Parameter scheme for abnormal noise analysis of electric components (I)

Modulation, Envelope

If there is obvious amplitude modulation (AM) in the noise, BK Connect software also provides envelope analysis (Envelope). Envelope analysis can separate the modulated signal from the carrier signal. As shown in the figure below, the red is the carrier signal and the black is the modulated signal. After envelope analysis, the modulation frequency fm in the lower right corner can be obtained.

In addition to being used for demodulation analysis of modulated signals, envelope analysis is also suitable for fault detection of motor bearings. It can identify weak fault signals from bearing vibration noise signals and detect faults in the early stages of a fault.

As shown in the example on the right of the figure below, after bandpass filtering the frequency band where the carrier is located (125±100Hz), FFT analysis is performed on the envelope time history curve. The Y-axis amplitude is the spectrum amplitude of the envelope. Obvious peaks are found in the spectrum, indicating that there is amplitude modulation in the signal. In the FFT analysis results of the noise signal, as shown in the left figure below, we found that there are adjacent peaks at 126Hz and 120Hz, where the higher amplitude 126Hz is the carrier frequency, and the lower amplitude 120Hz is the modulation signal frequency. According to the ratio of the modulation signal amplitude to the carrier signal amplitude, the degree of modulation can be calculated, which is 0.082/0.395≈21%.

If the frequency range of the carrier frequency cannot be determined before analysis, envelope analysis of different frequency bands can be performed first to investigate whether there is modulation in the data of these frequency bands. As shown in the example below, the X-axis horizontal coordinate is the modulation frequency, the Y-axis vertical coordinate on the left is the center frequency of the carrier frequency band of the envelope analysis, and the Z-axis color coordinate is the modulation signal amplitude. From the figure, it can be found that there are obvious peaks at the 2500Hz carrier frequency band and 80Hz modulation frequency (the yellow circle position in the figure), and at other carrier frequency bands, such as 500Hz, there are also multiple modulation frequency peaks (the orange circle position in the figure), indicating that there is obvious modulation at these positions.

***Tone-to-noise Ratio, Prominence Ratio***

(ProminenceRatio), Tonality

Due to the structural characteristics of the motor and transmission mechanism, the rotating mechanism often emits some pure tones. These pure tones may make the motor noise sound sharp, harsh, discordant, and irritating. When analyzing the pure tone problem, it is necessary to consider the masking effect of the broadband signal. When there is a pure tone in a broadband signal, the sound pressure level amplitude of the pure tone exceeds the sound pressure level amplitude of the critical frequency band where the pure tone is located by more than 6 dB, and the human ear can hear this pure tone.

If it is less than 6dB, the human ear cannot hear the pure tone, which is masked by the broadband sound in the surrounding frequency band. The figure below shows the legend and definition of pure tone ratio and prominence ratio. In view of this characteristic of the human ear, BK Connect software provides a variety of pure tone analysis parameters, including pure tone ratio (Tone-to-noise Ratio), prominence ratio (Prominence Ratio) and tone (Tonality), which can automatically determine whether there is an obvious pure tone in the noise.

FFT, 1/3 Octave,

Critical Band

FFT and 1/3 octave are mostly used to describe the distribution of sound energy in the frequency domain. Their ordinate is the sound amplitude, such as sound pressure level dB, A-weighted sound pressure level dB(A), etc., and the abscissa is the frequency. One of the most obvious differences between FFT and 1/3 octave is the resolution of the frequency coordinate. The frequency resolution of FFT is consistent at all frequencies (as shown in the left figure below), while the frequency resolution of 1/3 octave is a fixed percentage of the center frequency (about 23%). The lower the frequency, the narrower the frequency band represented by each bar graph, that is, the smaller the frequency resolution. Conversely, the higher the frequency, the wider the frequency band and the greater the frequency resolution (as shown in the figure below).

The essential difference between FFT and 1/3 octave is the calculation method of noise amplitude. FFT uses the principle of Fourier transform to convert time domain to frequency domain to obtain the noise amplitude of each frequency. 1/3 octave uses the time domain filter method to perform bandpass filtering on the time domain signal to obtain the noise amplitude of each frequency band.

FFT and 1/3 octave mainly describe the energy distribution of sound, without considering the frequency domain masking effect in psychoacoustics, so there is a certain gap between them and people's subjective perception of sound. In order to reduce this gap, critical bands (unit bark) are usually used to describe noise of different frequencies when performing psychoacoustic analysis. Critical bands divide the audible frequency band of human ears into 0-24 bark (as shown in the right figure below).

More parameters

If the noise signal has non-steady-state characteristics or even transient characteristics, in addition to the above common parameters, other parameters are often used. They are derived from conventional parameters and are also often used to evaluate the abnormal noise of electric components.

Percentile

When studying noise problems, in addition to the characteristic components related to its own formation mechanism, noise often contains random components, which makes the noise data often distributed within a certain range of values. In order to facilitate the quantification of the distribution characteristics of noise data, some statistical methods are introduced, such as percentiles. Common percentiles include the 1st, 50th (also called the median), and 99th percentiles.

Taking three sets of data from repeated tests as an example, in the loudness time history in the figure below, due to the influence of factors such as the operating status of the object under test, it is impossible to guarantee that the maximum values ​​of the three sets of data are the most stable and consistent. The table summarizes the 0th to 15th percentiles of the three sets of data, where the standard deviation of the 0th percentile (the third row, i.e. the maximum loudness) is 0.33Sone, while the standard deviations of the 3rd and 4th percentiles (the fourth and fifth rows) are significantly smaller than the standard deviation of the 0th percentile, so the 3rd and 4th percentiles have higher consistency.

(Characteristic Loudness Integrated Area) Percentile Frequency

When using critical bands to describe the loudness of different frequencies, we call the result characteristic loudness, as shown in the following figure, the horizontal axis is the critical band (bark), and the vertical axis is the loudness. When studying the frequency domain distribution characteristics of noise, the characteristic loudness curve can be used to calculate the integral area.

The frequency corresponding to the percentile of the integrated area is then used to describe the spectral distribution characteristics. The percentile result is the characteristic frequency, in Bark. Take the following figure as an example. The left and right figures are the characteristic loudness data of two different sounds. The total loudness of the two sounds is 13.00 Sone and 12.74 Sone, which are relatively close, but their spectral distribution is different. For example, near 9 Bark, the value of the left figure is smaller than that of the right figure, and near 16.5 Bark, the value of the left figure is larger than that of the right figure. Taking the 50th percentile frequency as an example, the integrated area below a certain characteristic frequency accounts for 50% of the total integrated area. It can also be described as the characteristic frequency integrated area above and below a certain critical frequency band is equal, that is, the characteristic loudness "center of gravity". The characteristic frequency corresponding to this critical frequency band is the 50th percentile frequency.

In the above two sets of data, the 50th percentile frequencies of the left and right figures are 12.5 Bark and 11 Bark respectively (the red vertical lines in the figure), which means that the characteristic loudness "center of gravity" of the left figure is more to the right than that of the right figure, indicating that the characteristic loudness is mainly concentrated in the higher frequency band. In addition to using the 50th percentile frequency, other percentiles (such as the 70th percentile) can also be used to represent the distribution characteristics of the characteristic loudness.

For transient noise signals, in the time history of transient noise, the frequency distribution changes with time. Taking the characteristic loudness time domain history when the motor stops running as an example, as shown in the figure on the left of the figure below, transient impact noise appears at 0.09-0.25 seconds (the time range in the blue box), and the characteristic loudness is different at different times. The percentile frequency is calculated using the characteristic loudness at each moment, thereby reflecting the frequency distribution change trend at each moment. The figure on the right of the figure below is the time history of the 70th percentile frequency, where 0 seconds is the peak moment of the impact noise, as the starting moment. The 70th percentile frequency curve gradually decreases over time, which means that the frequency components in the characteristic loudness gradually shift to lower frequencies over time.