Data cleaning and fault diagnosis method for quay crane trolley traction mechanism

Abstract

The invention belongs to the technical field of fault diagnosis, and particularly relates to a data cleaning and fault diagnosis method for a quay crane trolley traction mechanism. The data cleaning and fault diagnosis method for the quay crane trolley traction mechanism comprises the following steps that S1, a vibration acceleration sensor is arranged at a trolley motor driving end measuring point or a trolley gearbox upper measuring point for signal collection and preprocessing; s2, the collected signals are screened; s3, signal processing; and S4, identifying fault features. The data cleaning and fault diagnosis method for the quay crane trolley traction mechanism has the effect of realizing quantitative analysis and automatic diagnosis of faults of the trolley traction mechanism.

Description

Technical Field

[0001] The invention belongs to the technical field of fault diagnosis, and in particular relates to a data cleaning and fault diagnosis method for a trolley traction mechanism of a quay crane. Background Art

[0002] Shore container crane (abbreviated as quay crane) is an important process equipment in the port. With the changes in relevant parameters such as the large-scale size of port container machinery and the heavy rated lifting weight under the spreader, the increase in the lifting weight, lifting speed, trolley speed, etc. of the quay crane makes the motor, reduction box, drum and other components of the trolley traction mechanism of the quay crane heavier, and there is a trend of frequent failures. The diagnosis of the trolley migration mechanism of the existing quay crane mainly relies on manual regular inspection to listen for abnormal noise, or install vibration sensors, and evaluate the equipment status through vibration indicators or machine learning. The above two existing technology solutions require high manpower costs for the former, and require personnel with rich experience to give a relatively accurate conclusion. The latter has the speed and load of the trolley traction mechanism of the quay crane directly controlled by humans, resulting in unstable and frequent speed changes. The collected vibration signal will contain shutdown data and abnormal impact caused by the brake device, which makes the latter monitoring accuracy low and the implementation cycle very long.

[0003] The above-mentioned existing technical solutions have the following defects: they require a large manpower cost, and due to the influence of abnormal impact caused by shutdown data and braking devices, the accuracy of the monitoring solution of the existing technology is low and the implementation cycle is very long. Summary of the invention

[0004] The purpose of the present invention is to provide a data cleaning and fault diagnosis method for the trolley traction mechanism of the quay crane, so as to solve the technical problems of high labor cost demand, low accuracy and long implementation cycle, so as to achieve the purpose of quantitative analysis and automatic diagnosis of the trolley traction mechanism fault.

[0005] In order to solve the above technical problems, the present invention provides a data cleaning and fault diagnosis method for a trolley traction mechanism of a quay crane, the method comprising the following steps:

[0006] S1, a vibration acceleration sensor is provided at the measuring point of the motor drive end of the trolley or the measuring point on the gear box of the trolley for signal collection and preprocessing;

[0007] S2, screening the collected signals;

[0008] Screening the collected signal includes the following steps:

[0009] S2-1, shutdown and abnormal impact data elimination;

[0010] Divide the motor side point signal or gearbox measurement point signal into I groups of signals y with a length of N i(n), i = 1, 2 ... I; n = 1, 2 ... N, and perform the first data screening, and record the data after the first data screening as the slice index B;

[0011] S2-2, stable phase identification;

[0012] According to the slice index B from the signal y i (n) Filter out several groups of signals yc with length N j (n), and obtain the segmented RMS, centroid frequency and envelope spectrum kurtosis of each group of signals to form a feature vector and calculate the correlation coefficient of the feature vectors of two adjacent groups of signals, and then perform a second data screening to form a slice index E, and then perform a third data screening on the slice index E and enter the next stage;

[0013] S2-3, data reconstruction;

[0014] Use the slice index E to slice the signal from the motor side point or the gearbox measurement point yc j (n) and complete the signal reconstruction in index order, and finally obtain the signal xc 0 or signal xc 1 ;

[0015] S3, signal processing;

[0016] Signal processing includes the following steps:

[0017] S3-1, signal xc 0 or signal xc 1 Perform adaptive tensor singular spectrum decomposition to obtain l IMF components;

[0018] S3-2, calculate the linear kurtosis of each IMF component;

[0019] S3-3, select the IMF components whose linear kurtosis is greater than the linear kurtosis threshold for reconstruction, and obtain the denoised signal xci 0 and xci 1 .

[0020] S4. Identify fault characteristics and conduct quantitative analysis of gearbox faults.

[0021] Furthermore, in step S1, signal acquisition and preprocessing includes the following steps:

[0022] S1-1, trigger collection and equipped with sampling device;

[0023] S1-2, the measuring point at the driving end of the trolley motor or the measuring point on the gearbox of the trolley continuously collects vibration acceleration and calculates the root mean square value. When the root mean square value exceeds the set trigger threshold, the sampling device immediately collects the vibration acceleration signal with a sampling frequency of fs and a duration of T;

[0024] S1-3. Remove DC bias components and low-frequency disturbances from the signal.

[0025] Furthermore, in step S2-1, the first data screening includes the following steps:

[0026] S2-1-1. Calculate the root mean square (RMS) and kurtosis factor (KU) value of each group of I signals yi(n) using the following formula:

[0027]

[0028]

[0029] Where σ is the standard deviation of yi(n);

[0030] S2-1-2. Compare RMS i , KU i With the power-on threshold T RMS and abnormal shock threshold T ku The size of the slice index is retained to form the array A in accordance with the following relationship:

[0031] Rm i >T RMS AndKU i <T ku

[0032] S2-1-3. Sort array A by index from small to large, and divide A into g one-dimensional arrays based on continuity, retaining the array with the longest length, thus completing the first data screening.

[0033] Furthermore, in step S2-2, the stable phase identification includes the following steps:

[0034] S2-2-1. From signal y according to slice index B i (n) and obtain J groups of signals yc with a length of N j (n), j = 1, 2…J;

[0035] S2-2-2. Calculate signal yc j (n) The segmented RMS, centroid frequency and envelope spectrum kurtosis of each group of signals;

[0036] S2-2-3, the signal yc j (n) The calculated segment RMS, center of gravity frequency and envelope spectrum kurtosis value constitute the signal feature vector C j , and calculate the correlation coefficient of two adjacent groups of signal feature vectors;

[0037] S2-2-4, determine the relationship between the correlation coefficient and the correlation coefficient deviation threshold, perform a second data screening, and record the data after the second data screening as the slice index E;

[0038] S2-2-5, determine the relationship between the slice number of slice index E and the slice number threshold, perform a third data screening, and the slice index E that meets the third data screening proceeds to the next step;

[0039] Furthermore, the segmented root mean square (RMS jm ) is calculated as follows:

[0040] Select M frequency ranges [f m ,f m+1 ], m=1,2…M, and yc is calculated by the following formula j (n) Perform frequency domain bandpass filtering to obtain the filtered signal ycl jm (m),

[0041]

[0042] Where e is a natural constant, j 1 is the imaginary unit, H(k) is the frequency response function,

[0043]

[0044] Where t is the duration of each signal segment;

[0045] For each segmented signal ycl after filtering jm (m) Calculate the root mean square and obtain the segmented root mean square (RMS jm ).

[0046] Furthermore, the center of gravity frequency (FC j ) is calculated as follows:

[0047]

[0048] Among them, F j is the signal yc j (n) The frequency domain signal is obtained by fast Fourier transform, where f is the frequency.

[0049] Furthermore, the envelope spectrum kurtosis (SK j ) is calculated as follows:

[0050]

[0051] Among them, e j is the signal yc j (n) Envelope signal obtained after Hilbert demodulation, σ e for e jStandard deviation.

[0052] Furthermore, the calculation method of the correlation coefficient of two adjacent groups of signal feature vectors is as follows:

[0053]

[0054] Among them, r is the correlation coefficient, Cov(C j ,C j+1 ) is the covariance of two adjacent signal eigenvectors, is the variance of the eigenvector.

[0055] Furthermore, in step S2-2-4, the second data screening includes the following steps:

[0056] S2-2-4-1. Retain the slice index of the correlation coefficients of two adjacent groups of signals that satisfy the following relationship to form a one-dimensional array D,

[0057] r j-1 >ε r

[0058] Among them, ε r is the correlation coefficient deviation threshold;

[0059] S2-2-4-2. Sort the one-dimensional array D by index from small to large, and divide D into h one-dimensional arrays based on continuity. Keep the array with the longest length, which is the slice index E finally retained after the second data screening.

[0060] Furthermore, in step S2-2-5, the specific location of the third data screening is as follows:

[0061] Determine whether the number of slices of slice index E satisfies the following relationship, if yes, proceed to the next step:

[0062] length(E)>ε l

[0063] Where, length(E) is the number of slices of slice index E, ε k is the slice number threshold.

[0064] Furthermore, in step S3-1, the specific steps of adaptive tensor singular spectrum decomposition are as follows:

[0065] S3-1-1, xc 0 or xc 1 Constructed into a third-order tensor Z, the tensor Z is expressed as follows,

[0066]

[0067] S3-1-2. Perform singular value decomposition on the tensor Z and obtain the following:

[0068]

[0069] Where L is the first-order dimension of the tensor S;

[0070] The tensor multiplication operation rules mentioned above are as follows:

[0071]

[0072] Among them, variable b 1 ,b 2 ,…b c+d-2o is a positive integer, a represents the order of the tensors involved in the tensor multiplication operation, <1,1> means that the order of operations of tensor X and tensor Y during tensor multiplication is positive; c is the order of tensor X, d is the order of tensor Y; S 1 is the second-order dimension of the tensor X, S o is the dimension of the tensor X at the o+1th order;

[0073] S3-1-3, let A l =U× 1,<1,1> S l,: × 2,<1,1> V T And the tensor A l The first column is imf l , then we get the following:

[0074]

[0075] Among them, IMF l Signal xc 0 The lth IMF component of .

[0076] Furthermore, in step S3-2, the specific calculation formula of linear kurtosis is as follows:

[0077]

[0078] Suppose the sample sequence of the signal is {X}, and the sample values ​​are arranged in order of size as x (1) ≤x (2) ≤x (3) ≤…≤x (n) , p is the total number of samples in the sequence, and the unbiased estimate of the signal sample weight probability moment is:

[0079]

[0080] Where r is the order of the random variable sequence.

[0081] Furthermore, in step S3-3, the specific formula for screening the IMF component is as follows:

[0082] L_kurtosis>ε lk

[0083] Among them, ε lk is the linear kurtosis threshold.

[0084] Furthermore, in step S4, fault feature identification includes the following steps:

[0085] S4-1, signal xci 0 or signal xci 1 Perform fast Fourier transform to obtain the frequency domain signal F(x), and extract the frequency domain signal F(x) in [f c0 -ε fc ,f c0 +ε fc The frequency coordinate corresponding to the maximum amplitude value in the interval is denoted as f c , where f c0 is the preset ideal frequency, ε fc is the frequency error threshold;

[0086] S4-2, extract the frequency domain signal F(x) in [q*(f c -ε fc ),q*(f c +ε fc )] or in [w*(f mesh -ε gear ),w*(f mesh +ε gear The frequency coordinate corresponding to the maximum amplitude value in the interval is recorded as the frequency multiple fc q ,q=1,2…Q, or gear meshing frequency fg w ,w=1,2…W, Q, W are the order; multiply the frequency fc of each order of the frequency conversion q Combined into vector f cx , or the meshing frequencies of the gears at each order fg w Combined into vector f gear , where ε gear is the error of gear meshing frequency, f mesh is the gear meshing frequency;

[0087] S4-3, signal xci 0 or signal xci 1 Perform discrete Hilbert transform and fast Fourier transform to obtain the envelope spectrum of the component and take the modulus to obtain the frequency domain signal H(x). Extract H(x) in the following four intervals [fif*f c -ε bf ,fif*f c +ε bf ], [bsf*f c -ε bf ,bsf*f c +ε bf ]、[bpfi*f c -ε bf ,bpfi*f c +ε bf ], [bpfo*f c -ε bf ,bpfo*f c +ε bf The frequency coordinates corresponding to the maximum amplitude values ​​within ] are recorded as bearing characteristic frequencies f fif 、f bsf 、f bpfi 、f bpfo , forming the vector f bf , where ε bf is the bearing characteristic error threshold, fif, bsf, bpfi, bpfo are rolling bearing characteristic coefficients;

[0088] S4-4, xci 0 The signals are extracted separately cx The amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f bf The frequency and the corresponding amplitude form a two-dimensional array E, which can be used to quantitatively analyze the motor fault;

[0089] or xci 1 The signal is extracted F(x) at f cx 、f gear The amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f gear 、f bf The frequency and the corresponding amplitude form a two-dimensional array F, which can be used to quantitatively analyze the gearbox fault.

[0090] The beneficial effect of the present invention is to realize quantitative analysis and automatic diagnosis of faults of the trolley traction mechanism.

[0091] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0093] Figure 1 It is a flow chart of a data cleaning and fault diagnosis method for a quay crane trolley traction mechanism of the present invention. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0095] Example:

[0096] like Figure 1 As shown, a data cleaning and fault diagnosis method for the trolley traction mechanism of the quay crane is provided. A vibration acceleration sensor is installed at the motor drive end of the trolley traction mechanism, and three vibration acceleration sensors are installed on the gear box. The specific positions can be adjusted according to the actual structure and requirements. The data cleaning and fault diagnosis method for the trolley traction mechanism of the quay crane of the present invention includes the following steps:

[0097] S1. A vibration acceleration sensor is provided at a measuring point on the motor drive end of the trolley or a measuring point on the gear box of the trolley for signal collection and preprocessing. The signal collection and preprocessing includes the following steps:

[0098] S1-1, trigger collection and equipped with sampling device;

[0099] S1-2, the measuring point at the driving end of the trolley motor or the measuring point on the gearbox of the trolley continuously collects vibration acceleration and calculates the root mean square value. When the root mean square value exceeds the set trigger threshold, the sampling device immediately collects the vibration acceleration signal with a sampling frequency of fs and a duration of T;

[0100] S1-3. Use the DETREND function in MATLAB software to remove the DC bias component and low-frequency disturbance in the signal.

[0101] S2. Filter the collected signals. The trolley traction mechanism starts and stops frequently and has a short stabilization time. There is a high possibility of shutdown data in the method of triggering the collection of fixed-duration signals. In addition, the on-site working conditions of the quay crane are complex. Equipment braking, ship docking, and the start and stop of other mechanisms may cause impacts and affect the quality of the signal. Therefore, the data containing these two situations needs to be eliminated. The equipment working conditions are not stable during the signal collection process, and the vibration signals corresponding to the stable stage need to be filtered out;

[0102] Screening the collected signal includes the following steps:

[0103] S2-1, shutdown and abnormal impact data are eliminated, and the motor side point signal or gearbox measurement point signal is divided into I groups of signals y with a length of N. i (n), i = 1, 2 ... I; n = 1, 2 ... N, and perform the first data screening, and record the data after the first data screening as the slice index B; wherein the first data screening includes the following steps:

[0104] S2-1-1. Calculate the root mean square (RMS) and kurtosis factor (KU) value of each group of I signals yi(n) using the following formula:

[0105]

[0106]

[0107] Where σ is the standard deviation of yi(n);

[0108] S2-1-2. Compare RMS i , KU i With the power-on threshold T RMS and abnormal shock threshold T ku The size of the slice index is retained to form the array A in accordance with the following relationship:

[0109] Rm i >T RMS AndKU i <T ku

[0110] S2-1-3. Sort array A by index from small to large, and divide A into g one-dimensional arrays based on continuity, retaining the array with the longest length, thus completing the first data screening.

[0111] S2-2, stable phase identification;

[0112] Stable phase identification includes the following steps:

[0113] S2-2-1. From signal y according to slice index B i(n) and obtain J groups of signals yc with a length of N j (n), j = 1, 2…J;

[0114] S2-2-2. Calculate signal yc j (n) The segmented RMS, centroid frequency and envelope spectrum kurtosis of each group of signals;

[0115] S2-2-3, the signal yc j (n) The calculated segment RMS, center of gravity frequency and envelope spectrum kurtosis value constitute the signal feature vector C j , and calculate the correlation coefficient of two adjacent groups of signal feature vectors;

[0116] Segmental root mean square (RMS jm ) is calculated as follows:

[0117] Select M frequency ranges [f m ,f m+1 ], m=1,2…M, and yc is calculated by the following formula j (n) Perform frequency domain bandpass filtering to obtain the filtered signal ycl jm (m),

[0118]

[0119] Where e is a natural constant, j 1 is the imaginary unit, H(k) is the frequency response function,

[0120]

[0121] Where t is the duration of each signal segment;

[0122] For each segmented signal ycl after filtering jm (m) Calculate the root mean square and obtain the segmented root mean square (RMS jm ).

[0123] Center of gravity frequency (FC j ) is calculated as follows:

[0124]

[0125] Among them, F j is the signal yc j (n) The frequency domain signal is obtained by fast Fourier transform, where f is the frequency.

[0126] , envelope spectrum kurtosis (SK j ) is calculated as follows:

[0127]

[0128] Among them, e j is the signal yc j (n) Envelope signal obtained after Hilbert demodulation, σ e for e j Standard deviation.

[0129] The calculation method of the correlation coefficient feature vector of two adjacent groups of signals is as follows:

[0130]

[0131] Among them, r is the correlation coefficient, Cov(C j ,C j+1 ) is the covariance of two adjacent signal eigenvectors, is the variance of the eigenvector.

[0132] S2-2-4. Determine the relationship between the correlation coefficient and the correlation coefficient deviation threshold, perform a second data screening, and record the data after the second data screening as the slice index E. The second data screening includes the following steps:

[0133] S2-2-4-1. Retain the slice index of the correlation coefficients of two adjacent groups of signals that satisfy the following relationship to form a one-dimensional array D,

[0134] r j-1 >ε r

[0135] Among them, ε r is the correlation coefficient deviation threshold;

[0136] S2-2-4-2. Sort the one-dimensional array D by index from small to large, and divide D into h one-dimensional arrays based on continuity. Keep the array with the longest length, which is the slice index E finally retained after the second data screening.

[0137] S2-2-5. Determine the relationship between the slice number of slice index E and the slice number threshold, perform the third data screening, and the slice index E that meets the third data screening proceeds to the next step; the specific position of the third data screening is as follows:

[0138] Determine whether the number of slices of slice index E satisfies the following relationship, if yes, proceed to the next step:

[0139] length(E)>ε l

[0140] Where, length(E) is the number of slices of slice index E, ε l is the slice number threshold.

[0141] S2-3, data reconstruction, using the slice index E to slice the signal from the motor side point or the gearbox measurement point slice signal yc j (n) and complete the signal reconstruction in index order, and finally obtain the signal xc 0 or signal xc 1 .

[0142] S3, signal processing;

[0143] Signal processing includes the following steps:

[0144] S3-1, signal xc 0 or signal xc 1 Perform adaptive tensor singular spectrum decomposition to obtain l IMF components; the specific steps of adaptive tensor singular spectrum decomposition are as follows:

[0145] S3-1-1, xc 0 or xc 1 Constructed into a third-order tensor Z, the tensor Z is expressed as follows,

[0146]

[0147] S3-1-2. Perform singular value decomposition on the tensor Z and obtain the following:

[0148]

[0149] Where L is the first-order dimension of the tensor S;

[0150] The tensor multiplication operation rules mentioned above are as follows:

[0151]

[0152] Among them, variable b 1 ,b 2 ,…b c+d-2o is a positive integer, a represents the order of the tensors involved in the tensor multiplication operation, <1,1> means that the order of operations of tensor X and tensor Y during tensor multiplication is positive; c is the order of tensor X, d is the order of tensor Y; S 1 is the second-order dimension of the tensor X, S o is the dimension of the tensor X at the o+1th order;

[0153] S3-1-3, let A l =U× 1,<1,1> S l,: × 2,<1,1> V T And the tensor A l The first column is imf l , then we get the following:

[0154]

[0155] Among them, IMF l Signal xc 0 The lth IMF component of .

[0156] S3-2. Calculate the linear kurtosis of each IMF component. The specific calculation formula of linear kurtosis is as follows:

[0157]

[0158] Suppose the sample sequence of the signal is {X}, and the sample values ​​are arranged in order of size as x (1) ≤x (2) ≤x (3) ≤…≤x (n) , p is the total number of samples in the sequence, and the unbiased estimate of the signal sample weight probability moment is:

[0159]

[0160] Where r is the order of the random variable sequence.

[0161] S3-3, select the IMF components whose linear kurtosis is greater than the linear kurtosis threshold for reconstruction, and obtain the denoised signal xci 0 and xci 1 ; The specific formula for screening IMF components is as follows:

[0162] L_kurtosis>ε lk

[0163] Among them, ε lk is the linear kurtosis threshold.

[0164] S4, fault feature identification;

[0165] Fault feature identification includes the following steps:

[0166] S4-1, signal xci 0 or signal xci 1 Perform fast Fourier transform to obtain the frequency domain signal F(x), and extract the frequency domain signal F(x) in [f c0 -ε fc ,f c0 +ε fc The frequency coordinate corresponding to the maximum amplitude value in the interval is denoted as f c , where f c0 is the preset ideal transfer frequency, ε fc is the frequency error threshold;

[0167] S4-2, extract the frequency domain signal F(x) in [q*(fc -ε fc ),q*(f c +ε fc )] or in [w*(f mesh -ε gear ),w*(f mesh +ε gear The frequency coordinate corresponding to the maximum amplitude value in the interval is recorded as the frequency multiple fc q ,q=1,2…Q, or gear meshing frequency fg w ,w=1,2…W, Q, W are the order; multiply the frequency fc of each order of the frequency conversion q Combined into vector f cx , or the meshing frequencies of the gears at each order fg w Combined into vector f gear , where ε gear is the error of gear meshing frequency, f mesh is the gear meshing frequency;

[0168] S4-3, signal xci 0 or signal xci 1 Perform discrete Hilbert transform and fast Fourier transform to obtain the envelope spectrum of the component and take the modulus to obtain the frequency domain signal H(x). Extract H(x) in the following four intervals [fif*f c -ε bf ,fif*f c +ε bf ], [bsf*f c -ε bf ,bsf*f c +ε bf ]、[bpfi*f c -ε bf ,bpfi*f c +ε bf ], [bpfo*f c -ε bf ,bpfo*f c +ε bf The frequency coordinates corresponding to the maximum amplitude values ​​within ] are recorded as bearing characteristic frequencies f fif 、f bsf 、f bpfi 、f bpfo , forming the vector f bf , where ε bf is the bearing characteristic error threshold, fif, bsf, bpfi, bpfo are rolling bearing characteristic coefficients;

[0169] S4-4, xci 0 The signals are extracted separately cxThe amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f bf The frequency and the corresponding amplitude form a two-dimensional array E, which can be used to quantitatively analyze the motor fault;

[0170] or xci 1 The signal is extracted F(x) at f cx 、f gear The amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f gear 、f bf The frequency and the corresponding amplitude form a two-dimensional array F, which can be used to quantitatively analyze the gearbox fault.

[0171] In summary, the present invention relates to a method for automatically identifying the speed stabilization stage and filtering out abnormal signals during the operation of the traction mechanism of the quay crane trolley. The method comprises the following steps: the long-term data is segmented in detail through data slicing, and some abnormal data are filtered out by using vibration energy and impact; indicators sensitive to stable working conditions and speed-changing working conditions are selected, and the data of the stable stage are screened out by the similarity between the segmented data indicators; the data of the stable stage are subjected to tensor singular spectrum decomposition, and the linear kurtosis is used to screen out valuable signals to complete signal reconstruction; the corresponding frequency components are extracted by combining different fault characteristic frequency calculation formulas, so as to realize quantitative analysis and automatic diagnosis of the faults of the trolley traction mechanism, which has important practicality and engineering value.

[0172] The various devices selected in this application are all universal standard parts or parts known to those skilled in the art, and their structures and principles can be known to those skilled in the art through technical manuals or conventional experimental methods.

[0173] In the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0174] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0175] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A data cleaning and fault diagnosis method for the traction mechanism of the quay crane trolley, It is characterized in that The method comprises the following steps: S1, a vibration acceleration sensor is provided at the measuring point of the motor drive end of the trolley or the measuring point on the gear box of the trolley for signal collection and preprocessing; S2, screening the collected signals; Screening the collected signal includes the following steps: S2-1, shutdown and abnormal impact data elimination; Divide the motor side point signal or gearbox measurement point signal into I groups of signals y with a length of N i (n), i = 1, 2 ... I; n = 1, 2 ... N, and perform the first data screening, and record the data after the first data screening as the slice index B; S2-2, stable phase identification; According to the slice index B from the signal y i (n) Filter out several groups of signals yc with length N j (n), and obtain the segmented RMS, centroid frequency and envelope spectrum kurtosis of each group of signals to form a feature vector and calculate the correlation coefficient of the feature vectors of two adjacent groups of signals, and then perform a second data screening to form a slice index E, and then perform a third data screening on the slice index E and enter the next stage; S2-3, data reconstruction; Use the slice index E to slice the signal from the motor side point or the gearbox measurement point yc j (n) and complete the signal reconstruction in index order, and finally obtain the signal xc 0 or signal xc 1 ; S3, signal processing; Signal processing includes the following steps: S3-1, signal xc 0 or signal xc 1 Perform adaptive tensor singular spectrum decomposition to obtain l IMF components; S3-2, calculate the linear kurtosis of each IMF component; S3-3, select the IMF components whose linear kurtosis is greater than the linear kurtosis threshold for reconstruction, and obtain the denoised signal xci 0 and xci 1 . S4. Identify fault characteristics and conduct quantitative analysis of gearbox faults.

2. A data cleaning and fault diagnosis method for the traction mechanism of the quay crane trolley as claimed in claim 1, It is characterized in that In step S1, signal acquisition and preprocessing includes the following steps: S1-1, trigger collection and equipped with sampling device; S1-2, the measuring point at the driving end of the trolley motor or the measuring point on the gearbox of the trolley continuously collects vibration acceleration and calculates the root mean square value. When the root mean square value exceeds the set trigger threshold, the sampling device immediately collects the vibration acceleration signal with a sampling frequency of fs and a duration of T; S1-3. Remove DC bias components and low-frequency disturbances from the signal.

3. A data cleaning and fault diagnosis method for the traction mechanism of the quay crane trolley as claimed in claim 1, It is characterized in that In step S2-1, the first data screening includes the following steps: S2-1-1. Calculate the root mean square (RMS) and kurtosis factor (KU) value of each group of I signals yi(n) using the following formula: Where σ is the standard deviation of yi(n); S2-1-2. Compare RMS i , KU i With the power-on threshold T RMS and abnormal shock threshold T ku The size of the slice index is retained to form the array A in accordance with the following relationship: Rms i >T RMS and KU i <T ku S2-1-3. Sort array A by index from small to large, and divide A into g one-dimensional arrays based on continuity, retaining the array with the longest length, thus completing the first data screening.

4. A data cleaning and fault diagnosis method for the traction mechanism of a quay crane trolley as claimed in claim 1, It is characterized in that In step S2-2, the stable phase identification includes the following steps: S2-2-1. From signal y according to slice index B i (n) and obtain J groups of signals yc with a length of N j (n), j = 1, 2…J; S2-2-2. Calculate signal yc j (n) The segmented RMS, centroid frequency and envelope spectrum kurtosis of each group of signals; S2-2-3, the signal yc j (n) The calculated segment RMS, center of gravity frequency and envelope spectrum kurtosis value constitute the signal feature vector C j , and calculate the correlation coefficient of two adjacent groups of signal feature vectors; S2-2-4, determine the relationship between the correlation coefficient and the correlation coefficient deviation threshold, perform a second data screening, and record the data after the second data screening as the slice index E; S2-2-5, determine the relationship between the slice number of slice index E and the slice number threshold, perform a third data screening, and the slice index E that meets the third data screening proceeds to the next step; 5. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that Segmental root mean square (RMS jm ) is calculated as follows: Select M frequency ranges [f m ,f m+1 ], m=1,2…M, and yc is calculated by the following formula j (n) Perform frequency domain bandpass filtering to obtain the filtered signal ycl jm (m), Where e is a natural constant, j 1 is the imaginary unit, H(k) is the frequency response function, Where t is the duration of each signal segment; For each segmented signal ycl after filtering jm (m) Calculate the root mean square and obtain the segmented root mean square (RMS jm ).

6. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that Center of gravity frequency (FC j ) is calculated as follows: Among them, F j is the signal yc j (n) The frequency domain signal is obtained by fast Fourier transform, where f is the frequency.

7. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that Envelope spectral kurtosis (SK j ) is calculated as follows: Among them, e j is the signal yc j (n) Envelope signal obtained after Hilbert demodulation, σ e for e j Standard deviation.

8. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that The calculation method of the correlation coefficient of two adjacent groups of signal feature vectors is as follows: Among them, r is the correlation coefficient, Cov(C j ,C j+1 ) is the covariance of two adjacent signal eigenvectors, is the variance of the eigenvector.

9. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that In step S2-2-4, the second data screening includes the following steps: S2-2-4-1. Retain the slice index of the correlation coefficients of two adjacent groups of signals that satisfy the following relationship to form a one-dimensional array D, r j-1 >e r Among them, ε r is the correlation coefficient deviation threshold; S2-2-4-2. Sort the one-dimensional array D by index from small to large, and divide D into h one-dimensional arrays based on continuity. Keep the array with the longest length, which is the slice index E finally retained after the second data screening.

10. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that In step S2-2-5, the specific location of the third data screening is as follows: Determine whether the number of slices of slice index E satisfies the following relationship, if yes, proceed to the next step: length(E)>ε l Where, length(E) is the number of slices of slice index E, ε l is the slice number threshold.

11. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that In step S3-1, the specific steps of adaptive tensor singular spectrum decomposition are as follows: S3-1-1, xc 0 or xc 1 Constructed into a third-order tensor Z, the tensor Z is expressed as follows, S3-1-2. Perform singular value decomposition on the tensor Z and obtain the following: Where L is the first-order dimension of the tensor S; The tensor multiplication operation rules mentioned above are as follows: Among them, variable b 1 ,b 2 ,…b c+d-2o is a positive integer, a represents the order of the tensors involved in the tensor multiplication operation, <1,1> means that the order of operations of tensor X and tensor Y during tensor multiplication is positive; c is the order of tensor X, d is the order of tensor Y; S 1 is the second-order dimension of the tensor X, S o is the dimension of the tensor X at the o+1th order; S3-1-3, let A l =U× 1,<1,1> S l,: × 2,<1,1> V T And the tensor A l The first column is imf l , then we get the following: Among them, IMF l Signal xc 0 The lth IMF component of .

12. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 11, It is characterized in that In step S3-2, the specific calculation formula of linear kurtosis is as follows: Suppose the sample sequence of the signal is {X}, and the sample values ​​are arranged in order of size as x (1) ≤x (2) ≤x (3) ≤…≤x (n) , p is the total number of samples in the sequence, and the unbiased estimate of the signal sample weight probability moment is: Where r is the order of the random variable sequence.

13. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 12, It is characterized in that In step S3-3, the specific formula for screening IMF components is as follows: L_kurtosis>ε lk Among them, ε lk is the linear kurtosis threshold.

14. A method for data cleaning and fault diagnosis of a trolley traction mechanism of a quay crane as claimed in claim 1, It is characterized in that In step S4, fault feature identification includes the following steps: S4-1, signal xci 0 or signal xci 1 Perform fast Fourier transform to obtain the frequency domain signal F(x), and extract the frequency domain signal F(x) in [f c0 -ε fc ,f c0 +ε fc The frequency coordinate corresponding to the maximum amplitude value in the interval is denoted as f c , where f c0 is the preset ideal frequency, ε fc is the frequency error threshold; S4-2, extract the frequency domain signal F(x) in [q*(f c -ε fc ),q*(f c +ε fc )] or in [w*(f mesh -ε gear ),w*(f mesh +ε gear The frequency coordinate corresponding to the maximum amplitude value in the interval is recorded as the frequency multiple fc q ,q=1,2…Q, or gear meshing frequency fg w ,w=1,2…W, Q, W are the order; multiply the frequency fc of each order of the frequency conversion q Combined into vector f cx , or the meshing frequencies of the gears at each order fg w Combined into vector f gear , where ε gear is the error of gear meshing frequency, f mesh is the gear meshing frequency; S4-3, signal xci 0 or signal xci 1 Perform discrete Hilbert transform and fast Fourier transform to obtain the envelope spectrum of the component and take the modulus to obtain the frequency domain signal H(x). Extract H(x) in the following four intervals [fif*f c -ε bf ,fif*f c +ε bf ], [bsf*f c -ε bf ,bsf*f c +ε bf ]、[bpfi*f c -ε bf ,bpfi*f c +ε bf ], [bpfo*f c -ε bf ,bpfo*f c +ε bf The frequency coordinates corresponding to the maximum amplitude values ​​within ] are recorded as bearing characteristic frequencies f fif 、f bsf 、f bpfi 、f bpfo , forming the vector f bf , where ε bf is the bearing characteristic error threshold, fif, bsf, bpfi, bpfo are rolling bearing characteristic coefficients; S4-4, xci 0 The signals are extracted separately cx The amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f bf The frequency and the corresponding amplitude form a two-dimensional array E, which can be used to quantitatively analyze the motor fault; or xci 1 The signal is extracted F(x) at f cx 、f gear The amplitude corresponding to the frequency, H(x) at f bf The amplitude corresponding to the frequency, f cx 、f gear 、f bf The frequency and the corresponding amplitude form a two-dimensional array F, which can be used to quantitatively analyze the gearbox fault.

Images