Вы находитесь на странице: 1из 29

VIBRATIONAL ANALYSIS TO

DETECT FAULT IN POWERTRAIN


OF VEHICLE
SAURABH PATIL
121795025
GUIDE : K.P.WANI
Contents

 Causes of gearbox fault and defects


 Data collection
 Signal analysis techniques
 Time domain(Kurtosis) analysis
 Frequency domain analysis
 Demodulation( envelope analysis) Hilbert transform
 Cepstrum analysis
 Order analysis

2
Causes of gearbox fault and defect

1. Gear
 Tooth missing fault
 Misalignment
 Worn or cracked teeth
 Eccentric gear

2. Shaft
 Unbalance
 Bent shaft
 Misalignment
 Cracked shaft

3. Bearing
 Pitting of race or ball/roller
 Crack on race

3
Data collection

ADXL 335 Specifications


 3-axis sensing
 Acceleration with minimum full scale range of -3g to +3g
 Voltage accepted 3.3 V to 5 V

4
ADXL 335 program arduino

5
ADXL 335 data {Condition: First Gear (1500 rpm) Standing}

6
Excel to Matlab

7
Signal Classification

 Signals essentially convey information, features of which can be constant over a


period of time or vary with time.
 Classification
 Stationary Signal (e.g vibration signal from healthy gear box running at constant speed)
 Non Stationary Signal (e.g impact signal , vibration signal from healthy gear box
running at Variable speed)
 Static Signal (Constant with respect to time)
 Dynamic Signal (Changing with time)

8
Signal processing techniques

 Time domain (kurtosis)


 Frequency domain(side band due to modulation)
 Cepstrum ( group of sideband coming from one source detection) ex. Multistage gear
box
 Wavelet(Time frequency analysis to detect impact)
 SFFT(Time frequency analysis to detect impact)

9
Time domain technique

 Estimate feature
 For Gear Box
Kurtosis more than 6 means sure faulty system
 Only information related to presence of fault

Source: Mohanty, Amiya R. Machinery Condition Monitoring


Principles and Practices. 10
Matlab program for time domain analysis

11
Time domain (Feature estimation)

12
Time domain technique(Beating Phenomenon)

 Signal beating is a phenomenon that occurs when two signals that are very close in
frequency are present, and the resultant signal is given as a summation of such
independent signals.

Source: Mohanty, Amiya R. Machinery Condition Monitoring Principles


and Practices. 13
Time Domain Signal analysis techniques

 Waveform features
 Statistical signal processing (rms , mean etc. )
 Time domain synchronous averaging( to avoid signal smearing)

14
Need of frequency Analysis

 Every mechanical component has characteristic frequency


 Signature of machine component is unique
Characteristic Frequency:
1.Gear meshing frequency :For spur gears, the gear mesh frequency is the product of
the number of gear teeth and the rotational speed .
2. Ball bearing
The ball pass outer frequency (BPFO)
ball pass frequency inner (BPFI)
ball spin frequency (BSF)
fundamental train frequency (FTF)

Source: Mohanty, Amiya R. Machinery Condition Monitoring 15


Principles and Practices.
Frequency Domain Analysis

Frequency Estimation
 Signal Hetrodyning (Frequency oscillator to match frequency)
 Filter (low pass , high pass, band pass, notch filter )
 Fourier Series (Periodic) –mathematical form required and cannot handle random
signal
 Fourier Integral /transform (non periodic )-random /signal transient /periodic signals
 Discrete Fourier Transform (digitally implement Fourier transform) requires digitally
acquired time data (computer aided data acquisition)
 Fast Fourier Transform

16
Fast Fourier Transform

 Computational time required for DFT is more


 N log10 N computational operations were required instead of the N2 operations(DFT)
 N number sample time data converted to N/2 in frequency (mirror image)
Leakage Error : FFT analyzer take block of data and assumes it repeat itself
(Discontinuity in signal responsible for leakage error. Discontinuity in signal responsible for
leakage error.)

Source: Mohanty, Amiya R. Machinery Condition


Monitoring Principles and Practices 17
FFT in Matlab

 Matlab command : sptool IC engine F(max)= 5000Hz


Gear box = 5000 Hz
 File—Import to SPTool Ref: iitnoise

18
FFT Spectrum

19
FFT Spectrum Estimation

Amplitude Vs Time
Amplitude Vs Frequency
20
Representation of Fourier Transform data

 The Fourier transform of a real signal x(t) is a complex number X(f) = X(kΔf)
 the complex number can be represented in the real/imaginary format or the magnitude
phase format
 The magnitude and phase of a complex number are displayed in a bode plot, and the
imaginary versus real part of the Fourier transform is displayed in a Nyquist plot.

Source: Mohanty, Amiya R. Machinery Condition


Monitoring Principles and Practices
21
Representation of Fourier Transform data
Autopower Spectrum (Power Spectrum)---energy content in signal
The autopower spectrum of a signal x(t) is denoted as Sxx(f)

X *(f) is the complex conjugate of the Fourier transform of the signal x(t). The asterisk
symbol (*) denotes a complex conjugate operation.
For a transient signal, the frequency spectrum is continuous, and the power of such
signals is usually denoted by power spectral density, PSD.

• To express the linear spectrum magnitude of the above two quantities, the square roots of
Sxx(f) and PSD(f) are obtained.

22
Representation of Fourier Transform data

Cross- Power Spectrum


 When there are two signals, x(t) and y(t), present in a system, to estimate the relative
phase difference between the two signals the cross spectrum, Sxy(f) of the two signals
is estimated as

 The cross spectrum is a complex quantity


Frequency Response Function
Ratio of cross spectrum and auto power spectrum
Complex quantity

23
Signal Demodulation

 When the amplitude or frequency of a signal varies with time, it is called a modulated
signal.
 Machine rotating speed undergoes small variation in speed ,which means violation of
stationary signal property because of that after averaging Signal Smearing error
occurs.
 Envelope analysis (Hilbert Transform) is used for calculation of characteristics
frequency.
 Matlab Command : hilbert

24
FFT after Hilbert

25
Cepstrum Analysis

 When the families of sidebands in signals become many and the signals have a poor
signal-to-noise ratio, cepstrum analysis is used to recover such signals in the time
domain
 Y axis – Quefrency
 Matlab : cceps(xf)

26
Order analysis

X axis : rotation of shaft(1 rotation = 1 order)


 The key to sound and vibration analysis on rotating machine is measurement of
rotational speed of machine simultaneously using measurement of sound and vibration
Ex. Key phaser to check rotation and at same time accelerometer used for vibration.
 When analysis is related to rotational speed it is called order analysis.
Methods of order analysis :
1. FFT based order analysis
2. Time signal recording and short time fourier transform

27
Order analysis

 Waterfall diagram

Source: Mohanty, Amiya R. Machinery Condition


Monitoring Principles and Practices

28
THANK YOU

29

Вам также может понравиться