Bearing fault diagnosis method based on integrated support vector dynamic neural network

Abstract

The invention discloses a bearing fault diagnosis method integrated with a support vector dynamic neural network, and the method comprises the following steps: obtaining vibration data of a key position of a bearing of a rotating machine, carrying out the data preprocessing of the vibration data, and obtaining a preprocessed vibration signal data set; constructing a support vector machine model with a quadratic programming problem form based on the vibration signal data set, and designing a corresponding penalty function and a Lagrange function to simplify the support vector machine model; designing a variable parameter dynamics solving equation, and establishing a final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model; and simultaneously training a plurality of support vector dynamic neural network bearing fault diagnosis models based on a bagging classifier method, carrying out bearing state diagnosis on the vibration signal data set, and outputting a diagnosis result. According to the invention, the accuracy of the machine learning method, the convergence proof of the neural network solver and the interpretability of the intelligent fault diagnosis model are improved.

Description

Technical Field

[0001] The invention relates to the technical field of fault diagnosis of rotating machinery bearings, and in particular to a bearing fault diagnosis method integrating support vector dynamic neural network. Background Art

[0002] Rotating machinery and equipment are important components of industrial fields such as petroleum, chemical industry, and energy. They provide a constant power source for the manufacturing industry and are an important part of industrial machinery. Rotating machinery including steam turbines, gas turbines, centrifugal compressors, generators, water pumps, ventilators, and motors rely on bearings to achieve their rotation and power transmission. Since bearings are usually in a high-speed motion state, once a bearing fails, it may directly lead to the paralysis of the entire mechanical system, causing economic losses and safety accidents. During the working process, due to high speed, high load and changing environment, the failure rate of rolling bearings is usually very high. About 45%-55% of mechanical failures are caused by bearing failures. Therefore, real-time fault diagnosis of bearings is crucial to ensure reliable operation and prevent accidents. Bearing faults usually generate signals with specific characteristics. The standard fault diagnosis method is based on temperature, vibration, acoustic emission and motor current signals. Among all the fault characteristic quantities, vibration signal diagnosis is widely used in bearing fault diagnosis because of its simplicity and accuracy.

[0003] At present, the mainstream vibration analysis method used in industry is to monitor the vibration signal of the bearing system to detect the type (such as damage, looseness) and degree of bearing faults. Vibration analysts first install vibration sensors at key locations of the equipment to record the vibration signals generated during the operation of the bearing, and then diagnose the bearing vibration signals manually using spectrum analysis, time domain analysis, axis trajectory diagram and other technologies. Since manual diagnosis is time-consuming and labor-intensive, researchers have been trying to use intelligent fault diagnosis methods for diagnosis in recent years. Mainstream intelligent fault diagnosis methods include principal component analysis, k-nearest neighbor, support vector machine, Bayesian network, random forest and neural network. However, due to the shortcomings of these methods such as large data demand, long training time and data imbalance, the current intelligent fault diagnosis methods cannot perform fault diagnosis well. Summary of the invention

[0004] In order to overcome the defects and shortcomings of the prior art, the present invention provides an integrated support vector dynamic neural network bearing fault diagnosis method, which combines a support vector machine with a time-varying neural network to improve the accuracy of the machine learning method, the convergence proof of the neural network solver and the interpretability of the intelligent fault diagnosis model.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The second object of the present invention is to provide an integrated support vector dynamic neural network bearing fault diagnosis system;

[0007] A third object of the present invention is to provide a computer readable storage medium;

[0008] A fourth object of the present invention is to provide a computer device.

[0009] The present invention provides a bearing fault diagnosis method using an integrated support vector dynamic neural network, comprising the following steps:

[0010] Acquire vibration data of key positions of bearings of rotating machinery, perform data preprocessing on the vibration data, and obtain a preprocessed vibration signal data set;

[0011] Based on the vibration signal data set, a support vector machine model in the form of a quadratic programming problem is constructed, and the corresponding penalty function and Lagrangian function are designed to simplify the support vector machine model.

[0012] Design the variable parameter dynamics solution equation, establish the final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model;

[0013] Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously, the bearing status diagnosis is performed on the vibration signal dataset, and the diagnosis results are output.

[0014] As a preferred technical solution, the vibration data is preprocessed by using z-score normalization and random undersampling method, wherein the z-score normalization is expressed as:

[0015]

[0016] Among them, z i Represents vibration data, and s are the mean and standard deviation of all samples respectively, x i Represents vibration data after z-score normalization.

[0017] As a preferred technical solution, a support vector machine model in the form of a quadratic programming problem is constructed based on the vibration signal data set, and the corresponding penalty function and Lagrangian function are designed to simplify the support vector machine model, specifically including:

[0018] Define a linear decision boundary, expressed as:

[0019]

[0020] Where w represents the coefficient vector, b represents the bias scalar, represents the operation of projection to high-dimensional space, x i represents the preprocessed vibration signal dataset;

[0021] Define the soft margin support vector machine, expressed as:

[0022]

[0023] st i ≥0

[0024]

[0025] Where C represents the slack variable ζ i The coefficient of , n represents the number of vibration signals;

[0026] Define the Lagrangian multipliers α and β, and transform the optimization problem into the corresponding Lagrangian function, which is expressed as:

[0027]

[0028] Solve to get the simplified Lagrangian function:

[0029]

[0030] Adding Gaussian kernel function K(x i ,x j )replace It is expressed as:

[0031]

[0032] Among them, σ represents the parameter of the kernel function;

[0033] The kernel function is used to fit the nonlinear vibration data into the nonlinear space, and the formula is converted to:

[0034]

[0035] 0≤α1,α2,...,α m ≤C

[0036] After simplifying the model, we get:

[0037]

[0038] stQα=0

[0039] 0≤α≤C

[0040] Among them, Q, H, α, Q are the parameters of the quadratic programming problem, namely:

[0041]

[0042] Y=[y1 y2...y n ]·[y1 y2...y n ] T

[0043]

[0044] in, It is based on the multiplication operation of matrix elements;

[0045] Definition J = [EE] T ∈R 2m×m and d = [C0], where E∈R m×m is the unit matrix, and the final simplified model of the support vector machine model with the quadratic programming problem is:

[0046]

[0047] stQα=0

[0048] Jα≤d

[0049] Simplify the inequality constraints and define the penalty function as:

[0050]

[0051] N i (α) = d i -J i α

[0052] in, is the penalty factor, σ>0 is the parameter of M(·), and after eliminating the inequality constraint by the penalty function method, the quadratic programming QP problem is rewritten as:

[0053]

[0054] stQα=0.

[0055] As the preferred technical solution, the variable parameter dynamics solution equation is designed, and the final integrated support vector dynamic neural network bearing fault diagnosis model is established to solve the support vector machine model, which specifically includes:

[0056] Introducing the Lagrange multiplier λ and solving it, we get: μ(t) = Sz(t) - g(t);

[0057] in,

[0058] Define the error function μ(t):

[0059] μ(t)=Sz(t)-g(t)

[0060] According to the variable parameter neural dynamics method, we get:

[0061]

[0062] Among them, Ψ(·) is a monotonically increasing odd function, and the final single integrated support vector dynamic neural network bearing fault diagnosis model is:

[0063]

[0064] The final integrated support vector dynamic neural network bearing fault diagnosis model solves the support vector machine model to achieve parameter optimization.

[0065] As a preferred technical solution, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously based on the bagging classifier method, specifically including:

[0066] Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously, a multi-classification problem is decomposed into several sub-problems, and each sub-problem is solved with an independent single support vector dynamic neural network bearing fault diagnosis model;

[0067] The random undersampling method is applied to the training process of the integrated support vector dynamic neural network bearing fault diagnosis model. For each bearing state, the data of this type and the remaining data are extracted as positive samples and negative samples respectively. In the prediction process, k support vector dynamic neural network bearing fault diagnosis models are used to distinguish the positive and negative classes of the result output, where k is the number of possible bearing states.

[0068] The bearing status is diagnosed on the vibration signal data set, and the diagnosis results are output. The final diagnosis results are:

[0069]

[0070] Among them, r i represents the output of the i-th support vector dynamic neural network bearing fault diagnosis model, and argmax is the output of the i-th support vector dynamic neural network bearing fault diagnosis model with the largest value.

[0071] The present invention also provides an integrated support vector dynamic neural network bearing fault diagnosis system, including: a vibration data acquisition module, a data preprocessing module, a support vector machine model construction module, a support vector machine model simplification module, an equation solving design module, a bearing fault diagnosis model construction module, a model training module, and a diagnosis result output module;

[0072] The vibration data acquisition module is used to acquire vibration data of key positions of bearings of rotating machinery;

[0073] The data preprocessing module is used to perform data preprocessing on the vibration data to obtain a preprocessed vibration signal data set;

[0074] The support vector machine model building module is used to build a support vector machine model in the form of a quadratic programming problem based on the vibration signal data set;

[0075] The support vector machine model simplification module is used to design corresponding penalty functions and Lagrangian functions to simplify the support vector machine model;

[0076] The equation solving design module is used to design variable parameter dynamics solution equations;

[0077] The bearing fault diagnosis model building module is used to establish a final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model;

[0078] The model training module is used to simultaneously train multiple support vector dynamic neural network bearing fault diagnosis models based on a bagging classifier method;

[0079] The diagnosis result output module is used to perform bearing status diagnosis on the vibration signal data set and output the diagnosis result.

[0080] The present invention also provides a computer-readable storage medium storing a program, which, when executed by a processor, implements the above-mentioned integrated support vector dynamic neural network bearing fault diagnosis method.

[0081] The present invention also provides a computer device, comprising a processor and a memory for storing a program executable by the processor, wherein when the processor executes the program stored in the memory, the above-mentioned integrated support vector dynamic neural network bearing fault diagnosis method is implemented.

[0082] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0083] (1) The present invention adopts a technical solution of combining a support vector machine with a time-varying neural network, which solves the problem of long training time of existing deep learning networks, reduces the number of model parameters, and enables the model to be fitted with less data.

[0084] (2) The present invention adopts the penalty function method and variable parameter neural dynamics design in the process of parameter optimization solution, which solves the problem of long fitting time of previous neural network models and can converge to 0 in a short time. This enables the method of the present invention to quickly fit the parameters of the model and greatly reduce the fitting time of the model, thereby improving the real-time performance of the bearing monitoring technology.

[0085] (3) The present invention uses a random under-sampling data processing method to solve the data imbalance problem in traditional intelligent fault diagnosis methods, pre-processes the unbalanced data for bearing fault diagnosis, and uses cluster intelligence to use multiple classifiers to perform intelligent diagnosis on vibration data, thereby reducing the impact of unbalanced data on a single classifier. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 The figure is a flow chart of the bearing fault diagnosis method integrating support vector dynamic neural network of the present invention. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0088] Example 1

[0089] like Figure 1 As shown, this embodiment provides a bearing fault diagnosis method integrated with a support vector dynamic neural network, comprising the following steps:

[0090] S1: Install acceleration vibration sensors at key positions of the bearings of rotating machinery. The collected data set containing n vibration signals and n sample labels can be defined as:

[0091] Z={(z1,y1),(z2,y2),...,(z n ,y n )}

[0092] Among them, z i ∈R m represents the i-th eigenvector, y i ∈{-1,1} represents the state of the i-th vibration signal; if y i is equal to 1, then the i-th sample is a positive sample, otherwise, when y i When it is equal to -1, the i-th sample is a negative sample.

[0093] S2: The vibration data is preprocessed using z-score normalization and random undersampling methods. For a vibration signal z contained in the vibration signal dataset i , using the following z-score normalization method:

[0094]

[0095] in, and s are set as the mean and standard deviation of all samples respectively to eliminate the adverse effects caused by individual odd sample data. All vibration signal data sets after normalization are D = {(x1, y1), (x2, y2), ..., (x n ,y n )}.

[0096] Since the bagging method is used for integrated diagnosis, there are k classifiers, and each classifier uses a random undersampling method to process data. That is, assuming that there are k categories of bearing conditions, each bagging classifier only takes the data of one category as the positive sample, and randomly takes out 90% of the data from the remaining k-1 categories for negative samples.

[0097] S3: Based on the existing data, a support vector machine model in the form of a quadratic programming problem is established, and the corresponding penalty function and Lagrangian function are designed to simplify the support vector machine model, including:

[0098] S31: For the data set D generated in step S2, a hyperplane of a support vector machine is defined as a linear decision boundary including a coefficient vector w and a bias scalar b, which can be expressed as:

[0099]

[0100] in, It is the operation of projecting x into a higher-dimensional space.

[0101] The present invention introduces the relaxation variable ζ i To reduce the impact of constraints, reduce the impact of some abnormal data and unbalanced data on the entire model, define a soft margin support vector machine, whose equation can be rewritten as:

[0102]

[0103] st i ≥0

[0104]

[0105] Where C represents the slack variable ζ i The coefficient of .

[0106] S32: Define Lagrangian multipliers α and β, and transform the optimization problem into the corresponding Lagrangian function, that is:

[0107]

[0108] By solving the above formula, we can get the simplified Lagrangian function, namely:

[0109]

[0110] The present invention adds a kernel function K(x i ,x j ) instead K(·) is a Gaussian kernel function, which is in the form of:

[0111]

[0112] Among them, σ represents the parameter of the kernel function.

[0113] Using the kernel function to fit the nonlinear vibration data into the nonlinear space, the formula can be converted to:

[0114]

[0115] 0≤α1,α2,...,α m ≤C

[0116] Among them, σ represents the parameter of the kernel function, and K(·) is a Gaussian kernel function;

[0117] And simplifying the above model, we can get:

[0118]

[0119] stQα=0

[0120] 0≤α≤C

[0121] Among them, Q, H, α, Q are the parameters of the quadratic programming problem, namely:

[0122]

[0123] Y=[y1 y2...y n ]·[y1 y2...y n ] T

[0124]

[0125] in, It is a matrix element-based multiplication operation.

[0126] For the convenience of calculation, the present invention defines J = [EE] T ∈R 2m×m and d = [C0], where E∈R m×m is the identity matrix. Therefore, the final simplified model of the support vector machine model with the form of a quadratic programming problem can be obtained as follows:

[0127]

[0128] stQα=0

[0129] Jα≤d.

[0130] S33: According to the support vector machine model based on the quadratic programming problem in step S32, which includes equality and inequality constraints, the inequality constraints are simplified and a penalty function is defined:

[0131]

[0132] N i (α) = d i -J i α

[0133] in, is the penalty factor, σ>0 is the parameter of M(·), and after eliminating the inequality constraint by the penalty function method, the quadratic programming QP problem can be rewritten as:

[0134]

[0135] stQα=0.

[0136] S4: Design the variable parameter dynamics solution equation and establish the final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model, including:

[0137] Introducing the Lagrange multiplier λ and solving it, we get μ(t) = Sz(t) - g(t). Among them, Assuming that both z and g change with time t, define an error function μ(t):

[0138] μ(t)=Sz(t)-g(t)

[0139] According to the variable parameter neural dynamics method Among them, Ψ(·) is a monotonically increasing odd function, and the final single integrated support vector dynamic neural network bearing fault diagnosis model is obtained:

[0140]

[0141] In this embodiment, the final integrated support vector dynamic neural network bearing fault diagnosis model solves the support vector machine model, that is, realizes parameter optimization, and the solved support vector machine model can be used for fault diagnosis. The integrated support vector dynamic neural network bearing fault diagnosis model is equivalent to a model that integrates the support vector machine model and the dynamic neural network.

[0142] S5: Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously to diagnose vibration signals. A multi-classification problem is decomposed into several sub-problems, and each sub-problem is solved with an independent single support vector dynamic neural network bearing fault diagnosis model to ensure the accuracy and stability of the output results, realize the multi-classification fault diagnosis method of the model, and improve the generalization ability and fault diagnosis ability of the model;

[0143] After the training of multiple support vector dynamic neural network bearing fault diagnosis models is completed, the bearing status diagnosis is performed and the diagnosis results are output;

[0144] The random undersampling method is applied to the training process of the integrated support vector dynamic neural network bearing fault diagnosis model. During the training process of a single support vector dynamic neural network bearing fault diagnosis model, the bearing data of each state is used to train an independent support vector dynamic neural network bearing fault diagnosis model. For each bearing state, the data of this class and the remaining 90% of the data are extracted as positive samples and negative samples respectively. This method helps to reduce the impact of the imbalance of fault samples of different categories and enhance the robustness of each classifier. In the prediction process, k support vector dynamic neural network bearing fault diagnosis models are used to distinguish the positive and negative classes of the samples for the result output, where k is the number of possible bearing states. For the output results of all support vector dynamic neural network bearing fault diagnosis models, the output of the i-th support vector dynamic neural network bearing fault diagnosis model can be expressed as r i , then the final diagnosis result is:

[0145]

[0146] Among them, argmax is the output of the largest i-th support vector dynamic neural network bearing fault diagnosis model. For example, if there are 10 bearing states, 10 support vector dynamic neural network bearing fault diagnosis models are generated. When the output r3 of the third support vector dynamic neural network bearing fault diagnosis model is the largest output value among all support vector dynamic neural network bearing fault diagnosis models, the third fault state is the final diagnosis result of the integrated support vector dynamic neural network bearing fault diagnosis model.

[0147] Example 2

[0148] This embodiment provides an integrated support vector dynamic neural network bearing fault diagnosis system, which is used to implement the integrated support vector dynamic neural network bearing fault diagnosis method of the above-mentioned embodiment 1. The system includes: a vibration data acquisition module, a data preprocessing module, a support vector machine model construction module, a support vector machine model simplification module, an equation solving design module, a bearing fault diagnosis model construction module, a model training module, and a diagnosis result output module;

[0149] In this embodiment, the vibration data acquisition module is used to acquire vibration data of key positions of bearings of the rotating machinery;

[0150] In this embodiment, the data preprocessing module is used to perform data preprocessing on the vibration data to obtain a preprocessed vibration signal data set;

[0151] In this embodiment, the support vector machine model building module is used to build a support vector machine model in the form of a quadratic programming problem based on the vibration signal data set;

[0152] In this embodiment, the support vector machine model simplification module is used to design corresponding penalty functions and Lagrangian functions to simplify the support vector machine model;

[0153] In this embodiment, the equation solving design module is used to design a variable parameter dynamics solution equation;

[0154] In this embodiment, the bearing fault diagnosis model building module is used to establish the final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model;

[0155] In this embodiment, the model training module is used to simultaneously train multiple support vector dynamic neural network bearing fault diagnosis models based on the bagging classifier method;

[0156] In this embodiment, the diagnosis result output module is used to perform bearing status diagnosis on the vibration signal data set and output the diagnosis result.

[0157] Example 3

[0158] This embodiment provides a storage medium, which may be a ROM, RAM, disk, CD or other storage medium, and the storage medium stores one or more programs. When the program is executed by the processor, the integrated support vector dynamic neural network bearing fault diagnosis method of embodiment 1 is implemented.

[0159] Example 4

[0160] This embodiment provides a computing device, which can be a desktop computer, a laptop computer, a smart phone, a PDA handheld terminal, a tablet computer or other terminal device with a display function. The computing device includes a processor and a memory. The memory stores one or more programs. When the processor executes the program stored in the memory, the integrated support vector dynamic neural network bearing fault diagnosis method of embodiment 1 is implemented.

[0161] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A bearing fault diagnosis method based on integrated support vector dynamic neural network, characterized in that: The steps include: Acquire vibration data of key positions of bearings of rotating machinery, perform data preprocessing on the vibration data, and obtain a preprocessed vibration signal data set; Based on the vibration signal data set, a support vector machine model in the form of a quadratic programming problem is constructed, and the corresponding penalty function and Lagrangian function are designed to simplify the support vector machine model. Design the variable parameter dynamics solution equation, establish the final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model; Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously, the bearing status diagnosis is performed on the vibration signal dataset, and the diagnosis results are output.

2. The integrated support vector dynamic neural network bearing fault diagnosis method according to claim 1 is characterized in that: The vibration data is preprocessed by using z-score normalization and random undersampling methods, wherein the z-score normalization is expressed as: Among them, z i represents vibration data, z and s are the mean and standard deviation of all samples respectively, x i Represents vibration data after z-score normalization.

3. The integrated support vector dynamic neural network bearing fault diagnosis method according to claim 1 is characterized in that: Based on the vibration signal data set, a support vector machine model in the form of a quadratic programming problem is constructed, and the corresponding penalty function and Lagrangian function are designed to simplify the support vector machine model, including: Define a linear decision boundary, expressed as: Where w represents the coefficient vector, b represents the bias scalar, represents the operation of projection to high-dimensional space, x i represents the preprocessed vibration signal dataset; Define the soft margin support vector machine, expressed as: Where C represents the slack variable ζ i The coefficient of , n represents the number of vibration signals; Define the Lagrangian multipliers α and β, and transform the optimization problem into the corresponding Lagrangian function, which is expressed as: Solve to get the simplified Lagrangian function: Adding Gaussian kernel function K(x i ,x j )replace It is expressed as: Among them, σ represents the parameter of the kernel function; The kernel function is used to fit the nonlinear vibration data into the nonlinear space, and the formula is converted to: After simplifying the model, we get: Among them, Q, H, α, Q are the parameters of the quadratic programming problem, namely: in, It is based on the multiplication operation of matrix elements; Definition J = [EE] T ∈R 2m×m and d = [C0], where E∈R m×m is the unit matrix, and the final simplified model of the support vector machine model with the quadratic programming problem is: Simplify the inequality constraints and define the penalty function as: in, is the penalty factor, σ>0 is the parameter of M(·), and after eliminating the inequality constraint by the penalty function method, the quadratic programming QP problem is rewritten as:

4. The integrated support vector dynamic neural network bearing fault diagnosis method according to claim 3 is characterized in that: Design the variable parameter dynamics solution equation, establish the final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model, including: Introducing the Lagrange multiplier λ and solving it, we get: μ(t) = Sz(t) - g(t); in, Define the error function μ(t): μ(t)=Sz(t)-g(t) According to the variable parameter neural dynamics method, we get: Among them, Ψ(·) is a monotonically increasing odd function, and the final single integrated support vector dynamic neural network bearing fault diagnosis model is: The final integrated support vector dynamic neural network bearing fault diagnosis model solves the support vector machine model to achieve parameter optimization.

5. The integrated support vector dynamic neural network bearing fault diagnosis method according to claim 1 is characterized in that: Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously, including: Based on the bagging classifier method, multiple support vector dynamic neural network bearing fault diagnosis models are trained simultaneously, a multi-classification problem is decomposed into several sub-problems, and each sub-problem is solved with an independent single support vector dynamic neural network bearing fault diagnosis model; The random undersampling method is applied to the training process of the integrated support vector dynamic neural network bearing fault diagnosis model. For each bearing state, the data of this type and the remaining data are extracted as positive samples and negative samples respectively. In the prediction process, k support vector dynamic neural network bearing fault diagnosis models are used to distinguish the positive and negative classes of the result output, where k is the number of possible bearing states. The bearing status is diagnosed on the vibration signal data set, and the diagnosis results are output. The final diagnosis results are: Among them, r i represents the output of the i-th support vector dynamic neural network bearing fault diagnosis model, and argmax is the output of the i-th support vector dynamic neural network bearing fault diagnosis model with the largest value.

6. An integrated support vector dynamic neural network bearing fault diagnosis system, characterized in that: include: Vibration data acquisition module, data preprocessing module, support vector machine model building module, support vector machine model simplification module, equation solving design module, bearing fault diagnosis model building module, model training module, diagnosis result output module; The vibration data acquisition module is used to acquire vibration data of key positions of bearings of rotating machinery; The data preprocessing module is used to perform data preprocessing on the vibration data to obtain a preprocessed vibration signal data set; The support vector machine model building module is used to build a support vector machine model in the form of a quadratic programming problem based on the vibration signal data set; The support vector machine model simplification module is used to design corresponding penalty functions and Lagrangian functions to simplify the support vector machine model; The equation solving design module is used to design variable parameter dynamics solution equations; The bearing fault diagnosis model building module is used to establish a final integrated support vector dynamic neural network bearing fault diagnosis model to solve the support vector machine model; The model training module is used to simultaneously train multiple support vector dynamic neural network bearing fault diagnosis models based on a bagging classifier method; The diagnosis result output module is used to perform bearing status diagnosis on the vibration signal data set and output the diagnosis result.

7. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the integrated support vector dynamic neural network bearing fault diagnosis method as described in any one of claims 1 to 5 is implemented.

8. A computer device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, it implements the integrated support vector dynamic neural network bearing fault diagnosis method as described in any one of claims 1 to 5.

Images