Data-driven rotating machinery abnormal vibration alarm threshold setting method and system

Abstract

The invention discloses a data-driven rotating machine abnormal vibration alarm threshold setting method. The method comprises the following steps: S1, calculating mean values and peak values of low-frequency, intermediate-frequency and high-frequency signals in an original acceleration signal of a rotating machine; s2, according to the mean value and the peak value, fitting vibration reference values under different rotating speeds, and connecting the vibration reference values to form a vibration reference value curve; and S3, according to the vibration reference value curve, calculating and generating different rotating speed early warning threshold value curves and different rotating speed alarm threshold value curves. The invention also discloses a data-driven rotating machinery abnormal vibration alarm threshold setting system. By automatically fitting the vibration reference values of the low-frequency, intermediate-frequency and high-frequency signals at different rotating speeds and manually fitting the vibration reference values of the low-frequency, intermediate-frequency and high-frequency signals at different rotating speeds, data-driven rotating machinery abnormal vibration alarm threshold setting is realized, and the problems in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to a rotating mechanical equipment trend warning technology, and more specifically, to a data-driven rotating mechanical abnormal vibration alarm threshold setting method and system. Background Art

[0002] Rotating machinery and equipment plays an extremely important role in the development of modern industry. Ensuring the safe and stable operation of rotating machinery and equipment is the basis for safe production and industrial development.

[0003] Early warning is one of the most basic and effective technical means of equipment status monitoring. At present, a large part of early warning means still remain at fixed threshold alarms. This threshold setting method requires a lot of manual experience and is not universal. For the same type of equipment in different working environments, the threshold setting value will be different, so the alarm threshold needs to be set based on manual experience; and false alarms often occur, which may cause the equipment monitoring parameters to not exceed the alarm threshold when the alarm is triggered, wasting human and material resources; or cause the equipment monitoring parameters to exceed the alarm threshold when the alarm is triggered, and the equipment has already experienced obvious failures, and the effect of early warning is not achieved. Summary of the invention

[0004] In view of the defects existing in the prior art, the purpose of the present invention is to provide a data-driven method and system for setting the abnormal vibration alarm threshold of rotating machinery, which can realize data-driven setting of the abnormal vibration alarm threshold of rotating machinery by automatically fitting the vibration reference values ​​of low-frequency, medium-frequency and high-frequency signals at different speeds, and manually fitting the vibration reference values ​​of low-frequency, medium-frequency and high-frequency signals at different speeds, so as to solve the problems of the prior art.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A first aspect of the present invention provides a data-driven method for setting an abnormal vibration alarm threshold of a rotating machinery, comprising the following steps:

[0007] S1. Calculate the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals in the original acceleration signal of the rotating machinery;

[0008] S2. fitting vibration reference values ​​at different rotation speeds according to the mean value and the peak value, and connecting them into a vibration reference value curve;

[0009] S3. Calculate and generate different speed warning threshold curves and different speed alarm threshold curves according to the vibration reference value curve.

[0010] Preferably, fitting the vibration reference values ​​at different speeds in step S2 specifically includes:

[0011] Automatically fit the vibration reference value at different speeds through the least square method; or

[0012] The vibration reference values ​​at different speeds are manually fitted by manually selecting the corresponding vibration values ​​at three speeds.

[0013] Preferably, automatically fitting the vibration reference values ​​at different speeds by the least squares method comprises the following steps:

[0014] 1) Select a period of historical data at different speeds up to the current collection time from the mean value and the peak value, expressed as (X i ,Y i ),i=1,2,…,n;

[0015] Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed;

[0016] 2) Performing least squares fitting on the data selected in step 1):

[0017] Assume that the curve equation of the least squares fitting method is y=a0+a1x+a2x 2 +…+a k x k ;

[0018] Among them, a0, a1, …, a k is the coefficient of the curve equation; k is the order of the curve equation;

[0019] 3) Substituting the coefficients of the curve equation solved in step 2) into the curve equation to obtain the vibration reference value, and calculating the correlation coefficient between the signal of each frequency band and the vibration reference value.

[0020] Preferably, the step 2) further comprises:

[0021] The curve equation needs to fit the ordinate y i The actual vertical coordinate Y i The sum of squares of the difference is the smallest, that is, the error The value of is the smallest, and the following partial derivatives of the error ε need to be calculated respectively, and each partial derivative is set to 0:

[0022]

[0023]

[0024] …

[0025]

[0026] Then substitute the values ​​of each point into the above formula to get the value.

[0027] Preferably, the correlation coefficient in step 3) is calculated as follows:

[0028]

[0029] Preferably, manually selecting the corresponding vibration values ​​at three rotational speeds to manually fit the vibration reference values ​​at different rotational speeds comprises the following steps:

[0030] A) Select a period of historical data at different speeds up to the current collection time from the mean value and the peak value, expressed as (X i ,Y i ),i=1,2,…,n;

[0031] Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed;

[0032] B) performing three-point fitting on the data selected in step A):

[0033] Assume that the equation of the three-point fitting curve is y=ax 2 +bx+c;

[0034] Among them, a, b, c are the coefficients of the curve equation;

[0035] Manually select the corresponding vibration values ​​(X1, Y1), (X2, Y2), (X3, Y3) at three speeds and substitute them into the three-point fitting curve equation to find the values ​​of a, b, and c;

[0036] C) Substituting the coefficients of the curve equation solved in step B) into the curve equation to obtain the vibration reference value, and calculating the correlation coefficient between the signal of each frequency band and the vibration reference value.

[0037] Preferably, the correlation coefficient in step C) is calculated as follows:

[0038]

[0039] Preferably, the step S3 further comprises:

[0040] According to the vibration reference value curve automatically fitted to obtain the vibration reference value at different speeds, calculate the double value of the vibration reference value curve to be the different speed warning threshold curve; or

[0041] The vibration reference value curve obtained by manually fitting the vibration reference values ​​at different rotation speeds is calculated as the different rotation speed alarm threshold curve by calculating 4 times the value.

[0042] A second aspect of the present invention provides a data-driven rotating machinery abnormal vibration alarm threshold setting system, comprising:

[0043] The signal frequency band separation module is used to calculate the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals respectively;

[0044] The module for automatically fitting vibration reference values ​​at different speeds can automatically fit the vibration reference values ​​at different speeds through the least square method according to the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals, connect them into a smooth vibration reference value curve, and calculate and give the correlation coefficient between the signal of each frequency band and the vibration reference value;

[0045] The manual fitting module for vibration reference values ​​at different speeds manually fits the vibration reference values ​​at different speeds by manually selecting the corresponding vibration values ​​at three speeds based on the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals, connects them into a smooth vibration reference value curve, and calculates and gives the correlation coefficient between the signal in each frequency band and the vibration reference value;

[0046] The alarm threshold setting module calculates and generates different speed warning threshold curves and different speed alarm threshold curves based on the vibration reference value curve obtained by the vibration reference value automatic or manual fitting module, and is used to determine whether the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals at the current speed exceed the different speed warning threshold curves and different speed alarm threshold curves obtained at the speed.

[0047] Preferably, the signal frequency band separation module filters out low-frequency, medium-frequency and high-frequency signals in the original acceleration signal of the rotating machinery through a bandpass filter, and then calculates the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals respectively.

[0048] The present invention provides a data-driven abnormal vibration alarm threshold setting method and system for rotating machinery. The method selects a period of historical data at different speeds up to the current acquisition time, and filters out the low-frequency, medium-frequency and high-frequency signals in the original signal through the signal frequency band separation module; then the least square method and the three-point fitting method are used to fit the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals, respectively, to obtain the vibration reference value curve, and the correlation coefficient between the signals of each frequency band and the vibration reference value is calculated to evaluate the fitting effect; finally, the warning threshold curve is set to 2 times the value of the vibration reference value curve, and the alarm threshold curve is set to 4 times the value of the vibration reference value curve, and it is judged whether the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals at the current speed exceed the warning threshold curve and the alarm threshold curve obtained by fitting at the speed, so as to realize the early warning of early failure of mechanical equipment. Compared with the traditional simple threshold alarm, the present invention establishes a data-driven alarm threshold curve based on big data, which has the advantages of high warning accuracy and intelligent warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of a method for setting an abnormal vibration alarm threshold value of a rotating machinery according to the present invention;

[0050] Figure 2 It is a flow chart of signal frequency band separation in the method for setting the alarm threshold value of abnormal vibration of rotating machinery of the present invention;

[0051] Figure 3 It is a schematic flow chart of automatic fitting of vibrations at different rotation speeds in the method for setting abnormal vibration alarm threshold of rotating machinery of the present invention;

[0052] Figure 4 It is a flow chart of manual fitting of vibrations at different rotation speeds in the method for setting abnormal vibration alarm threshold of rotating machinery of the present invention. DETAILED DESCRIPTION

[0053] In order to better understand the above technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0054] The present invention provides a data-driven abnormal vibration alarm threshold setting system for rotating machinery, comprising a signal frequency band separation module, an automatic fitting module for vibration reference values ​​at different speeds, a manual fitting module for vibration reference values ​​at different speeds and an alarm threshold setting module.

[0055] Combination Figure 1 As shown, the present invention also provides a data-driven method for setting an abnormal vibration alarm threshold of a rotating machinery, which specifically includes:

[0056] The signal frequency band separation module firstly filters out the low-frequency, medium-frequency and high-frequency signals in the original acceleration signal of the rotating machinery through a bandpass filter, and then calculates the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals respectively;

[0057] The automatic fitting module for vibration reference values ​​at different speeds automatically fits the optimal vibration reference values ​​at different speeds through the least square method according to the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals, connects them into a smooth optimal vibration reference value curve, and calculates and gives the correlation coefficient between the signal of each frequency band and the optimal vibration reference value;

[0058] The manual fitting module for vibration reference values ​​at different speeds manually fits the vibration reference values ​​at different speeds by manually selecting the corresponding vibration values ​​at three speeds according to the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals, connects them into a smooth vibration reference value curve, and calculates and gives the correlation coefficient between the signal of each frequency band and the vibration reference value;

[0059] The alarm threshold setting module calculates and generates different speed warning threshold curves and different speed alarm threshold curves based on the vibration reference value curve automatically or manually fitted by the vibration reference value fitting module at different speeds, and is used to determine whether the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals at the current speed exceed the different speed warning threshold curves and different speed alarm threshold curves obtained at the speed.

[0060] Combination Figure 2 As shown, the low-frequency, medium-frequency and high-frequency signals in the original acceleration signal are filtered out respectively by band-pass filters, wherein the low-frequency signal refers to the mean and peak values ​​of the velocity signal obtained by converting the original acceleration signal into a velocity signal by the frequency domain integration method and then filtering it through a band-pass filter with a filtering range of 100 to 1000 Hz, and the unit is mm / s; the medium-frequency signal refers to the mean and peak values ​​of the acceleration signal obtained by filtering the original acceleration signal through a band-pass filter with a filtering range of 1000 to 10000 Hz, and the unit is gravity acceleration G; the high-frequency signal refers to the mean and peak values ​​of the envelope signal obtained by filtering out the highest frequency band information of the original acceleration signal through a band-pass filter with a filtering range of 10000 Hz to half the sampling frequency and then Hilbert transforming it, and the unit is gravity acceleration G.

[0061] Combination Figure 3 As shown, the following steps are included to automatically fit the vibration reference values ​​at different speeds using the least squares method:

[0062] 1) Select a period of historical data at different speeds from the mean and peak values ​​up to the current acquisition time, expressed as (X i ,Y i ),i=1,2,…,n;

[0063] Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed;

[0064] 2) Perform least squares fitting on the data selected in step 1):

[0065] Assume that the curve equation of the least squares fitting method is y=a0+a1x+a2x 2 +…+a k x k ;

[0066] Among them, a0, a1, …, a k is the coefficient of the curve equation; k is the order of the curve equation;

[0067] In order to achieve the best fitting effect, the curve equation needs to fit the ordinate y i The actual vertical coordinate Y iThe sum of squares of the difference is the smallest, that is, the error The value of is the smallest, and the following partial derivatives of the error ε need to be calculated respectively, and each partial derivative is set to 0:

[0068]

[0069]

[0070] …

[0071]

[0072] Then substitute the values ​​of each point into the above formula to get the value;

[0073] 3) Substitute the coefficients of the curve equation solved in step 2) into the curve equation to obtain the vibration reference value, and calculate the correlation coefficient between the signal of each frequency band and the vibration reference value. The correlation coefficient is calculated as follows:

[0074]

[0075] Combination Figure 4 As shown, manually selecting the corresponding vibration values ​​at three speeds to manually fit the vibration reference values ​​at different speeds includes the following steps:

[0076] A) Select a period of historical data at different speeds from the mean and peak values ​​up to the current acquisition time, expressed as (X i ,Y i ),i=1,2,…,n;

[0077] Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed;

[0078] B) Perform three-point fitting on the data selected in step A):

[0079] Assume that the equation of the three-point fitting curve is y=ax 2 +bx+c;

[0080] Among them, a, b, c are the coefficients of the curve equation;

[0081] Manually select the corresponding vibration values ​​(X1, Y1), (X2, Y2), (X3, Y3) at three speeds and substitute them into the three-point fitting curve equation to find the values ​​of a, b, and c;

[0082] C) Substitute the coefficients of the curve equation solved in step B) into the curve equation to obtain the vibration reference value, and calculate the correlation coefficient between the signal of each frequency band and the vibration reference value. The correlation coefficient is calculated as follows:

[0083]

[0084] Example

[0085] This embodiment selects the historical acceleration data of the output side of the gearbox output shaft at a speed of 500-655r / min from the data collected within 1 hour. Through the frequency domain integration method, the above historical acceleration signal is input into a bandpass filter with a filtering range of 100-1000Hz to obtain a speed signal, and the mean range and peak range at 577r / min are calculated to be 0.2523-1.1124m / s and 0.4326-1.3523m / s respectively. The above historical acceleration signal is filtered through a bandpass filter with a filtering range of 1000-10000Hz to obtain an acceleration signal, and the mean and peak values ​​at 577r / min are calculated to be 0.0016-0.0815G and 0.0811-0.1679G respectively.

[0086] The above historical acceleration signal is passed through a bandpass filter with a filtering range of 10000 Hz to half the sampling frequency, and then the envelope signal of the aluminum foil signal is calculated by Hilbert transform. Finally, the mean and peak values ​​at 577 r / min are calculated to be 0.0149~0.0721 G and 0.2983~0.6958 G respectively.

[0087] The fitting results of the mean and peak values ​​of the output side of the gearbox output shaft at 577 r / min were calculated using the least squares method. Among them, the mean and peak values ​​at 100-1000 Hz are 0.6128 m / s and 0.7436 m / s, respectively, the mean and peak values ​​at 1000-10000 Hz are 0.0421 G and 0.1237 G, respectively, and the mean and peak values ​​at 10000 Hz to half the sampling frequency are 0.0327 G and 0.4817 G, respectively.

[0088] The correlation coefficients between the fitting value and the actual value in the three frequency bands of low, medium and high frequencies calculated by the correlation coefficient calculation formula are 91.96%, 66.68% and 22.74% respectively. The fitting results of the mean and peak values ​​of the output side of the output shaft of the gearbox at 577r / min are calculated by the three-point fitting method. Among them, the mean and peak values ​​at 100-1000Hz are 0.5963m / s and 0.7124m / s respectively, the mean and peak values ​​at 1000-10000Hz are 0.0364G and 0.1197G respectively, and the mean and peak values ​​at 10000Hz to half the sampling frequency are 0.0298G and 0.5027G respectively.

[0089] The correlation coefficients between the fitted values ​​and the actual values ​​in the low, medium and high frequency bands calculated using the correlation coefficient calculation formula are 59.31%, 62.61% and 32.33% respectively. Therefore, the results of fitting the low-frequency data using the least squares method are better. We use the fitting results of the least squares method to set the warning threshold curve, which is equal to 2 times the fitting value: mean 1.2256m / s and peak 1.4872m / s; the alarm threshold curve is equal to 4 times the fitting value: mean 2.4512m / s and peak 2.9744m / s.

[0090] The results of fitting the intermediate frequency data using the least squares method are better. The warning threshold curve is set using the fitting results of the least squares method, which is equal to twice the fitting value: mean 0.0842G and peak 0.2474G; the alarm threshold curve is equal to four times the fitting value: mean 0.1684G and peak 0.4948G.

[0091] The results of fitting high-frequency data using the three-point fitting method are better. We use the fitting results of the three-point fitting method to set the warning threshold curve, which is equal to 2 times the fitting value: mean 0.0596G and peak 1.0054G; the alarm threshold curve is equal to 4 times the fitting value: mean 0.1192G and 2.0108G peak. So far, the subsequent acceleration signal of the output shaft of the gearbox at a speed of 577r / min, after filtering in the low, medium and high frequency bands, the mean and peak values ​​are used for warning and alarm according to the above results.

[0092] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention, and are not intended to limit the present invention. As long as they are within the spirit of the present invention, any changes or modifications to the above embodiments will fall within the scope of the claims of the present invention.

Claims

1. A data-driven method for setting abnormal vibration alarm threshold of rotating machinery, characterized in that: The following steps are involved: S1. Calculate the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals in the original acceleration signal of the rotating machinery; S2. fitting vibration reference values ​​at different rotation speeds according to the mean value and the peak value, and connecting them into a vibration reference value curve; S3. Calculate and generate different speed warning threshold curves and different speed alarm threshold curves according to the vibration reference value curve.

2. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 1, characterized in that: The step S2 of fitting the vibration reference values ​​at different speeds specifically includes: Automatically fit the vibration reference value at different speeds through the least square method; or The vibration reference values ​​at different speeds are manually fitted by manually selecting the corresponding vibration values ​​at three speeds.

3. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 2, characterized in that: Automatically fitting the vibration reference values ​​at different speeds by the least squares method includes the following steps: 1) Select a period of historical data at different speeds up to the current collection time from the mean value and the peak value, expressed as (X i ,Y i ),i=1,2,…,n; Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed; 2) Performing least squares fitting on the data selected in step 1): Assume that the curve equation of the least squares fitting method is y=a0+a1x+a2x 2 +…+a k x k ; Among them, a0, a1, …, a k is the coefficient of the curve equation; k is the order of the curve equation; 3) Substituting the coefficients of the curve equation solved in step 2) into the curve equation to obtain the vibration reference value, and calculating the correlation coefficient between the signal of each frequency band and the vibration reference value.

4. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 3 is characterized in that: The step 2) further comprises: The curve equation needs to fit the ordinate y i The actual vertical coordinate Y i The sum of squares of the difference is the smallest, that is, the error The value of is the smallest, and the following partial derivatives of the error ε need to be calculated respectively, and each partial derivative is set to 0: … Then substitute the values ​​of each point into the above formula to get the value.

5. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 4, characterized in that: In step 3), the correlation coefficient is calculated as follows:

6. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 2, characterized in that: Manually selecting the corresponding vibration values ​​at three rotational speeds to manually fit the vibration reference values ​​at different rotational speeds comprises the following steps: A) Select a period of historical data at different speeds up to the current collection time from the mean value and the peak value, expressed as (X i ,Y i ),i=1,2,…,n; Among them, X i is the specific speed; Y i is the vibration value at the corresponding speed; B) performing three-point fitting on the data selected in step A): Assume that the equation of the three-point fitting curve is y=ax 2 +bx+c; Among them, a, b, c are the coefficients of the curve equation; Manually select the corresponding vibration values ​​(X1, Y1), (X2, Y2), (X3, Y3) at three speeds and substitute them into the three-point fitting curve equation to find the values ​​of a, b, and c; C) Substituting the coefficients of the curve equation solved in step B) into the curve equation to obtain the vibration reference value, and calculating the correlation coefficient between the signal of each frequency band and the vibration reference value.

7. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 6, characterized in that: In step C), the correlation coefficient is calculated as follows:

8. The data-driven rotating machinery abnormal vibration alarm threshold setting method according to claim 5 or 7, characterized in that: The step S3 further comprises: According to the vibration reference value curve automatically fitted to obtain the vibration reference value at different speeds, calculate the double value of the vibration reference value curve to be the different speed warning threshold curve; or The vibration reference value curve obtained by manually fitting the vibration reference values ​​at different rotation speeds is calculated as the different rotation speed alarm threshold curve by calculating 4 times the value.

9. A data-driven rotating machinery abnormal vibration alarm threshold setting system, characterized in that: include: The signal frequency band separation module is used to calculate the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals respectively; The module for automatically fitting vibration reference values ​​at different speeds can automatically fit the vibration reference values ​​at different speeds through the least square method according to the mean and peak values ​​of low-frequency, medium-frequency and high-frequency signals, connect them into a smooth vibration reference value curve, and calculate and give the correlation coefficient between the signal of each frequency band and the vibration reference value; The manual fitting module for vibration reference values ​​at different speeds manually fits the vibration reference values ​​at different speeds by manually selecting the corresponding vibration values ​​at three speeds based on the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals, connects them into a smooth vibration reference value curve, and calculates and gives the correlation coefficient between the signal in each frequency band and the vibration reference value; The alarm threshold setting module calculates and generates different speed warning threshold curves and different speed alarm threshold curves based on the vibration reference value curve obtained by the vibration reference value automatic or manual fitting module, and is used to determine whether the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals at the current speed exceed the different speed warning threshold curves and different speed alarm threshold curves obtained at the speed.

10. The data-driven rotating machinery abnormal vibration alarm threshold setting system according to claim 9, characterized in that: The signal frequency band separation module filters out the low-frequency, medium-frequency and high-frequency signals in the original acceleration signal of the rotating machinery through a bandpass filter, and then calculates the mean and peak values ​​of the low-frequency, medium-frequency and high-frequency signals respectively.

Images