Method for calculating bottom noise characteristics of continuous type band impact acoustic emission signal

Abstract

The invention belongs to the technical field of acoustic emission signal analysis, and particularly relates to a continuous impact acoustic emission signal floor noise feature calculation method. The acoustic emission waveform signal mainly has two forms, namely an impact signal and bottom noise, and the proportion of the bottom noise is larger, so that the distribution of all characteristic values obtained by calculating the characteristic values of the acoustic emission signal in a window with a certain width and drawing a sliding window is obtained, and the position where a first peak value appears in the distribution is the effective value of the bottom noise. The ground noise signal is also a high-frequency signal, and the sampling rate of the acoustic emission signal reaches 1M or even higher when the rotary machine fault diagnosis is carried out by using an acoustic emission monitoring means, so that the ground noise signal also contains abundant information, and the characteristic value of the ground noise comprises the characteristics of an effective value, a kurtosis value and the like. And calculating the continuous acoustic emission signal characteristics. The ground noise signal features are used for evaluating the indexes of the early abnormal state of the rotating machine.

Description

Technical Field

[0001] The present invention belongs to the technical field of acoustic emission signal analysis, and particularly relates to a method for calculating the background noise characteristics of continuous impact acoustic emission signals. Background Art

[0002] The acoustic emission monitoring technology has high sensitivity and can effectively identify the stress wave signals released during the event. At present, it has been applied to the crack and leakage monitoring of pipelines and containers. Such acoustic emission signals belong to burst-type signals, and the biggest feature of burst-type acoustic emission signals is small noise interference. However, when applying the acoustic emission technology to the fault diagnosis of rotating machinery, external noise is inevitably introduced. On the one hand, the characteristics of the background noise reflect the magnitude of the external noise. On the other hand, it reflects the characteristic energy introduced by the fault. However, due to the typical impact signal characteristics commonly existing in acoustic emission signals, there are certain difficulties in determining the characteristics of the background noise signal. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for calculating the background noise characteristics of continuous impact acoustic emission signals and use the background noise signal characteristics as an index for evaluating the early abnormal state of rotating machinery.

[0004] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] A method for calculating the background noise characteristics of continuous impact acoustic emission signals, the specific steps are as follows:

[0006] (1) Set a sliding window W with a length of N, and set the overlap rate of the sliding window to ε;

[0007] (2) Set an empty eigenvalue list Charater = [], and continuously add eigenvalues to the list through calculation;

[0008] (3) Starting from the first data of the acoustic emission signal, incorporate the first N values in the acoustic emission signal sequence Data = [X 1 , X 2 , …, X i , …] into the sliding window W to form the signal sequence Data_W1 within the sliding window = [X 1 , X 2 , …, X i , …X N , calculate the eigenvalue of the signal sequence Data_W1 within the sliding window, calculate the root mean square value according to formula (1), and add it to the eigenvalue list to obtain Charater = [RMS 1 , and the formula for calculating the root mean square value RMS of the data sequence [X 1 , X 2 , …, X N is as follows;

[0009]

[0010] (4) Slide the calculation window \(W\) backward along the acoustic emission signal sequence \(Data = [X 1 , X 2 , …, X i , …]\) according to the overlap rate \(\varepsilon\). The sliding distance \(STEP\) is calculated by formula (2):

[0011] \(STEP=\max(\lfloor(1 - \varepsilon)\times N\rfloor, 1)\ (2)

[0012] where \(\lfloor\ \rfloor\) represents rounding down to an integer; if the calculated \(STEP\) is 0, then \(STEP\) is taken as 1;

[0013] (5) Slide the window \(W\) according to the sliding distance \(STEP\) to obtain a new sliding window data sequence \(Data\_W2 = [X 1+STEP , X 2+STEP , …, X N+STEP \), and recalculate the eigenvalues of the sliding window data sequence \(Data\_W2\), including the effective value shown in formula (1), and add it to the eigenvalue list to obtain \(Charater = [RMS 1 , RMS 2 ;

[0014] (6) Repeat steps (4) and (5) until the last element \(X 1+(K-1)*STEP \) in the last \(K\) - th sliding window data sequence \(Data\_WK = [X 2+(K-1)*STEP , X N+(K-1)*STEP \) exceeds the last element of the signal data sequence \(Data = [X N+(K-1)*STEP , X 1 , …, X 2 , …]\). The sliding window terminates. Add the remaining values of the acoustic emission signal sequence to the \(K\) - th sliding window data sequence \(Data\_WK\), calculate the corresponding eigenvalues, and add them to the eigenvalue list. Finally, obtain the eigenvalue list \(Charater = [RMS i , RMS 1 , …, RMS 2 , …, RMS K ;

[0015] (7) Divide the eigenvalues into \(M\) equal parts from the minimum value to the maximum value, and plot the distribution histogram of the feature list \(Charater = [RMS 1 , RMS 2 , …, RMS K \). The abscissa represents the eigenvalues, and the calculation accuracy \(\gamma\) of the eigenvalues is calculated by the formula shown in (3):

[0016] γ = (max(Charater) - min(Charater)) / M (3)

[0017] (8) According to the abscissa of the first highest point of the distribution histogram of the feature list, the eigenvalue of the acoustic emission background noise signal can be obtained.

[0018] Selection of the length N of the sliding window W: Typical acoustic emission signals include the noise length S and the impact signal length L; the length N cannot exceed the length S of the background noise signal between two impact signals at most; initialize the length N of the sliding window W to the minimum of the impact signal length L and 0.25 times S, that is, N = min(L, 0.25 * S); make the window function slide in the background noise without impact signals, calculate the background noise eigenvalues within the sliding window, and form a background noise eigenvalue sequence L = [C 1 , C 2 , …, C M , calculate the variance of the background noise feature sequence. If the variance exceeds the set value Ω, it means that the eigenvalue waveform is large and the window length is too short. Increase the window function length by 10% and repeat the above process until the variance does not exceed the set value Ω.

[0019] Selection of the overlap rate ε: Initialize ε to 0.99 by default. If the drawn histogram is not smooth enough, increase the overlap rate ε; if the histogram is very smooth but the calculation time is long, appropriately reduce the overlap rate ε.

[0020] After the rotating equipment is installed, the form of the acoustic emission signal is determined. After selecting the appropriate length N and overlap rate ε, it can be directly used for the analysis of the continuous acoustic emission monitoring signal and the regular inspection acoustic emission signal of the same type of rotating equipment.

[0021] The beneficial effects achieved by the present invention are as follows:

[0022] Taking the monitoring and diagnosis of different states of mechanical seals as an example, analyze the background noise value of the acoustic emission signal, and then judge what state the mechanical seal is in. The first peak of the eigenvalue distribution histogram within the window drawn according to the eigenvalue list is the eigenvalue of the background noise. The background noise values calculated for different states of the mechanical seal show different background noise eigenvalues in different states. By calculating the background noise eigenvalues, the state of the mechanical seal can be analyzed and judged. Therefore, the background noise eigenvalue can be used as a typical index for the condition monitoring and diagnosis of rotating machinery. Specific embodiments

[0023] The present invention will be described in detail below with reference to specific embodiments.

[0024] When applying acoustic emission monitoring technology to the early fault monitoring and diagnosis of rotating machinery, the acoustic emission signals are mainly background noise signals, but there are also obvious impact signals. The impact signals are significantly higher than the background noise signals, and the patterns of the impact signals are not obvious. The impact signals cannot be effectively filtered out by means such as filtering. How to bypass the influence of the impact signals and calculate the characteristic values of the background noise signals is the problem solved by this invention.

[0025] There are mainly two forms in the acoustic emission waveform signals, namely impact signals and background noise. Among them, the proportion of background noise is larger. Therefore, by calculating the characteristic values of the acoustic emission signals within a window of a certain width and plotting the distribution of all the characteristic values obtained from the sliding window, the position where the first peak appears in the distribution is the effective value of the background noise.

[0026] The background noise signals are also high-frequency signals. When using acoustic emission monitoring means to carry out fault diagnosis of rotating machinery, the sampling rate of the acoustic emission signals reaches 1M or even higher. Therefore, the background noise signals also contain rich information. The characteristic values of the background noise include characteristic values such as effective values and kurtosis values. Taking the calculation of the effective value characteristic of the background noise signals as an example, this invention calculates the characteristics of continuous acoustic emission signals. The specific steps are as follows:

[0027] (1) Set a sliding window W with a length of N, and set the overlap rate of the sliding window to ε;

[0028] (2) Set an empty characteristic value list Charater = [], and continuously add characteristic values to the list through calculation;

[0029] (3) Starting from the first data of the acoustic emission signal, incorporate the first N values in the acoustic emission signal sequence Data = [X 1 , X 2 , …, X i , …] into the sliding window W to form the signal sequence Data_W1 within the sliding window = [X 1 , X 2 , …, X i , …X N . Calculate the characteristic values of the signal sequence Data_W1 within the sliding window, calculate the effective value according to formula (1), and add it to the characteristic value list to obtain Charater = [RMS 1 . The formula for calculating the effective value RMS of the data sequence [X 1 , X 2 , …, X N is as follows;

[0030]

[0031] (4) Along the acoustic emission signal sequence Data = [X 1 , X 2, …, X i , …] Slide the calculation window W backward, and the sliding distance STEP is calculated by formula (2):

[0032] STEP = max(Floor((1 - ε) × N), 1) (2)

[0033] where Floor represents taking the integer part downward; if the calculated STEP is 0, then STEP is taken as 1;

[0034] (5) Slide the window W according to the sliding distance STEP to obtain a new sliding window data sequence Data_W2 = [X 1+STEP , X 2+STEP , …, X N+STEP , and recalculate the eigenvalues of the sliding window data sequence Data_W2, including the effective value shown in formula (1), and add it to the eigenvalue list to obtain Charater = [RMS 1 , RMS 2 ;

[0035] (6) Repeat steps (4) and (5) until the last element X 1+(K-1)*STEP in the last Kth sliding window data sequence Data_WK = [X 2+(K-1)*STEP , X N+(K-1)*STEP exceeds the last element of the signal data sequence Data = [X N+(K-1)*STEP , X 1 , …, X 2 , …, X i , …], the sliding window terminates, add the remaining values of the acoustic emission signal sequence to the Kth sliding window data sequence Data_WK, and calculate the corresponding eigenvalues (such as the effective value), and add it to the eigenvalue list. Finally, obtain the eigenvalue list Charater = [RMS 1 , RMS 2 , …, RMS K ;

[0036] (7) Divide the eigenvalues into M equal parts from the minimum value to the maximum value, and draw the distribution histogram of the feature list Charater = [RMS 1 , RMS 2 , …, RMS K , the abscissa represents the eigenvalue, and the calculation formula of the calculation accuracy γ of the eigenvalue is as shown in formula (3):

[0037] γ = (max(Charater) - min(Charater)) / M (3)

[0038] (8) According to the abscissa of the first highest point of the histogram of the feature list distribution, the eigenvalue of the acoustic emission background noise signal can be obtained.

[0039] (9) Selection of the length N of the window W. The selection of N should not be too large nor too small. Typical acoustic emission signals include the noise length S and the impact signal length L.

[0040] The selection principle of the sliding window W is as follows:

[0041] When W is too long, the background noise and the impact signal will appear in the window simultaneously, resulting in the calculated feature being the comprehensive feature of the background noise and the impact signal when the window slides. Therefore, the maximum length of the window should not exceed the length S of the background noise signal between two impact signals, and to ensure that when the window slides, in most cases, only the noise signal is included;

[0042] When the length of the window is too short, the calculation of the eigenvalue in the window will be unstable, and the calculated eigenvalue cannot truly reflect the eigenvalue of the background noise signal. When the window W slides in the background noise, if the fluctuation of the eigenvalue calculated from the background noise in the window is too large, it indicates that the length of the window W is too short. The length N of the window W can be confirmed according to the following principle:

[0043] a. Initialize the length N of the window W as the minimum value of the impact signal length L and 0.25 times S, that is, N = min(L, 0.25 * S)

[0044] b. Make the window function slide in the background noise that does not contain the impact signal, calculate the eigenvalue of the background noise in the sliding window, and form a sequence of background noise eigenvalues L = [C 1 , C 2 , …, C M , calculate the variance of the background noise eigenvalue sequence. If the variance exceeds the set value Ω, it means that the eigenvalue waveform is large and the window length is too short. The length of the window function can be increased by 10%, and the above process can be repeated until the variance does not exceed the set value Ω.

[0045] (10) Selection of setting the overlap rate ε of the sliding window. Theoretically, the larger the ε is, the better, but too large will lead to a greater amount of calculation. When initializing, ε can be defaulted to 0.99. If the drawn histogram is not smooth enough, increase the overlap rate ε of the sliding window; if the histogram is very smooth but the calculation time is long, the overlap rate ε of the sliding window can be appropriately reduced;

[0046] (11) After the rotating equipment is installed, the form of the acoustic emission signal is basically determined. After selecting the appropriate sliding window length N and the sliding window overlap rate ε according to steps (9) and (10), this method can be directly used for the analysis of continuous acoustic emission monitoring signals and regular inspection acoustic emission signals of the same type of rotating equipment.

Claims

1. A calculation method for the background noise characteristics of a continuous impact acoustic emission signal, characterized in that: The specific steps are as follows: (1) Set a sliding window W with a length of N, and set the overlap rate of the sliding window to ε; (2) Set an empty eigenvalue list Charater = [], and continuously add eigenvalues to the list through calculation; (3) Starting from the first data of the acoustic emission signal, incorporate the first N values in the acoustic emission signal sequence Data = [X 1 , X 2 ,...., X i ,...] into the sliding window W to form the signal sequence Data_W1 within the sliding window = [X 1 , X 2 ,...., X i ,...X N . Calculate the eigenvalue of the signal sequence Data_W1 within the sliding window, calculate the effective value according to formula (1), and add it to the eigenvalue list to obtain Charater = [RMS 1 . The calculation formula for the effective value RMS of the data sequence [X 1 , X 2 ,..., X N is as follows; (4) Slide the calculation window W backward along the acoustic emission signal sequence Data = [X 1 , X 2 ,...., X i ,...] according to the overlap rate ε, and the sliding distance STEP is calculated by formula (2): STEP = max(Floor((1 - ε) × N), 1) (2) where Floor represents rounding down to an integer; if the calculated STEP is 0, then STEP takes 1; (5) Slide the window W according to the sliding distance STEP to obtain a new sliding window data sequence Data_W2 = [X 1+STEP , X 2+STEP ,...., X N+STEP , and recalculate the eigenvalues of the sliding window data sequence Data_W2, including the effective value shown in equation (1), and add it to the eigenvalue list to obtain Charater = [RMS 1 , RMS 2 ; (6) Repeat steps (4) and (5) until the last element X 1+(K-1)*STEP in the last sliding window data sequence Data_WK = [X 2+(K-1)*STEP , X N+(K-1)*STEP exceeds the last element of the signal data sequence Data = [X N+(K-1)*STEP , X 1 , X 2 ,...., X i ,...]. The sliding window terminates. Add the remaining values of the acoustic emission signal sequence to the K-th sliding window data sequence Data_WK, calculate the corresponding eigenvalues, and add them to the eigenvalue list. Finally, obtain the eigenvalue list Charater = [RMS 1 , RMS 2 , …, RMS K ; (7) Divide the eigenvalues into M equal parts from the minimum value to the maximum value, and plot the feature list Charater = [RMS 1 , RMS 2 , …, RMS K distribution histogram, where the abscissa represents the eigenvalue, and the calculation formula for the calculation accuracy γ of the eigenvalue is shown in (3): γ = (max(Charater) - min(Charater)) / M (3) (8) According to the abscissa of the first highest point of the distribution histogram of the feature list, the eigenvalue of the acoustic emission background noise signal can be obtained.

2. The calculation method for the background noise characteristics of a continuous impact acoustic emission signal according to claim 1, characterized in that: Selection of the length N of the sliding window W: Typical acoustic emission signals include the noise length S and the impact signal length L; the maximum length N cannot exceed the length S of the background noise signal between two impact signals; initialize the length N of the sliding window W as the minimum of the impact signal length L and 0.25 times S, that is, N = min(L, 0.25*S); make the window function slide in the background noise without impact signals, calculate the background noise eigenvalue in the sliding window, and form a background noise eigenvalue sequence L = [C 1 , C 2 , …, C M , calculate the variance of the background noise feature sequence. If the variance exceeds the set value Ω, it means that the eigenvalue waveform is large and the window length is too short. Increase the window function length by 10% and repeat the above process until the variance does not exceed the set value Ω.

3. The calculation method for the background noise characteristics of a continuous impact acoustic emission signal according to claim 1, characterized in that: Selection of the overlap rate ε: At initialization, ε is defaulted to 0.

99. If the drawn histogram is not smooth enough, increase the overlap rate ε; If the histogram is very smooth but the calculation time is long, appropriately reduce the overlap rate ε.

4. The calculation method for the background noise characteristics of a continuous impact acoustic emission signal according to claim 1, characterized in that: After the rotating equipment is installed, the form of the acoustic emission signal is determined. After selecting the appropriate length N and overlap rate ε, it can be directly used for the analysis of the continuous acoustic emission monitoring signal and the regular inspection acoustic emission signal of the same type of rotating equipment.

Images