MACHINERY VIBRATION

Machinery Vibration Monitoring and Analysis

Machinery Vibration Monitoring and Analysis

Introduction

Vibration analysis is a basic approach for fault diagnosis of rotating machinery. When a fault begins to develop, the vibration profile will change its characteristic shape. By using an appropriate signal processing method, it is feasible to detect changes in vibration signals caused by faulty components and to judge the conditions of the machinery. Traditional analysis has generally relied upon spectrum analysis based on FFT or STFT.

Fourier analysis is suitable for stationary signal processing, but provides a poor representation of signal well localised in time. This limitation of the Fourier transform therefore led to the introduction of time-frequency or time-scale signal processing tools. Wavelet transform, a new method for time-varying or non-stationary signal analysis, has attracted the attention of many researchers recently. Wavelet theory provides a unified framework for a number of techniques that had been developed independently for various signal processing applications.

Now wavelet theory has been used widely in fault diagnosis of rotating machinery, but few researches could be found on how to identify faults quantitatively with wavelet techniques.

Typical faulty signals of an experimental rotor are analysed in this paper based on continuous wavelet transform (CWT). The CWT scalogram could be used to distinguish these faults but without impersonality. Two features, wavelet grey moment (WGM) and wavelet grey moment vector (WGMV), are presented to diagnose faults automatically. The effectiveness of these two features is demonstrated by test data.

The faults discussed in this paper are imbalance (IB), misalignment (MA), oil whipping (OW), shaft crack (SC), pedestal looseness (PL) and rotor/stator friction (RSF). The first four faults are detected by shaft vibration signals, the last two are detected both by shaft vibration signals and bearing shell vibration signals. Thus, in this paper, pedestal looseness of shaft vibration signals is abbreviated to 'PL1', pedestal looseness of shell vibration signals to 'PL2', rotor/stator friction of shaft vibration to 'RSF1' and rotor/stator friction of shell vibration to 'RSF2'. Additionally, normal signals (NS) detected by shaft vibration signals are analysed too. All signal data come from the faults simulative tests with an experimental rotor in Huazhong university of Science & Technology.

Experiments

Arrangement of experimental rotor

The experimental rotor is shown schematically in Fig. 1. It is a double span shafting with three rotary tables in one span and two rotary tables in another. The shafting is supported by four sliding bearings. The overall length of the shaft is 650 mm and shaft diameter is 8 mm. Thickness and diameter of the rotary table are 12 and 70 mm, respectively. The shafting is driven by a direct current electromotor, giving an output of 0-10000 rpm.

Fig. 1. Experimental test rig.

This rig mounts four eddy current probes close to the shaft and an accelerometer on the bearing housing. The signals from the sensors are fed into a signal preprocessing device and then sampled by synchronous sampling. The sampling frequency is 32 times the basic frequency of the rotor: the sampled signals, each containing 512 data points, are carefully stored in ...