Motor Fault Acoustic Detection System Based on LabVIEW

　　A motor fault sound frequency detection system was built using virtual instrument technology. The system is based on the LabVIEW virtual instrument software platform, which realizes the real-time acquisition, analysis and post-processing of motor sound signals to achieve the purpose of detecting faulty motors. In addition to power spectrum and 1/3 octave analysis, the system uses the Matlab script node method to complete wavelet transform under the LabVIEW platform to extract the feature vector of the noise signal. Field applications show that the system works well.

　　1 Composition of Virtual Instrument System

　　Virtual instrument is a popular instrument configuration and detection control scheme in the world today. Virtual instrument is an open system that combines a computer platform with hardware modules with standard interfaces and development and testing software. It has the characteristics of good versatility and ease of use. Its typical hardware structure is: sensor - signal conditioner - data acquisition device - computer. The overall structure of the motor fault acoustic detection system is shown in Figure 1, which consists of a monitoring head (microphone, amplification and protection circuit), an audio card and a computer [1].

　　

 

　　Figure 1 Schematic diagram of the motor fault acoustic detection system

　　The monitoring head uses multiple microphones to pick up multi-point noise signals of the motor under test and convert the air vibration signals into electrical signals; the audio card uses a sound card to realize the mutual conversion between noise electrical signals (analog signals) and digital signals (WAV format); the computer records the WAV format digital signals and processes the waveform to determine whether any fault occurs.

　　2 Motor Fault Sound Detection Software System

　　LabVIEW is the originator of the concept of virtual instrument. This software platform integrates all the functions of hardware communication such as GPIB, VXL PXIRS-232 RS-485 and data acquisition card, and provides a large number of signal processing functions and signal analysis tools, which facilitates users to quickly and easily build virtual instrument test systems. Therefore, the software part of this system adopts the graphical software LabVIEW. The overall software structure block diagram of the system is shown in Figure 2.

　　

 

　　Figure 2: Software block diagram of motor fault acoustic detection system

　　The main tasks completed by the system software are:

　　(1) Display and record of motor noise signals;

　　(2) Signal analysis (including file analysis and real-time analysis) uses wavelet analysis and frequency domain analysis to alarm and display faults for abnormal signals:

　　(3) Saving and printing of documents;

　　(4) Motor and sensor parameter settings.

　　2.1 Signal Acquisition

　　This system uses a sound card as a noise collection tool. In terms of resolution, the general microcomputer multimedia sound card is 16 bits and the sampling frequency is 44.1/48kHz. Most mainstream mid-to-high-end sound cards have a sampling accuracy of 96kHz/24 bits, and some even reach 32 bits. The noise level and total harmonic distortion are higher, surpassing the indicators of most analog devices, and the price is relatively cheap. Therefore, it is feasible to use a sound card in the system.

　　LabVIEW provides a complete sound card control module. This article selects the "Sound Input" module. This module contains multiple functions to implement the settings, start, acquisition, stop and clear memory of the sound card. Figure 3 shows a block diagram of the sound acquisition program for a channel. The channel parameters in the system are set as follows: input is single channel, 16-bit sampling bit, 44.1kHz sampling frequency; output is 16-bit single channel. Figure 4 shows the noise signal of a motor with electromagnetic fault.

　　

 

　　Figure 3 Sound collection program flowchart

　　

 

　　Figure 4 Noise signal of a motor with electromagnetic fault

2.2 Signal analysis

 

　　After the signal is collected, the program automatically processes and analyzes it. The power spectrum and 1/3 frequency spectrum of the noise signal are obtained by frequency domain analysis. This is convenient for testers to observe. Figure 4 is the power spectrum of the motor in Figure 3. The "Sound and Vibration" toolkit can be purchased for LabVIEW software, which can easily perform various related analyses. Figure 5 shows the power spectrum of the noise of a motor with electromagnetic fault.

　　

 

　　Figure 5 Noise power spectrum of a motor with electromagnetic fault

　　The motor is a very complex mechanical system, and its noise signal contains rich information about the equipment status. Since the measured sound information contains various components and interference, it is a non-stationary signal, and the traditional filtering method based on Fourier transform has contradictions in improving the signal-to-noise ratio and spatial resolution. Therefore, the system uses wavelet technology as a feature extraction tool. The main methods used are:

　　(1) Wavelet transform soft threshold denoising method. In the one-dimensional signal denoising algorithm, the most critical part is the selection of threshold and the quantization of threshold. The soft threshold denoising method can extract signal features more accurately.

　　(2) Arbitrary scale reconstruction of wavelet transform: continuous wavelet can be reconstructed at a selected scale to extract signal features.

　　In LabVIEW, you can use an external signal processing toolkit to implement wavelet transform, or you can call Matlab in LabVIEW, that is, the Matlab Script node. The Matlab Script node allows users to import .m programs into the flowchart, and edit the M program in the flowchart according to the syntax of the Matlab program. In this way, users can use the powerful numerical calculation function of Matlab in LabVIEW.

　　When using Matlab script nodes, you must pay attention to the following: ① Matlab script nodes can only be used on Windows platforms; ② Matlab must be installed on the machine to use Matlab script nodes; ③ When combining LabVIEW and Matlab, you must pay attention to the matching of data types inside and outside the Matlab script node, otherwise errors or wrong information will be generated when LabVIEW is running.

　　Figure 6 shows the noise signal of the rear bearing scraping fault. The signal is denoised using the function in Matlab and the soft threshold filtering algorithm, and the denoising effect is shown in Figure 7. It can be seen that a good denoising effect is achieved in some sudden changes or peaks.

　　

 

　　Figure 6 Noise signal of a rear bearing scraper motor

　　

 

　　Figure 7 Signal of a rear bearing scraper motor after wavelet denoising

　　Determining the characteristic value of motor fault is one of the keys to fault diagnosis. The energy of the component signals in each frequency band contains a wealth of fault information, and the energy of certain frequency bands contains certain fault characteristics. This paper uses multi-resolution analysis to perform wavelet decomposition of the noise signal, and uses the energy characteristic value of each frequency band as the criterion to gradually perform fault diagnosis from low frequency to high frequency. Figure 8 shows the third-layer high-frequency coefficients of the wavelet decomposition of a motor with a bearing fault. From Figure 8, we can clearly observe the moment when the fault occurs, and we can also clearly capture the fault characteristic information of the noise signal in different frequency bands for feature extraction.

　　

 

　　Figure 8 Wavelet decomposition coefficients of a motor with bearing failure