A gearbox fault detection method and system based on intuitionistic fuzzy FS-SVDD

Abstract

The present invention provides a gearbox fault detection method and system based on intuitive fuzzy FS-SVDD, wherein the method comprises: analyzing the collected gearbox vibration signal according to the pre-constructed intuitive fuzzy multi-classification support vector data description to determine the gearbox fault. The gearbox fault detection method and system based on intuitive fuzzy FS-SVDD of the present invention, based on the intuitive fuzzy idea, calculates the intuitive fuzzy score of each data point in the data set, thereby reflecting the dependence of each data point on the category and improving the feature selection performance; based on the introduced intuitive fuzzy idea, constructs an intuitive fuzzy multi-classification support vector data description classification algorithm, and applies it to gearbox fault detection, further improving the recognition accuracy of gearbox faults.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a gearbox fault detection method and system based on intuitionistic fuzzy FS-SVDD. Background Art

[0002] Gear transmission is a common transmission method in mechanical equipment. As a key component in mechanical equipment systems, gearboxes are often affected by special environments and are prone to failure. Therefore, timely diagnosis of their operating status is essential.

[0003] Gearbox fault vibration signals are typically highly random and complex, making it difficult to accurately extract fault features and resulting in poor fault diagnosis. Features of gearbox vibration signals are extracted from different perspectives, including the time domain, frequency domain, and multi-scale entropy. While multi-angle feature extraction methods enrich feature information, they also introduce high dimensionality and redundancy issues in feature sets, necessitating further feature screening. Fisher score (FS) is an effective sample feature selection method. However, existing FS and its improved methods do not consider the influence of outliers and noise in intra-class datasets in feature selection applications, resulting in poor feature selection results.

[0004] In addition, support vector data description (SVDD) was used to identify various mechanical faults, achieving good recognition results. However, existing SVDD and evolutionary algorithms do not consider the impact of outliers and noise points in the training data on the decision sphere, which can affect the fault recognition effect to a certain extent. Summary of the Invention

[0005] One of the purposes of the present invention is to provide a gearbox fault detection method and system based on intuitionistic fuzzy FS-SVDD. The intuitionistic fuzzy idea is used to calculate the intuitionistic fuzzy score of each data point in the data set to reflect the dependence of each data point on the category and improve the feature selection performance. Based on the introduced intuitionistic fuzzy idea, an intuitionistic fuzzy multi-classification support vector data description classification algorithm is constructed and applied to gearbox fault detection to further improve the recognition accuracy of gearbox faults.

[0006] An embodiment of the present invention provides a gearbox fault detection method based on intuitionistic fuzzy FS-SVDD, comprising:

[0007] Based on the pre-built intuitionistic fuzzy multi-classification support vector data description, the collected gearbox vibration signal is analyzed to determine the gearbox fault;

[0008] The steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0009] Extract features from historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set;

[0010] The intuitionistic fuzzy set and Fisher score algorithm are used to optimally screen the high-dimensional feature set and obtain the optimal feature subset;

[0011] Based on the optimal feature subset, the intuitionistic fuzzy set and support vector data description are integrated to construct the intuitionistic fuzzy multi-classification support vector data description.

[0012] Preferably, the collection of the gearbox vibration signal includes:

[0013] The gearbox vibration signal is collected using an acceleration sensor group fixed to the gearbox.

[0014] Preferably, the feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0015] Extracting a preset first number of time domain features as first fault features from a time domain graph of a gearbox vibration signal using a statistical method;

[0016] Extracting a preset second number of frequency domain features as second fault features from a frequency domain diagram of the gearbox vibration signal based on a Fourier transform analysis method;

[0017] The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated. The multi-scale permutation entropy and multi-scale sample entropy are used to extract a preset third number of multi-scale entropy features as the third fault feature;

[0018] The first fault feature, the second fault feature, and the third fault feature are combined to obtain a high-dimensional feature set.

[0019] Preferably, the optimal screening steps for the high-dimensional feature set are as follows:

[0020] Given a training dataset {x i ,y i}, where x i is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label;

[0021] Let any data point x in class j be i The membership degree of is defined as:

[0022]

[0023] In the formula, δ is the preset first parameter, r + is the radius of the positive data set, X + is the class center of the positive class dataset;

[0024] Any data point x in class j i The non-membership degree of is defined as:

[0025] b(x i )=(1-a(x i ))ρ(x i );

[0026] Satisfying 0≤a(x i )+b(x i )≤1, and

[0027]

[0028] Wherein, t>0 and is the preset second parameter;

[0029] Any data point x in class j i The corresponding intuitionistic fuzzy score function is:

[0030]

[0031] Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n}, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is:

[0032]

[0033] in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-type dataset, For the i-th data point in the j-class dataset in feature f k The value of the following;

[0034] Based on each feature f in the feature set F k Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the score, and select the fourth number of features preset before as the optimal feature subset.

[0035] Preferably, the steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0036] Given a feature set {f1, f2, ..., fm} is taken from the training data x of c categories i ∈R m , i=1,2,…,l;

[0037] Class j dataset j=1,2,…,c, datasets other than the jth class Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as:

[0038]

[0039] st ||φ(x i )-a j || 2 ≤R j +ξ i ,ξ i ≥0, (i=1,2,…,n j ),

[0040] ||φ(x k )-a j || 2 ≤R j -ξ k ,ξ k ≥0, (k=1,2,…,ln j ),

[0041] Where a j 、R j are the center and radius of the j-th dataset, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class;

[0042] Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set. j and radius R j .

[0043] Preferably, the collected gearbox vibration signal is analyzed based on the pre-constructed intuitionistic fuzzy multi-classification support vector data description to determine the gearbox fault, including:

[0044] Process the collected gearbox vibration signal to obtain the corresponding optimal feature subset as fault data;

[0045] Calculate the distance between the fault data and the sphere center of each type of data set. The category to which the sphere center closest to the fault data belongs is the class corresponding to the fault data.

[0046] The present invention also provides a gearbox fault detection system based on intuitionistic fuzzy FS-SVDD, comprising: a fault analysis unit and a construction module; wherein the fault analysis unit analyzes the collected gearbox vibration signal based on the intuitionistic fuzzy multi-classification support vector data description pre-constructed by the construction module to determine the gearbox fault;

[0047] The steps for constructing the intuitionistic fuzzy multi-classification support vector data description in the construction module are as follows:

[0048] Extract features from historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set;

[0049] The intuitionistic fuzzy set and Fisher score algorithm are used to optimally screen the high-dimensional feature set and obtain the optimal feature subset;

[0050] Based on the optimal feature subset, the intuitionistic fuzzy set and support vector data description are integrated to construct the intuitionistic fuzzy multi-classification support vector data description.

[0051] Preferably, the collection of the gearbox vibration signal includes:

[0052] The gearbox vibration signal is collected using an acceleration sensor group fixed to the gearbox.

[0053] Preferably, the feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0054] Extracting a preset first number of time domain features as first fault features from a time domain graph of a gearbox vibration signal using a statistical method;

[0055] Extracting a preset second number of frequency domain features as second fault features from a frequency domain diagram of the gearbox vibration signal based on a Fourier transform analysis method;

[0056] The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated. The multi-scale permutation entropy and multi-scale sample entropy are used to extract a preset third number of multi-scale entropy features as the third fault feature;

[0057] The first fault feature, the second fault feature, and the third fault feature are combined to obtain a high-dimensional feature set.

[0058] Preferably, the optimal screening steps for the high-dimensional feature set are as follows:

[0059] Given a training dataset {x i ,y i}, where x i is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label;

[0060] Let any data point x in class j be i The membership degree of is defined as:

[0061]

[0062] In the formula, δ is the preset first parameter, r + is the radius of the positive dataset, C + is the class center of the positive class dataset;

[0063] Any data point x in class j i The non-membership degree of is defined as:

[0064] b(x i )=(1-a(x i ))ρ(x i );

[0065] Satisfying 0≤a(x i )+b(x i )≤1, and

[0066]

[0067] Wherein, t>0 and is the preset second parameter;

[0068] Any data point x in class j i The corresponding intuitionistic fuzzy score function is:

[0069]

[0070] Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n}, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is:

[0071]

[0072] in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-type dataset, For the i-th data point in the j-class dataset in feature f k The value of the following;

[0073] Based on each feature f in the feature set Fk Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the score, and select the fourth number of features preset before as the optimal feature subset.

[0074] Preferably, the steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0075] Given a feature set {f1, f2, ..., f m} is taken from the training data x of c categories i ∈R m , i=1,2,…,l;

[0076] Class j dataset j=1,2,…,c, datasets other than the jth class Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as:

[0077]

[0078] st ||φ(x i )-a j || 2 ≤R j +ξ i ,ξ i ≥0, (i=1,2,…,n j ),

[0079] ||φ(x k )-a j || 2 ≤R j -ξ k ,ξ k ≥0, (k=1,2,…,ln j ),

[0080] Where a j 、R j are the center and radius of the j-th dataset, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class;

[0081] Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set. j and radius R j .

[0082] Preferably, the fault analysis unit performs the following operations:

[0083] Process the collected gearbox vibration signal to obtain the corresponding optimal feature subset as fault data;

[0084] Calculate the distance between the fault data and the sphere center of each type of data set. The category to which the sphere center closest to the fault data belongs is the class corresponding to the fault data.

[0085] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0086] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0088] Figure 1 Schematic diagram of the steps for constructing intuitionistic fuzzy multi-classification support vector data description according to an embodiment of the present invention;

[0089] Figure 2 Schematic diagram of a gearbox fault detection system based on intuitionistic fuzzy FS-SVDD in an embodiment of the present invention. DETAILED DESCRIPTION

[0090] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0091] The embodiment of the present invention provides a gearbox fault detection method based on intuitionistic fuzzy FS-SVDD, comprising:

[0092] Based on the pre-built intuitionistic fuzzy multi-classification support vector data description, the collected gearbox vibration signal is analyzed to determine the gearbox fault;

[0093] Gearbox vibration signal collection includes: using an accelerometer group fixed to the gearbox to collect gearbox vibration signals. The accelerometer group includes multiple accelerometers to detect data in different directions, at least three accelerometers, with the monitoring directions of each pair of accelerometers perpendicular to each other. In addition, a gearbox operating status data collection platform can be built, and the accelerometers can be configured on the gearbox operating status data collection platform.

[0094] Among them, such as Figure 1 As shown in Figure 2, the steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0095] Step 1: Extract features from the historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set;

[0096] The feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0097] A statistical method is used to extract a preset first number of time domain features as a first fault feature from a time domain diagram of a gearbox vibration signal; for example, 16 time domain features are extracted, of which 10 dimensioned statistical parameters are the mean, standard deviation, square root amplitude, absolute mean, skewness, kurtosis, variance, maximum value, minimum value, and peak-to-peak value in the time domain diagram, and 6 dimensionless statistical parameters are the waveform index, peak index, variance pulse index, edge index, skewness index, and kurtosis index in the time domain diagram;

[0098] Extracting a preset second number of frequency domain features as second fault features from the frequency domain diagram of the gearbox vibration signal based on the Fourier transform analysis method; for example, extracting 13 frequency domain features, the 13 frequency domain features being the spectrum amplitude mean, spectrum amplitude variance, spectrum amplitude skewness, spectrum amplitude kurtosis, center of gravity frequency, spectrum amplitude standard deviation, spectrum root mean square value, main frequency band position, frequency energy concentration, frequency dispersion, frequency skewness, frequency kurtosis, and spectrum normalized mean in the frequency domain diagram;

[0099] The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated. Multiscale permutation entropy and multiscale sample entropy are used to extract a preset third number (30) of multiscale entropy features as the third fault feature. The parameter requirements of the multiscale permutation entropy and the multiscale sample entropy are as follows: the scale factors of the multiscale permutation entropy and the multiscale sample entropy are both 15, the embedding dimension of the multiscale permutation entropy is 6, the time delay is 1, the embedding dimension of the multiscale sample entropy is 2, and the similarity tolerance is 0.15 times the sample standard deviation.

[0100] Merging the first fault feature, the second fault feature, and the third fault feature to obtain a high-dimensional feature set;

[0101] Step 2: Use the intuitionistic fuzzy set and Fisher score algorithm to optimally screen the high-dimensional feature set and obtain the optimal feature subset;

[0102] The optimal screening steps for high-dimensional feature sets are as follows:

[0103] Given a training dataset {x i ,y i}, where xi is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label;

[0104] Let any data point x in class j be i The membership degree of is defined as:

[0105]

[0106] In the formula, δ is the preset first parameter, r + is the radius of the positive dataset, C + is the class center of the positive class dataset;

[0107] Any data point x in class j i The non-membership degree of is defined as:

[0108] b(x i )=(1-a(x i ))ρ(x i );

[0109] Satisfying 0≤a(x i )+b(x i )≤1, and

[0110]

[0111] Wherein, t>0 and is the preset second parameter;

[0112] Any data point x in class j i The corresponding intuitionistic fuzzy score function is:

[0113]

[0114] Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n}, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is:

[0115]

[0116] in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-type dataset, For the i-th data point in the j-class dataset in feature f k The value below;

[0117] Based on each feature f in the feature set F k Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the score, and select the fourth number (m and m<n) of features preset before as the optimal feature subset;

[0118] Step 3: Based on the optimal feature subset, the intuitionistic fuzzy set and support vector data description are integrated to construct the intuitionistic fuzzy multi-classification support vector data description.

[0119] The steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0120] Given a feature set {f1, f2, ..., f m} is taken from the training data x of c categories i ∈R m , i=1,2,…,l;

[0121] Class j dataset j=1,2,…,c, datasets other than the jth class Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as:

[0122]

[0123] st ||φ(x i )-a j || 2 ≤R j +ξ i ,ξ i ≥0, (i=1,2,…,n j ),

[0124] ||φ(x k )-a j || 2 ≤R j -ξ k ,ξ k ≥0, (k=1,2,…,ln j ),

[0125] Where a j 、R j are the center and radius of the j-th dataset, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class;

[0126] Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set. j and radius R j .

[0127] In one embodiment, the collected gearbox vibration signal is analyzed based on a pre-constructed intuitionistic fuzzy multi-classification support vector data description to determine the gearbox fault, including:

[0128] Process the collected gearbox vibration signal to obtain the corresponding optimal feature subset as fault data;

[0129] Calculate the distance between the fault data and the sphere center of each type of data set. The category to which the sphere center closest to the fault data belongs is the class corresponding to the fault data.

[0130] That is, the decision function is:

[0131] In the formula, x is the data sample to be tested, is the feature map, label(x) is the category to which x belongs, c is the number of categories of data, a j The center of the sphere of the j-th dataset.

[0132] In one embodiment, when determining a gearbox fault, the distance from the fault data to the sphere center of each class data set is calculated, and then the distances are compared. This requires calculating the distances for all class data sets, which is time-consuming. In this embodiment, an association value is assigned to each class data set, and the association value is used to guide the next class data set for distance calculation. The class data sets are also screened, and class data sets with association values ​​less than or equal to a preset threshold do not need to be calculated for distance values, thereby improving calculation efficiency.

[0133] The association value is determined by calculating the distance between the sphere centers of each data set, querying the pre-configured distance-association value correspondence table based on the distance, and determining the corresponding association value; the association value guidance method also needs to be combined with the previously calculated distance. Specifically, the distance is used to query the pre-configured distance-association value table corresponding to the corresponding class data set to determine the next or subsequent N calculated class data sets.

[0134] In addition, for the first class data set used for calculation, the class data set to be calculated first can be determined by predicting the probability of failure corresponding to each class data set in the gearbox; and the class data set with the minimum distance can be determined quickly with a high probability.

[0135] Among them, the steps of predicting the probability of failure are as follows: obtaining the operation record data of the gearbox; performing feature extraction on the operation record data and constructing a prediction parameter set based on the extracted feature values; calling the fault prediction data set corresponding to the prediction parameter set in the fault prediction library from the pre-configured fault prediction library through the prediction parameter set, and determining the probability corresponding to each type of data set based on each data in the fault prediction data set; wherein, the prediction parameter set includes: parameters representing the model of the gearbox, parameters representing the material of each gear of the gearbox, parameters representing the operating time of the gearbox in each mode, parameters representing the start time of the gearbox, etc.; each data in the fault prediction data set corresponds to the probability parameter of each gear failure, that is, it corresponds to each type of data set.

[0136] The present invention also provides a gearbox fault detection system based on intuitionistic fuzzy FS-SVDD, such as Figure 2 As shown, it includes: a fault analysis unit 1 and a construction module 2; wherein the fault analysis unit 1 analyzes the collected gearbox vibration signal based on the intuitionistic fuzzy multi-classification support vector data description constructed in advance by the construction module 2 to determine the gearbox fault;

[0137] The steps for constructing module 2 to construct the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0138] Extract features from historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set;

[0139] The intuitionistic fuzzy set and Fisher score algorithm are used to optimally screen the high-dimensional feature set and obtain the optimal feature subset;

[0140] Based on the optimal feature subset, the intuitionistic fuzzy set and support vector data description are integrated to construct the intuitionistic fuzzy multi-classification support vector data description.

[0141] Among them, the collection of gearbox vibration signals includes:

[0142] The gearbox vibration signal is collected using an acceleration sensor group fixed to the gearbox.

[0143] Preferably, the feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0144] Extracting a preset first number of time domain features as first fault features from a time domain graph of a gearbox vibration signal using a statistical method;

[0145] Extracting a preset second number of frequency domain features as second fault features from a frequency domain diagram of the gearbox vibration signal based on a Fourier transform analysis method;

[0146] The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated. The multi-scale permutation entropy and multi-scale sample entropy are used to extract a preset third number of multi-scale entropy features as the third fault feature;

[0147] The first fault feature, the second fault feature, and the third fault feature are combined to obtain a high-dimensional feature set.

[0148] Among them, the optimal screening steps for high-dimensional feature sets are as follows:

[0149] Given a training dataset {x i ,y i |, where x i is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label;

[0150] Let any data point x in class j be i The membership degree of is defined as:

[0151]

[0152] In the formula, δ is the preset first parameter, r + is the radius of the positive dataset, C + is the class center of the positive class dataset;

[0153] Any data point x in class j i The non-membership degree of is defined as:

[0154] b(x i )=(1-a(x i ))ρ(x i );

[0155] Satisfying 0≤a(x i )+b(x i )≤1, and

[0156]

[0157] Wherein, t>0 and is the preset second parameter;

[0158] Any data point x in class j i The corresponding intuitionistic fuzzy score function is:

[0159]

[0160] Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n}, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is:

[0161]

[0162] in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-type dataset, For the i-th data point in the j-class dataset in feature f k The value of the following;

[0163] Based on each feature f in the feature set F k Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the score, and select the fourth number of features preset before as the optimal feature subset.

[0164] The steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows:

[0165] Given a feature set {f1, f2, ..., f m} is taken from the training data x of c categories i ∈R m , i=1,2,…,l;

[0166] Class j dataset j=1,2,…,c, datasets other than the jth class Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as:

[0167]

[0168] st ||φ(x i )-a j || 2 ≤R j +ξ i ,ξ i ≥0, (i=1,2,…,n j ),

[0169] ||φ(x k )-a j || 2 ≤R j -ξ k ,ξ k ≥0, (k=1,2,…,ln j),

[0170] Where a j 、R j are the center and radius of the j-th dataset, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class;

[0171] Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set. j and radius R j .

[0172] The fault analysis unit performs the following operations:

[0173] Process the collected gearbox vibration signal to obtain the corresponding optimal feature subset as fault data;

[0174] Calculate the distance between the fault data and the sphere center of each type of data set. The category to which the sphere center closest to the fault data belongs is the class corresponding to the fault data.

[0175] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A gearbox fault detection method based on intuitionistic fuzzy FS-SVDD, characterized in that: include: According to the pre-built intuitive fuzzy multi-classification support vector data description, the collected gearbox vibration signal is analyzed to determine the gearbox fault; Among them, the steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows: Extract features from historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set; Intuitive fuzzy sets and Fisher score algorithm are used to optimally screen high-dimensional feature sets and obtain the optimal feature subset; Based on the optimal feature subset, the intuitive fuzzy set and support vector data description are integrated to construct the intuitive fuzzy multi-classification support vector data description. The steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows: Given a feature set {f1, f2, ..., f m } is taken from the training data x of c categories i ∈R m , i＝1,2，…，l； The jth dataset Datasets other than class j Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as: st‖φ(x i )-a j ‖ 2 ≤R j +ξ i ,x i ≥0, (i=1,2,…,n j ), ‖φ(x k )-a j ‖ 2 ≤R j -x k ,x k ≥0, (k=1,2,…,ln j ), In the formula, a j , R j are the center and radius of the j-th data set, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class; Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set j and radius R j .

2. The gearbox fault detection method based on intuitionistic fuzzy FS-SVDD as claimed in claim 1 is characterized in that: The collection of gearbox vibration signals includes: The gearbox vibration signal is collected by using an acceleration sensor group fixed to the gearbox.

3. The gearbox fault detection method based on intuitionistic fuzzy FS-SVDD as claimed in claim 1 is characterized in that: The feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows: Extracting a preset first number of time domain features as first fault features in a time domain diagram of a gearbox vibration signal by using a statistical method; Extracting a preset second number of frequency domain features as second fault features in the frequency domain diagram of the gearbox vibration signal based on a Fourier transform analysis method; The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated, and a preset third number of multi-scale entropy features are extracted using multi-scale arrangement entropy and multi-scale sample entropy as the third fault feature; The first fault feature, the second fault feature and the third fault feature are combined to obtain a high-dimensional feature set.

4. The gearbox fault detection method based on intuitionistic fuzzy FS-SVDD as claimed in claim 1, characterized in that: The optimal screening steps for high-dimensional feature sets are as follows: Given a training dataset {x i ,y i }, where x i is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label; Let any data point x in the jth class i The membership degree of is defined as: In the formula, δ is the preset first parameter, r + is the radius of the positive dataset, C + is the class center of the positive class data set; Any data point x in the jth class i The non-membership degree of is defined as: b(x i )=(1-a(x i ))ρ(x i ); Satisfies 0≤a(x i )+b(x i )≤1, and Wherein, t>0 and is the preset second parameter; Any data point x in the jth class i The corresponding intuitionistic fuzzy score function is: Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n }, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is: in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-class dataset, is the i-th data point in the j-type data set in feature f k The value of the following; Based on each feature f in the feature set F k Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the scores, and select the fourth number of features preset before as the optimal feature subset.

5. The gearbox fault detection method based on intuitionistic fuzzy FS-SVDD as claimed in claim 1, characterized in that: According to the pre-built intuitive fuzzy multi-classification support vector data description, the collected gearbox vibration signal is analyzed to determine the gearbox fault, including: Process the collected gearbox vibration signal to obtain the corresponding optimal feature subset as fault data; Calculate the distance from the fault data to the center of each type of data set. The category to which the center of the sphere closest to the fault data belongs is the class corresponding to the fault data.

6. A gearbox fault detection system based on intuitionistic fuzzy FS-SVDD, characterized in that: include: A fault analysis unit and a construction module; wherein the fault analysis unit analyzes the collected gearbox vibration signal based on the intuitive fuzzy multi-classification support vector data description constructed in advance by the construction module to determine the gearbox fault; Among them, the construction steps of the construction module to construct the intuitionistic fuzzy multi-classification support vector data description are as follows: Extract features from historically collected gearbox vibration signals from multiple angles to obtain a high-dimensional feature set; Intuitive fuzzy sets and Fisher score algorithm are used to optimally screen high-dimensional feature sets and obtain the optimal feature subset; Based on the optimal feature subset, the intuitive fuzzy set and support vector data description are integrated to construct the intuitive fuzzy multi-classification support vector data description. The steps for constructing the intuitionistic fuzzy multi-classification support vector data description are as follows: Given a feature set {f1, f2, ..., f m } is taken from the training data x of c categories i ∈R m , i＝1,2，…，l； The jth dataset Datasets other than class j Let the optimization problem of the intuitionistic fuzzy multi-classification support vector description of the j-th data set be described as: st‖φ(x i )-a j ‖ 2 ≤R j +ξ i ,x i ≥0, (i=1,2,…,n j ), ‖φ(x k )-a j ‖ 2 ≤R j -x k ,x k ≥0, (k=1,2,…,ln j ), In the formula, a j , R j are the center and radius of the j-th data set, respectively. C>0 is the preset penalty parameter. is the intuitionistic fuzzy score of the i-th data point in the j-th data set, is the intuitionistic fuzzy score of the kth data point in the dataset outside the jth class; Introducing kernel function and Lagrangian function, solving through quadratic programming, we can get the sphere center a of the j-th data set j and radius R j .

7. The gearbox fault detection system based on intuitionistic fuzzy FS-SVDD as claimed in claim 6, characterized in that: The collection of gearbox vibration signals includes: The gearbox vibration signal is collected by using an acceleration sensor group fixed to the gearbox.

8. The gearbox fault detection system based on intuitionistic fuzzy FS-SVDD as claimed in claim 6, characterized in that: The feature extraction steps in the process of constructing the intuitionistic fuzzy multi-classification support vector data description are as follows: Extracting a preset first number of time domain features as first fault features in a time domain diagram of a gearbox vibration signal by using a statistical method; Extracting a preset second number of frequency domain features as second fault features in the frequency domain diagram of the gearbox vibration signal based on a Fourier transform analysis method; The gearbox vibration signal is decomposed at different time scales, and the entropy value at each time scale is calculated, and a preset third number of multi-scale entropy features are extracted using multi-scale arrangement entropy and multi-scale sample entropy as the third fault feature; The first fault feature, the second fault feature and the third fault feature are combined to obtain a high-dimensional feature set.

9. The gearbox fault detection system based on intuitionistic fuzzy FS-SVDD as claimed in claim 6, characterized in that: The optimal screening steps for high-dimensional feature sets are as follows: Given a training dataset {x i ,y i }, where x i is an n-dimensional real vector, i = 1, 2, ..., l, y i =1,2,…,c,c≥2 and c is the corresponding class label; Let any data point x in the jth class i The membership degree of is defined as: In the formula, δ is the preset first parameter, r + is the radius of the positive dataset, C + is the class center of the positive class data set; Any data point x in the jth class i The non-membership degree of is defined as: b(x i )=(1-a(x i ))ρ(x i ); Satisfies 0≤a(x i )+b(x i )≤1, and Wherein, t>0 and is the preset second parameter; Any data point x in the jth class i The corresponding intuitionistic fuzzy score function is: Let the n-dimensional feature set of the training data set be F = {f1, f2, ..., f n }, then for any feature f k ∈F, its intuitionistic fuzzy Fisher score is: in, n p is the number of data points of type p, n q is the number of data points of type q, n pq For samples of class p and class q in feature f k The number of data points with the same value, n j is the number of data points of the jth class, is the intuitionistic fuzzy score of the i-th data point in the p-class dataset, is the i-th data point in the j-type data set in feature f k The value of the following; Based on each feature f in the feature set F k Calculate the intuitionistic fuzzy Fisher score NFS(f k ), and sort all the features in the feature set from high to low according to the scores, and select the fourth number of features preset before as the optimal feature subset.

Images