within the time record). then the real (cosine) or imaginary (sine) part or the phase may be displayed. 157 . i. the time record is a real quantity. Time Synchronous Averaging see “vector averaging” Transfer Function The transfer function is the ratio of the spectrum of channel 2 to the spectrum of channel 1. the magnitude of the spectrum is displayed. These two cases also have a great deal of amplitude variation between them (up to 4 dB). independent of phase. In general.Vibration Analysis on Rotating Equipment The spectrum is the basic measurement of an FFT analyzer. T Tachometer A device for measuring the speed of rotation. The magnitude is the square root of the FFT times its complex conjugate. It is simply the complex FFT. The phase is simply the arc tangent of the ratio of the imaginary and real parts of each frequency component. The phase is always relative to the start of the triggered time record. If there is phase information in the spectrum. A signal will appear as narrow as a single bin if its frequency is exactly equal to a frequency bin. cross-spectrum.

since the filter is narrow. It may not be identical from record to record. its points are multiplied by the window function. In the frequency domain. BlackmanHarris. time bin by time bin. Windows are functions defined across the time record which are periodic in the time record. With vector averaging. W Windowing What is windowing? Let's go back to the time record. it falls off from center rapidly. the signal can actually smear across the entire spectrum. 158 . Flattop. This smearing will also change wildly between records because the amount of mismatch between the starting value and ending value changes with each record. This means that even frequencies close to the bin may be attenuated somewhat. The different types of windows trade off selectivity. When the time record is windowed. The net result of windowing is to reduce the amount of smearing in the spectrum from signals not exactly periodic with the time record. a window acts like a filter. This is because the complex information is preserved. then only frequencies near the bin will contribute to the bin. and Kaiser. but it will be periodic (zero at each end). If the time record does not start and stop with the same data value. amplitude accuracy. (The real part is averaged separately from the imaginary part.) This can reduce the noise floor for random signals since they are not phase coherent from time record to time record.it selects a small range of frequencies around each bin. and noise floor. They start and stop at zero and are smooth functions in between. The signal of interest must be both periodic and phase synchronous with the trigger. Otherwise. the real and imaginary parts as well as phase displays are correctly averaged and displayed.Vibration Analysis on Rotating Equipment V Vector (Synchronous Time) Averaging Vector averaging averages the complex FFT spectrum. and the resulting time record is by definition periodic. The amplitude of each frequency bin is determined by centering this filter on each bin and measuring how much of the signal falls within the filter. However. This is true but it's actually a bit worse than that. If the filter is narrow. Vector averaging requires a trigger. A narrow filter is called a selective window . What happens if a signal is not exactly periodic within the time record? We said that its amplitude is divided into multiple adjacent frequency bins. Several types of window functions are available including Uniform (none). the real and imaginary parts of the signal will not add in phase and instead will cancel randomly. then frequencies far from the bin will contribute to the bin amplitude but those close by will not be attenuated significantly. Hanning. If the filter is wide.

131. 62. 109. 101. 104. 101. 131. 28. 69. 154 accelerometer. 152 Accelerometer. 133. 94. 41. 91. 120 Analysis. 71. 18. 29. 3. 6. 74. 26. 41. 15. 95. 13. 3. 6. 69. 14. 126 Critical speed. 7. 2. 153 Fundamental. 20. 154 acceleration. 31. 13. 62 envelope. 6. 104 Imbalanced phases. 152 Demodulation. 6. 110. 108. 2. 2. 2. 38. 152 Enveloping. 60. 42. 119 —H— Hanning. 26. 43. 142 fans. 140. 27. 9. 156. 91. 14. 13. 71. 71. 19. 94. 16. 6. 2. 40. 30. 6. 73. 21. 115. 23. 22. 155. 120 Hertz. 92. 9. 29. 100. 41. 43. 7. 118. 26. 87. 3. 36. 3. 119 Bin. 80. 45. 22. 93. 6. 114. 66. 153. 122. 64. 157. 40. 155. 151 Cracking. 17. 69 FFT. 80. 70. 66. 6. 35. 60 aerodynamic forces. 7. 7. 22. 13. 90. 9. 139. 152 displacement. 100. 32 Integrate. 7. 4. 78. 100. 39. 127 —E— eccentric rotor. 3. 78. 26. 90. 6. 112. 113. 5. 19. 99. 77. 130. 146 DC. 110. 2. 46. 106. 91 fan. 2. 104. 153 Armature. 45. 70. 89. 70. 130 Bearing. 153. 68. 114 Benzene. 74. 65. 153. 32 BPIR. 152. 48. 38. 3. 78. 21. 67. 67 Differentiate. 64. 148. 94. 156. 10. 158. 151. 61 Coherence. 151 Bode. 111. 129. 103. 4. 101. 75. 34. 27 Developed Fatigue. 103 Envelope. 130. 121. 47. 145. 92 —F— Failure type. 102. 156 Analyzer. 117. 132 Boltzman. 15. 154. 18. 159 Fourier. 93. 23. 20. 123 —I— Imbalance. 27 External manifestation. 93. 41 Compression mode. 117. 50. 6. 28. 159 Decibel. 103. 59. 3. 21. 119. 100. 67. 13 Bump test. 154 Frequency. 9. 6. 59 Air. 24. 150. 3. 93. 32 Average. 157. 6. 17. 24 Correlation. 72. 30. 153 Hydraulic valve. 28. 134. 37 Flattop. 3. 28. 70. 130 —G— Gas. 24 Aerodynamic cross coupling. 3. 67. 100. 79. 60 Hz. 99. 2. 150 Accelerometers. 93. 91 bearing. 32 Elasto Hydrodynamic Lubrication. 13. 153 —B— Balancing. 150. 3. 152. 149. 153 Harmonics. 152. 33 Impact hammer. 3. 159 Harmonic. 102. 60. 39. 121 bowed rotor. 99. 45. 12. 159 Filter. 2. 4. 3. 158. 2. 143. 152 Averaging. 3. 99. 65. 34. 106. 41. 5. 21. 157. 153. 6. 117. 11. 127. 6. 45. 42. 103. 104. 107. 88. 135. 127. 7. 85 Base cause. 63. 6. 154. 87. 151. 151 CPM. 154. 25. 158 —C— Campbell. 151 Comparitor Card. 72. 38. 37. 158 159 . 3. 120 Induction Motors. 49.Vibration Analysis on Rotating Equipment Index —A— AC. 106. 3. 153. 10. 124 Choking. 7. 62. 43. 22. 120 hysteresis whirl. 82. 155 Firing Cards. 103. 68. 115. 63. 154. 154 Displacement. 93. 7. 118. 121 Glycerin. 75. 157 Blackman-Harris. 2. 69. 2. 59. 154 —D— Damping. 131. 92. 132 Impeller. 36. 40. 152. 156. 13.

Vibration Analysis on Rotating Equipment Interval. 118. 159 —V— Vane pass. 157 —W— waterfall. 4. 87. 9. 18. 128. 130. 99. 10. 142. 22. 146. 101. 84. 74. 41. 104 lubricant wedge. 18 Peak Hold. 4. 13. 7. 91. 30. 117. 69. 7. 50. 15. 105 Shear mode. 139. 146 Mercury. 7. 85. 101. 72. 107. 158. 3. 135. 73. 158 Speed of sound. 8. 31. 38 Setup. 80. 31. 154 Looseness. 25. 6. 157 rotor bars. 53. 118 Logarithmic. 126. 57. 141. 158. 83. 84. 158 Technology. 131. 81. 2. 56. 7. 130. 155 Orbits. 13. 94. 117. 121. 143. 4. 159 Time. 4. 106. 119. 26. 155. 2. 122. 7. 100. 152. 127. 29. 122. 107. 157. 154. 6. 118. 95 Phase. 142 Newton. 110. 76 Transfer Function. 71. 142. 65. 27. 127 mounting techniques. 4. 91. 25. 140. 150. 29. 155. 153. 61. 50. 120 PRELOAD. 101. 154. 65. 19. 99. 152. 123. 6. 152. 64. 54. 151. 102. 34 Rotor Bars. 11. 101. 2. 113. 108. 15. 66. 103. 102. 51. 16. 127. 33. 112. 114. 101. 159 tolerance stack up. 64. 61 Surging. 7. 154 Liquids. 18. 3. 30. 73. 73. 32. 93. 124 waveform. 10. 30.C. 132. 118. 3. 106. 36. 73. 3. 157. 133. 100. 20. 127. 92. 133. 120. 141 RMS. 22. 22. 121 Spring. 2. 114. 2. 123. 32. 92 time. 58. 124 160 . 155. 50. 92 Severity. 39. 145 —K— Kaiser. 52. 34. 91. 15. 32. 22. 149 Resonance. 128. 92 RUB. 156. 151. 133. 16. 14. 154 Line. 129. 154. 24 Sideband. 155 Overlap. 91. 34 Route. 102. 79. 38. 130 whirl. 2. 72. 2. 155 —P— Parameter. 106. 21. 37. 114 Mode shape. 124. 132 —T— Tachometer. 159 velocity. 140. 23. 106. 33 Mass. 6. 59. 92 PEAK. 13. 94 spectrum. 3. 4. 18. 22. 2. 19. 159 Spectrum. 3. 139 —R— Resolution. 4. 6. 156. 113 Rubber. 158 Simple Harmonic Motion. 129. 60 —L— Leakage. 93. 50. 33. 90 —N— Natural frequency. 118. 33. 45. 109. 59. 4. 157. 80. 82 spall. 4. 147. 119. 69. 129. 28. 109 —O— Octave. 158. 7. 156 Piston pulse. 117. 94. 107. 139. 112. 101. 55. 110 Overall. 21. 144. 6. 33. 94. 83. 18. 31. 104. 156. 74. 101. 111. 11. 108. 120 vector. 154. 33. 157 resonance. 111. 99. 41. 7. 10. 158. 26. 157. 129. 112. 104. 102. 123. 24. 130. 31. 127. 64 —M— magnetic center. 112 —U— Uniform. 92. 11. 103. 100. 6. 115. 43. 112.R. 156. 124. 20. 113. 72. 153. 99. 92. 118. 154 Vibration isolators. 156 PFA. 151. 100. 155. 15 Single Plane. 21. 101. 145. 20. 158 Triggering. 63. 64. 116. 143. 34. 2. 94. 14. 119 Misalignment. 153. 117. 4. 15. 6. 4. 154. 150 Nyquist. 33. 19. 104. 23. 100. 8. 82. 9. 60. 69. 23. 7. 28. 150. 104. 2. 7. 3. 6. 121. 94. 17. 62. 89. 5. 2. 40. 117. 19. 156 Orbit. 79. 27. 29. 115. 4. 91. 113. 50. 109. 106. 121. 69. 20. 69. 7. 153 whip. 18. 84. 70. 101. 4. 26. 131. 158. 159 —S— S. 94. 6. 135. 154. 99. 3. 116. 12. 142 Stone Walling. 4.

71 Windows. 159 161 . 32. 4. 99. 33. 149 windings.Vibration Analysis on Rotating Equipment Whole body vibration. 148.