Unit 4 Vibrationdata Random Vibration 1 Random Vibration Examples Turbulent airflow passing over an aircraft wing Oncoming turbulent wind against a building Rocket vehicle liftoff acoustics Earthquake excitation of a building Vibrationdata 2 Random Vibration Characteristics Vibrationdata One common characteristic of these examples is that the motion varies randomly with time. Thus, the amplitude cannot be expressed in terms of a "deterministic" mathematical function. Dave Steinberg wrote: The most obvious characteristic of random vibration is that it is nonperiodic. A knowledge of the past history of random motion is adequate to predict the probability of occurrence of various acceleration and displacement magnitudes, but it is not sufficient to predict the precise magnitude at a specific instant. 3 Optics Analogy Vibrationdata Sinusoidal vibration is like a laser beam Random vibration is like “white light” White light passed through a prism produces a spectrum of colors 4 Music Analogy Vibrationdata Playing a single piano key produces sinusoidal vibration (fundamental + harmonics) Playing all 88 piano keys at once produces a signal which approximates random vibration 5 Types of Random Vibration Random vibration can be broadband or narrow band Random vibration can be stationary or nonstationary Vibrationdata Stationary random vibration is where the key statistical parameters remain constant with each consecutive time segment Parameters include: mean, standard deviation, histogram, power spectral density, etc. Shaker table tests can be controlled to be stationary for the test duration Measured data is usually nonstationary White noise and pink noise are two special cases of random vibration 6 White Noise Vibrationdata Commercial white noise generator designed to produce soothing random noise which masks household noise as a sleep aid. White noise and pink noise are two special cases of random vibration White noise is a random signal which has a constant power spectrum for a constant frequency bandwidth It is thus analogous to white light, which is composed of a continuous spectrum of colors Static noise over a non-operating TV or radio station channel tends to be white noise 7 Pink Noise Vibrationdata Waterfalls and oceans waves may generate pink noise Pink noise is a random signal which has a constant power spectrum for each octave band This noise is called pink because the low frequency or “red” end of the spectrum is emphasized Pink noise is used in acoustics to measure the frequency response of an audio system in a particular room It can thus be used to calibrate an analog graphic equalizer 8 Vibrationdata Sample Random Time History, Synthesized WHITE NOISE 5 4 mean =0 3 std dev =1 ACCEL (G) 2 Sample rate = 20K samples/sec 1 0 Band-limited to 2 KHz via lowpass filtering -1 -2 Stationary -3 -4 -5 0 2 4 6 8 10 TIME (SEC) Synthesize time history with Matlab GUI script: vibrationdata.m 9 Sample Random Time History, Close-up View Vibrationdata WHITE NOISE 5 4 3 ACCEL (G) 2 1 0 -1 -2 -3 -4 -5 2.00 2.02 2.04 2.06 2.08 2.10 TIME (SEC) 10 Vibrationdata Random Time History, Standard Deviation WHITE NOISE 5 Peak Absolute = 4.5 G 4 3 Std dev = 1 G ACCEL (G) 2 1 Crest Factor 0 -1 = (Peak Absolute / Std dev) -2 = (4.5 G/ 1 G) -3 = 4.5 -4 -5 0 2 4 6 8 10 TIME (SEC) 11 Histogram Comparison Vibrationdata Sine Vibration has bathtub shaped histogram Sine vibration tends to linger at its extreme values Random Vibration has a bell-shaped curve histogram Random vibration tends to dwell near zero Thus, there is no real way to directly compare sine and random vibration. But we can “sort of” make this comparison indirectly by taking a rainflow cycle count of the response of a system to each time history. Rainflow fatigue will be covered in future units. 12 Random Time History, Histogram Vibrationdata Histogram of white noise instantaneous amplitudes has a normal distribution. The amplitude is expressed in bins with unit of G. 13 Statistics of Sample Time History Parameter Value Duration 10 sec Sample Rate 20K sps Samples 200K Mean 0 Std Dev 1 RMS 1 Skewness 0 Kurtosis 3.0 Maximum 4.3 Minimum -4.5 Vibrationdata Consider limits: -4.49 to 4.49 Normal distribution Probability within limits 0.99999288 Probability of exceeding limits 7.1223174e-06 7.1223174e-06 * 200000 points = 1.4 Rounding to nearest integer . . . One point was expected to exceed 4.5 in terms of absolute value. 14 RMS and Standard Deviation Vibrationdata = standard deviation RMS = root-mean-square [ RMS ] 2 = [ ] 2 + [ mean ]2 RMS = assuming zero mean 15 Peak and RMS values Vibrationdata Pure sine vibration has a peak value that is 2 times its RMS value Random vibration has no fixed ratio between its peak and RMS values Again, the ratio between the absolute peak and RMS values in the previous example is 4.5 G / 1 G = 4.5 16 Vibrationdata Statistical Formulas Mean = 1 n Variance = n n Yi Y i Skewness = i 1 i 1 1 n 3 n Yi n 2 i 1 n Kurtosis = Y i 4 i 1 n 3 4 Standard Deviation is the square root of the variance where Yi is each instantaneous amplitude, n is the total number of points, is the mean, is the standard deviation 17 Statistics of Sample Time History Vibrationdata Random vibration is often considered to have a 3 peak for design purposes Need to differentiate between input and response levels Response is more important for design purposes, fatigue analysis, etc. Both input and response can have peaks > 3 even for stationary vibration 18 Probability Values for Random Signal Vibrationdata Normal Distribution, Instantaneous Amplitude Statement Probability Ratio Percent - < x < + 0.6827 68.27% -2 < x < +2 0.9545 95.45% -3 < x < +3 0.9973 99.73% 19 More Probability Vibrationdata Normal Distribution, Instantaneous Amplitude Statement Probability Ratio Percent |x|> 0.3173 31.73% | x | > 2 0.0455 4.55% | x | > 3 0.0027 0.27% 20 SDOF Response to White Noise Vibrationdata The equation of motion was previously derived in Webinar 2. Apply the white noise base input from the previous example as a base input to an SDOF system (fn=600 Hz, Q=10). 21 Solving the Equation of Motion Vibrationdata A convolution integral is used for the case where the base input acceleration is arbitrary. The convolution integral is numerically inefficient to solve in its equivalent digitalseries form. Instead, use… Smallwood, ramp invariant, digital recursive filtering relationship! 22 SDOF Response Vibrationdata mean =0 std dev =2.16 G Peak Absolute = 9.18 G Crest Factor = 9.18 G / 2.16 G = 4.25 The theoretical Crest Factor from the Rayleigh Distribution is 4.31 Rice Characteristic Frequency = 595 Hz 23 SDOF Response, Close-up View Vibrationdata SDOF system tends to vibrate at its natural frequency. 60 peaks / 0.1 sec = 600 Hz. 24 Histogram of SDOF Response Vibrationdata The response time history is narrowband random. The histogram has a normal distribution. 25 Histogram of SDOF Response Peaks Vibrationdata The histogram of the absolute response peaks has a Rayleigh distribution. 26 Rayleigh Distribution Vibrationdata Consider a lightly damped, single-degree-of-freedom system subjected to broadband random excitation The system will tend to behave as a bandpass filter The bandpass filter center frequency will occur at or near the system’s natural frequency. The system response will thus tend to be narrowband random. The probability distribution for its instantaneous values will tend to follow a Normal distribution, which the same distribution corresponding to a broadband random signal The absolute values of the system’s response peaks, however, will have a Rayleigh distribution 27 Rayleigh Distribution Vibrationdata R A Y L E IG H D IS T R IB U T IO N F O R = 1 0 .7 0 .6 0 .5 p (A ) 0 .4 0 .3 0 .2 0 .1 0 0 0 .5 1 .0 1 .5 2 .0 2 .5 3 .0 3 .5 4 .0 A 28 Rayleigh Probability Table Vibrationdata Rayleigh Distribution Probability Prob [ A > ] 0.5 88.25 % 1.0 60.65 % 1.5 32.47 % 2.0 13.53 % 2.5 4.39 % 3.0 1.11 % 3.5 0.22 % 4.0 0.034 % Thus, 1.11 % of the peaks will be above 3 sigma for a signal whose peaks follow the Rayleigh distribution. 29 Rayleigh Peak Response Formula Vibrationdata Consider a single-degree-of-freedom system with the index n. The maximum response can be estimated by the following equations. cn 2 ln fn T Cn cn Maximum Peak fn T ln n 0 . 5772 cn Cn n is the natural frequency is the duration is the natural logarithm function is the standard deviation of the oscillator response 30 Unit 4 Exercise 1 Vibrationdata Consider an avionics component. It is powered and monitored during a bench test. It passes this "functional test." Nevertheless, it may have some latent defects such as bad solder joints or bad parts. A decision is made to subject the component to a base excitation test on a shaker table to check for these defects. Which would be a more effective test: sine sweep or random vibration? Why? Reference: NAVMAT P9492, Section 3.1 31 Unit 4 Exercise 2 Vibrationdata Repeat the pervious examples on your own. Use the vibrationdata.m GUI script. Generate white noise vibrationdata > Miscellaneous Functions > Generate Signal > white noise Statistics vibrationdata > Signal Analysis Functions > Statistics Find probability from Normal distribution curve vibrationdata > Miscellaneous Functions > Statistical Distributions > Normal 32 Unit 4 Exercise 2 (cont) Vibrationdata SDOF Response vibrationdata > Signal Analysis Functions > SDOF Response to Base Input Estimated Peak Response from Rayleigh distribution vibrationdata > Miscellaneous Functions > SDOF Response: Peak Sigma 33