Continuous acoustic emission signal impact frequency and impact energy analysis method

Abstract

The invention belongs to the technical field of acoustic emission signal analysis, and particularly relates to a continuous acoustic emission signal impact frequency and impact energy analysis method. When the acoustic emission monitoring technology is used for early fault monitoring and diagnosis of the rotating machine, acoustic emission signals within a period of time are obtained, the number of impact events and impact energy distribution in the acoustic emission signals are counted, the number of impact events and impact energy distribution in different normal and abnormal states are analyzed, and the early fault diagnosis accuracy is improved. And thus, early failure monitoring and diagnosis of the rotating machinery are realized. Through impact statistics and impact energy distribution, identification and diagnosis of different operation states of the mechanical seal can be realized.

Description

Technical Field

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

[0002] The acoustic emission monitoring technology is highly sensitive and can effectively identify the stress wave signals released during an event. It has been applied to the crack and leakage monitoring of pipelines and containers, and such acoustic emission signals belong to burst-type signals. When cracks and damages appear on the surface of rotating machinery, impact signal characteristics will also be generated during the relative operation process. By using the acoustic emission monitoring technology to monitor the high-frequency impact signals released by the early minor faults of rotating machinery, the early faults of rotating machinery can be effectively identified. However, the acoustic emission signals of rotating machinery belong to continuous acoustic emission signals. Statistically analyzing the impact times and impact energy in the acoustic emission signals within a certain period of time is of great significance for identifying the early faults of rotating machinery. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for analyzing the impact times and impact energy of continuous acoustic emission signals, so as to monitor and diagnose the early faults of rotating machinery.

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

[0005] A method for analyzing the impact times and impact energy of continuous acoustic emission signals, the specific steps are as follows:

[0006] (1) Set a background noise sequence L with an initial length of N 0 =[k 1 , k 2 , …, k i , …, k N , and add the absolute value |X 1 | of the first value X 2 in the signal sequence Data = [X i , X 1 , …] to the background noise sequence L 1 to form a new background noise sequence L 0 = [|X 1 |, k 1 , …, k 2 , …, k i , …, k N , and calculate the average value of the background noise sequence L 1 according to formula (1):

[0007]

[0008] where y i is L 1Among the elements, except for y 1 , the rest of the elements are all 0;

[0009] (2) Calculate the absolute value |X 1 ,X 2 ,…,X i ,…] of the second value X 2 in the signal sequence Data = [X 2 |, and determine whether |X 2 | exceeds K times the average value of the noise floor sequence L 1 , that is, whether equation (2) holds:

[0010] |X 2 | > K * L 1_ave (2)

[0011] (3) If equation (2) holds, it means that X 2 deviates slightly more from the mean. To avoid affecting the mean, do not add |X 2 | to L 1 . Continue to calculate the absolute value |X 1 ,X 2 ,…,X i ,…] of the third value X 3 in the signal sequence Data = [X 3 |, and repeat the above steps;

[0012] (4) If equation (2) does not hold, it means that X 2 is within a reasonable range. Add |X 2 | to L 1 to form a new noise floor sequence L 2 = [|X 1 |,|X 2 |,…,k i ,…,k N , and calculate the average value L 2 of the new noise floor sequence L 2_ave according to equation (1); Continue to calculate the absolute value |X 1 ,X 2 ,…,X i ,…] of the third value X 3 in the signal sequence Data = [X 3 |, and repeat the above steps;

[0013] Repeat the above 4 steps of (1)-(4) until the noise floor sequence L is completely filled to form L N = [|X 1 |,|X 2 |,…,|X t |,…,|X N, and at the same time, the mean value L of the noise floor sequence is calculated through equation (1). N_ave ; Through the above four steps (1), (2), (3), and (4), the initialization of the noise floor sequence is completed;

[0014] (5) Set the identifier for whether an impact signal appears as Label. If no impact signal appears, then Label = False. If an impact signal appears, then Label = True. At the same time, define the impact signal energy list PulseEnergy, and initialize the identifier Label for whether an impact signal appears to be false, that is, Label = False, and initialize the impact signal energy list to be empty, that is, PulseEnergy0 = [].

[0015] (6) When the identifier Label for whether an impact signal appears is Label = False, continue to calculate the absolute value |X 1 , X 2 , …, X i , …] of the m-th value X m in the signal sequence Data, and determine whether |X m | exceeds M times the mean value of the noise floor sequence, that is, whether it satisfies equation (3): m |X

[0016] |X m | > M * L N_ave (3)

[0017] (7) If equation (3) holds, it means that X m deviates a lot from the mean value, and it is determined that there is an impact signal. Take |X m | as the first data point of the impact signal sequence L pulse , and form the impact signal sequence L pulse = [|X m |], and let the maximum value MAX(L pulse ) = |X m |, calculate the difference between the maximum value MAX(L pulse ) and the mean value L of the noise floor sequence N_ave , and take it as the amplitude GAP of the impact signal. The calculation formula for the amplitude GAP of the impact signal is shown in equation (4):

[0018] GAP = MAX(L pulse ) - L N_ave (4)

[0019] At the same time, set the identifier Label for whether an impact signal appears to Label = True;

[0020] (8) In step (7), if the impact signal identifier Label = True, continue to calculate the signal sequence Data = [X 1 ,X 2 ,…,X i ,…] the m+1th value X m+1 The absolute value of |X m+1 |, first determine |X m+1 |Is it greater than the maximum value MAX(L pulse ); if |X m+1 |Greater than MAX(L pulse ), then update the maximum value MAX(L pulse ) is |X m+1 |, and recalculate the updated impulse signal amplitude GAP according to formula (4); if |X m+1 |Not greater than MAX(L pulse ), then maintain the maximum value MAX(L pulse ) and the amplitude GAP of the impulse signal remain unchanged; then determine |X m+1 | and the mean value of the noise sequence L N_ave Whether the difference exceeds α times the GAP value, that is, it satisfies formula (5):

[0021] |X m+1 |- L N_ave > α*GAP (5)

[0022] (9) If |X m+1 | and the mean value of the noise sequence L N_ave The difference between |X satisfies formula (5), which means that m+1 |The energy is still within the range of impact energy attenuation, and |X m+1 | is added to the impulse signal sequence to form a new impulse signal sequence Lpulse=[|X m |,|X m+1 |], and continue to analyze the signal sequence Data = [X 1 ,X 2 ,…,X i ,…] the m+2th value X m+2 ;

[0023] (10) If |X m+1 | and the mean value of the noise sequence L N_ave The difference between |X does not satisfy formula (5), which means |X m+1 |The energy has basically decayed, indicating that the impact process has ended. m+1 |Add background noise sequence L N At the end of NThe value at the very front pops out to form the updated background noise sequence L N+1 = [|X 2 |, …, |X t |, …, |X N |, |X m+1 |] and the updated background noise sequence mean L N+1_ave , and at the same time set the impulse signal identifier Label to false, i.e., Label = False, and calculate the energy value Lpulse_Energy of the impulse signal sequence according to equation (6):

[0024]

[0025] Add L pulse _Energy to the impulse signal energy list PulseEnergy to form a new impulse signal energy list PulseEnergy1 = [L pulse _Energy]; on this basis, continue to analyze the (m + 2)-th value X 1 in the signal sequence Data = [X 2 , X i , …, X m+2 ;

[0026] (11) If equation (3) does not hold, it means that X m does not deviate far from the mean value and cannot be determined as the starting point of an impulse event. Add |X m | to the end of the background noise sequence L N , and at the same time pop out the value at the very front of the original background noise sequence L N to form the updated background noise sequence L N+1 = [|X 2 |, …, |X t |, …, |X N |, |X m |] and the updated background noise sequence mean L N+1_ave , and on this basis continue to analyze the (m + 1)-th value X 1 in the signal sequence Data = [X 2 , …, X i , …] according to step (6); m+1 ;

[0027] (12) Analyze the remaining data in the signal sequence Data = [X 1 , X 2 , …, X i , …] according to (6)-(11), and finally form the impulse signal energy list PulseEnergyM = [L pulse _Energy1, L pulse_Energy2,…,L pulse _EnergyM], the length of the impulse signal energy list PulseEnergyM is M, that is, the analyzed acoustic emission signal has M impulses. By drawing the distribution histogram of the impulse signal energy list PulseEnergyM, the distribution of the impulse energy can be obtained.

[0028] K is 2.

[0029] M is 5.

[0030] α is taken as 0.2.

[0031] The beneficial effects achieved by the present invention are:

[0032] Through impact statistics and impact energy distribution, the identification and diagnosis of different operating states of mechanical seals can be realized. The number of impact events and impact energy distribution of mechanical seals in different states are obviously different, which can be used to monitor and diagnose early abnormal states of mechanical seals. Improve the accuracy of impact signal detection. DETAILED DESCRIPTION

[0033] The present invention is described in detail below with reference to specific embodiments.

[0034] When the acoustic emission monitoring technology is used for early-stage fault monitoring and diagnosis of rotating machinery, the acoustic emission signals within a period of time are obtained, the number of impact events and the impact energy distribution in the acoustic emission signals are counted, and the number of impact events and the impact energy distribution under different normal and abnormal conditions are analyzed, thereby realizing the monitoring and diagnosis of early-stage faults of rotating machinery. The impact-type acoustic emission signal manifests itself as a sudden increase in the monitoring value, but since the acoustic emission signal itself has certain fluctuations, it is difficult to determine how much the acoustic emission signal exceeds as an impact event. The present invention analyzes and confirms the determination of the impact signal, the calculation of the impact signal energy, and the distribution of the impact signal energy. The specific steps are as follows:

[0035] (1) Set a background noise sequence L with an initial length of N 0 =[k 1 ,k 2 ,…,k i ,…,k N ], and the signal sequence Data = [X 1 ,X 2 ,…,X i ,…] the first value X 1 The absolute value of |X 1 |Add background noise sequence L 0 In the process, a new background noise sequence L is formed. 1 =[|X 1 |,k 2 ,…,ki , …, k N , and calculate the background noise sequence L according to formula (1) 1 for the average value of:

[0036]

[0037] where y i is an element in L 1 , and except for y 1 , the rest of the elements are all 0;

[0038] (2) Calculate the absolute value |X 1 | of the second value X 2 in the signal sequence Data = [X i , X 2 , …, X 2 , …], and determine whether |X 2 | exceeds K times (K can be taken as 2) the average value of the background noise sequence L 1 , that is, whether formula (2) holds:

[0039] |X 2 | > K * L 1_ave (2)

[0040] (3) If formula (2) holds, it means that X 2 deviates slightly more from the mean. To avoid affecting the mean, do not add |X 2 | to L 1 , and continue to calculate the absolute value |X 1 | of the third value X 2 in the signal sequence Data = [X i , X 3 , …, X 3 |, and repeat the above steps;

[0041] (4) If formula (2) does not hold, it means that X 2 is within a reasonable range. Add |X 2 | to L 1 to form a new background noise sequence L 2 = [|X 1 |, |X 2 |, …, k i , …, k N , and calculate the average value L 2 of the new background noise sequence L according to formula (1); continue to calculate the absolute value |X 2_ave | of the third value X 1 in the signal sequence Data = [X 2 , X i , …, X 3 | of the third value X3 , and repeat the above steps;

[0042] Repeat the above 4 steps (1)-(4) until the background noise sequence L is completely filled to form L N =[|X 1 |,|X 2 |,…,|X t |,…,|X N |], and at the same time calculate the mean value L of the background noise sequence through formula (1) N_ave ; Through the above four steps (1)(2)(3)(4), the initialization of the background noise sequence is completed.

[0043] (5) Set the identifier for whether an impact signal appears as Label. If no impact signal appears, then Label = False. If an impact signal appears, then Label = True. At the same time, define the impact signal energy list PulseEnergy, and initialize the identifier Label for whether an impact signal appears to false, that is, Label = False, and initialize the impact signal energy list to empty, that is, PulseEnergy0 = [];

[0044] (6) When the identifier for whether an impact signal appears Label = False, continue to calculate the absolute value |X 1 ,X 2 ,…,X i ,…] of the m-th value X m in the signal sequence Data, and judge whether |X m | exceeds M times (M can be taken as 5) of the mean value of the background noise sequence, that is, whether it satisfies formula (3): m |X

[0045] |>M*L m (3) N_ave (3)

[0046] (7) If formula (3) holds, it means that X m deviates a lot from the mean value, and it is determined that there is an impact signal. Take |X m | as the first data point of the impact signal sequence L pulse , and form the impact signal sequence L pulse =[|X m |], and let the maximum value MAX(L pulse ) = |X m |, calculate the difference between the maximum value MAX(L pulse ) and the mean value L of the background noise sequence N_ave , and take it as the amplitude GAP of the impact signal. The calculation formula for the amplitude GAP of the impact signal is shown in formula (4):

[0047] GAP = MAX(L pulse ) - L N_ave (4)

[0048] Meanwhile, set the identifier Label indicating the absence of impact signals to Label = True;

[0049] (8) When the impact signal identifier Label = True in step (7), continue to calculate the absolute value |X 1 , X 2 , …, X i , …] of the (m + 1)-th value X m+1 in the signal sequence Data. First, determine whether |X m+1 | is greater than the maximum value MAX(L m+1 ) of the existing impact signal sequence; if |X pulse | is greater than MAX(L m+1 ), then update the maximum value MAX(L pulse ) of the impact signal sequence to |X pulse |, and recalculate the amplitude GAP of the updated impact signal according to equation (4); if |X m+1 | is not greater than MAX(L m+1 ), then keep the maximum value MAX(L pulse ) of the impact signal sequence and the amplitude GAP of the impact signal unchanged; then determine whether the difference between |X pulse | and the mean value L m+1 of the background noise sequence exceeds α times (α can be taken as 0.2) of the GAP value, that is, satisfy equation (5): N_ave |X

[0050] |X m+1 | - L N_ave > α * GAP (5)

[0051] (9) If the difference between |X m+1 | and the mean value L N_ave of the background noise sequence satisfies equation (5), it means that the energy of |X m+1 | is still within the range of impact energy attenuation. Add |Xm + 1| to the impact signal sequence to form a new impact signal sequence Lpulse = [|Xm|, |Xm + 1|], and continue to analyze the (m + 2)-th value X 1 , X 2 , …, X i , …] in the signal sequence Data = [X m+2 ;

[0052] (10) If the difference between |X m+1 | and the mean value L N_aveIf the difference does not satisfy equation (5), it indicates that the energy of |X m+1 | has basically decayed, indicating that this impact process has ended. Add |X m+1 | to the end of the background noise sequence L N , and at the same time pop the value at the very front of the original background noise sequence L N to form the updated background noise sequence L N+1 = [|X 2 |,…,|X t |,…,|X N |, |Xm+1|] and the updated background noise sequence mean L N+1_ave . At the same time, set the impact signal identifier Label to false, i.e., Label = False, and calculate the energy value Lpulse_Energy of the impact signal sequence according to equation (6):

[0053]

[0054] Add L pulse _Energy to the impact signal energy list PulseEnergy to form a new impact signal energy list PulseEnergy1 = [L pulse _Energy]; On this basis, continue to analyze the (m + 2)-th value X 1 in the signal sequence Data = [X 2 , X i ,…, X m+2 according to step (6);

[0055] (11) If equation (3) does not hold, it indicates that X m does not deviate far from the mean value, and it cannot be determined as the starting point of an impact event. Add |X m | to the end of the background noise sequence L N , and at the same time pop the value at the very front of the original background noise sequence L N to form the updated background noise sequence L N+1 = [|X 2 |,…,|X t |,…,|X N |, |Xm|] and the updated background noise sequence mean L N+1_ave , and continue to analyze the (m + 1)-th value X 1 in the signal sequence Data = [X 2 , X i ,…, X m+1 according to step (6);

[0056] (12) Analyze the signal sequence Data = [X 1 , X 2, …, X i , …] to analyze the remaining data, and finally form the impact signal energy list PulseEnergyM = [L pulse _Energy1, L pulse _Energy2, …, L pulse _EnergyM]. The length of the impact signal energy list PulseEnergyM is M, that is, the analyzed acoustic emission signal has M impacts. By plotting the distribution histogram of the impact signal energy list PulseEnergyM, the distribution of the impact energy can be obtained.

Claims

1. A method for analyzing the impact times and impact energy of continuous acoustic emission signals, characterized in that: The specific steps are as follows: (1) Set a noise floor sequence \(L\) with an initial length of \(N\) 0 = [k 1 , k 2 , …, k i , …, k N , and take the absolute value \(|X 1 |\) of the first value \(X 2 \) in the signal sequence \(Data = [X i , X 1 , …, X 1 , …]\) and add it to the noise floor sequence \(L 0 \) to form a new noise floor sequence \(L 1 = [|X 1 |, k 2 , …, k i , …, k N , and calculate the average value of the noise floor sequence \(L 1 \) according to formula (1): where y i is an element in L 1 and all other elements except y 1 are 0; (2) Calculate the absolute value of the second value \(X_{ 2}\) in the signal sequence Data = \([X_{ 1}, X_{ 2}, \ldots, X_{ i}, \ldots]\), i.e., \(|X_{ 2}|\), and determine whether \(|X_{ 2}|\) exceeds \(K\) times the average value of the noise floor sequence \(L_{ 1}\), that is, whether Equation (2) holds:​​​​​​​​​​​​​​ |X 2 |>K*L 1_ave (2) (3) If equation (2) holds, it indicates that X 2 deviates slightly more from the mean. To avoid affecting the mean, |X 2 | is not added to L 1 . Continue to calculate the absolute value |X 1 | of the third value X 2 in the signal sequence Data = [X i , X 3 , …, X 3 , …], and repeat the above steps; (4) If equation (2) does not hold, it indicates that X 2 Within a reasonable range, add |X 2 | to L 1 to form a new noise floor sequence L 2 = [|X 1 |, |X 2 |, …, k i , …, k N , and calculate the average value L 2 of the new noise floor sequence L 2_ave according to equation (1); continue to calculate the absolute value |X 1 | of the third value X 2 in the signal sequence Data = [X i , X 3 , …, X 3 , …], and repeat the above steps; Repeat the above four steps (1)-(4) until the noise floor sequence L is completely filled to form L N =[|X 1 |,|X 2 |,…,|X t |,…,|X N |], and at the same time, calculate the mean value L of the noise floor sequence through Equation (1); N_ave Through the above four steps (1)(2)(3)(4), the initialization of the noise floor sequence is completed; (5) Set the identifier for whether an impact signal appears as Label. If no impact signal appears, then Label = False. If an impact signal appears, then Label = True. At the same time, define a list of impact signal energies PulseEnergy, and initialize the identifier Label for whether an impact signal appears to false, that is, Label = False, and initialize the list of impact signal energies to empty, that is, PulseEnergy0 = []; When the impact signal identifier Label = False appears, continue to calculate the absolute value of the m-th value X m in the signal sequence Data = [X 1 , X 2 , …, X i , …], and determine whether |X m | exceeds M times the mean value of the background noise sequence, that is, whether it satisfies equation (3): 1 , X 2 , …, X i , …] in the m-th value X m m of the absolute value |X m m |, determine whether |X m m | exceeds M times the mean value of the background noise sequence, that is, whether it satisfies equation (3): |X m |>M*L N_ave (3) (7) If equation (3) holds, it indicates that X m deviates significantly from the mean, and it is determined that there is an impact signal. Take |X m | as the first data point of the impact signal sequence L pulse , and form the impact signal sequence L pulse = [|X m |], and let the maximum value of the impact signal sequence MAX(L pulse ) = |X m |. Calculate the difference between the maximum value MAX(L pulse ) and the mean value of the background noise sequence L N_ave , and take it as the amplitude GAP of the impact signal. The calculation formula for the amplitude GAP of the impact signal is shown in equation (4) as follows: GAP = MAX(L pulse ) - L N_ave (4) At the same time, set the identifier Label for whether an impact signal appears to Label = True; (8) In step (7), if the impact signal identifier Label = True, continue to calculate the signal sequence Data = [X 1 ,X 2 ,…,X i ,…] the m+1th value X m+1 The absolute value of |X m+1 |, first determine |X m+1 |Is it greater than the maximum value MAX(L pulse ); if |X m+1 |Greater than MAX(L pulse ), then update the maximum value MAX(L pulse ) is |X m+1 |, and recalculate the updated impulse signal amplitude GAP according to formula (4); if |X m+1 |Not greater than MAX(L pulse ), then maintain the maximum value MAX(L pulse ) and the amplitude GAP of the impulse signal remain unchanged; then determine |X m+1 | and the mean value of the noise sequence L N_ave Whether the difference exceeds α times the GAP value, that is, it satisfies formula (5): |X m+1 |- L N_ave > α*GAP (5) (9) If the difference between |X m+1 | and the mean value L of the background noise sequence N_ave satisfies equation (5), it indicates that the energy of |X m+1 | is still within the range of impulse energy attenuation. Add |X m+1 | to the impulse signal sequence to form a new impulse signal sequence Lpulse = [|X m |, |X m+1 |], and continue to analyze the (m + 2)-th value X 1 in the signal sequence Data = [X 2 , X i , …, X m+2 according to step (8); (10) If |X m+1 | and the mean value L N_ave of the background noise sequence do not satisfy equation (5), it indicates that the energy of |X m+1 | has basically decayed, indicating that this impact process has ended. Add |X m+1 | to the end of the background noise sequence L N , and at the same time pop the value at the very front of the original background noise sequence L N to form the updated background noise sequence L N+1 = [|X 2 |,..., |X t |,..., |X N |, |X m+1 |] and the updated mean value L N+1_ave of the background noise sequence. At the same time, set the impact signal identifier Label to false, i.e., Label = False, and calculate the energy value Lpulse_Energy of the impact signal sequence according to equation (6): Add L pulse _Energy to the impact signal energy list PulseEnergy to form a new impact signal energy list PulseEnergy1 = [L pulse _Energy]; On this basis, continue to analyze the (m + 2)-th value X 1 , X 2 , …, X i , … in the signal sequence Data = [X m+2 ; (11) If equation (3) does not hold, it indicates that X m is not far from the mean value, and it cannot be determined as the starting point of a shock event. Add |X m | to the end of the background noise sequence L N , and at the same time pop the value at the very front of the original background noise sequence L N to form the updated background noise sequence L N+1 = [|X 2 |,…,|X t |,…,|X N |,|X m |] and the updated mean value L N+1_ave of the background noise sequence. Based on this, continue to analyze the (m + 1)-th value X 1 in the signal sequence Data = [X 2 , X i ,…, X m+1 ; (12) Analyze the remaining data in the signal sequence Data = [X 1 , X 2 , …, X i , …] according to (6)-(11), and finally form the impact signal energy list PulseEnergyM = [L pulse _Energy1, L pulse _Energy2, …, L pulse _EnergyM]. The length of the impact signal energy list PulseEnergyM is M, that is, the analyzed acoustic emission signal has M impacts. By plotting the distribution histogram of the impact signal energy list PulseEnergyM, the distribution of the impact energy can be obtained.

2. The method for analyzing the impact times and impact energy of continuous acoustic emission signals according to claim 1, characterized in that: K is taken as 2.

3. The method for analyzing the impact times and impact energy of continuous acoustic emission signals according to claim 1, characterized in that: M is taken as 5.

4. The method for analyzing the impact times and impact energy of continuous acoustic emission signals according to claim 1, characterized in that: α is taken as 0.2.

Images