Motor fault diagnosis method, system and medium based on multi-sensor data fusion

Abstract

The embodiment of the present invention discloses a motor fault diagnosis method, system and medium based on multi-sensor data fusion, the method comprising: obtaining initial signal data of each sensor; performing noise reduction processing on the initial signal data through singular value decomposition; extracting intrinsic mode function components of the initial signal data after noise reduction, and obtaining reconstructed data according to the intrinsic mode function components; constructing a back propagation neural network based on the reconstructed data, introducing the crossover probability and mutation probability in the genetic algorithm to correct the weights and thresholds of the back propagation neural network, and determining a hybrid network; obtaining the fault diagnosis probability of each sensor according to the hybrid network; determining an evidence reliability factor according to the inherent reliability coefficient of each sensor and the inherent reliability coefficient of the hybrid network; correcting the fault diagnosis probability of each sensor according to the evidence reliability factor, determining the evidence probability, predicting motor faults according to the evidence probability, and improving the accuracy and credibility of the overall diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of motor fault diagnosis, and in particular to a motor fault diagnosis method, system and medium based on multi-sensor data fusion. Background Art

[0002] With the rapid development of industrialization, informatization, and intelligentization, the global industrial automation process is also advancing continuously, and a variety of equipment is widely used in various important fields. As the core component of equipment operation, electric motors are in increasing demand and product quality requirements are becoming increasingly higher. Their reliability directly affects, and even determines, the stability and continuity of the entire production process. They play an irreplaceable and important role in industrial production. During daily operation, motors are often prone to failure due to improper user operation, poor maintenance, and aging of internal insulation. Once a failure occurs, the motor stops operating, directly disrupting the continuity of the production process, seriously endangering the lives and health of on-site workers, and easily causing serious safety accidents and huge economic losses. Therefore, the diagnosis of motor failures is extremely necessary and has strong economic benefits and practical significance.

[0003] Traditional motor fault diagnosis mostly relies on data from a single sensor. This reliance on a single sensor carries reliability risks, and a single sensor can only provide limited information, failing to fully reflect the motor's operating status. In reality, when a motor fault occurs, the diagnostic results from sensors of different types and locations can vary. Combining data from different sensors for a joint diagnosis can significantly improve diagnostic reliability. Summary of the Invention

[0004] Based on this, it is necessary to propose a motor fault diagnosis method, system and medium based on multi-sensor data fusion to address the above problems.

[0005] A motor fault diagnosis method based on multi-sensor data fusion, the method comprising:

[0006] Get the initial signal data of each sensor.

[0007] The initial signal data is denoised by singular value decomposition.

[0008] Extracting intrinsic mode function components of the initial signal data after noise reduction, and obtaining reconstructed data according to the intrinsic mode function components.

[0009] A back propagation neural network is constructed based on the reconstructed data, and the crossover probability and mutation probability in the genetic algorithm are introduced to correct the weights and thresholds of the back propagation neural network to determine a hybrid network.

[0010] The fault diagnosis probability of each sensor is obtained according to the hybrid network.

[0011] An evidence reliability factor is determined according to the inherent reliability coefficients of the sensors and the inherent reliability coefficient of the hybrid network.

[0012] The fault diagnosis probability of each sensor is corrected according to the evidence reliability factor to determine the evidence probability, and the motor fault prediction is performed based on the evidence probability.

[0013] The denoising process of the initial signal data by singular value decomposition specifically includes:

[0014] The Hankel matrix of the initial signal data is determined by phase space reconstruction, and the number of rows and columns of the Hankel matrix is ​​determined according to the number of signal samples of the initial signal data.

[0015] Performing singular value decomposition on the Hankel matrix to obtain singular values, and determining the degree of change between adjacent singular values.

[0016] It is determined that the degree of change between the current singular value and the next singular value has a sudden change, and the singular value is the dividing point.

[0017] All singular values ​​before the demarcation point are retained, and the remaining singular values ​​are set to zero, and then a new Hankel matrix is ​​obtained according to the inverse process of the singular value decomposition.

[0018] According to the inverse process of the phase space reconstruction, the new Hankel matrix is ​​converted into the initial signal data after noise reduction.

[0019] The determining of the Hankel matrix of the initial signal by phase space reconstruction, and determining the number of rows and columns of the Hankel matrix according to the number of signal samples of the initial signal, specifically includes:

[0020] pass Determine the Hankel matrix of the initial signal data, where x1, x2, ..., x N is the initial signal data, m is the number of rows of the Hankel matrix, n is the number of columns of the Hankel matrix, D m×n is the signal subspace without noise interference, W m×n is the noise signal subspace.

[0021] according to The number of rows of the Hankel matrix is ​​determined, where N is the number of signal samples.

[0022] The number of columns of the Hankel matrix is ​​determined according to n=N+1-m.

[0023] The step of performing singular value decomposition on the Hankel matrix to obtain singular values ​​and determining the degree of change between each singular value and the next singular value specifically includes:

[0024] according to The Hankel matrix is ​​subjected to singular value decomposition, where m is the number of rows of the Hankel matrix, n is the number of columns of the Hankel matrix, U and V are both orthogonal matrices, and Σ is a diagonal matrix.

[0025] according to Get the singular values, where S = diag(σ1,σ2,…,σ r ), r is the rank of the Hankel matrix, r∈[1,m], σ i is the i-th singular value in the Hankel matrix, and σ i ≥σ i+1 , i∈[1,r-1].

[0026] according to Determine how much each singular value varies from the next, where a i is the i-th singular value, a max is the maximum value among the singular values, a min is the minimum value among the singular values.

[0027] The extracting of the intrinsic mode function components of the initial signal data after noise reduction and obtaining the reconstructed data according to the intrinsic mode function components specifically includes:

[0028] Random white noise with zero mean and equal variance is added to the initial signal data after noise reduction to obtain a signal group after white noise is added.

[0029] Performing empirical mode decomposition on each signal in the signal group after adding white noise, and extracting several groups of intrinsic mode function components and residuals.

[0030] An eigenmode function component matrix and a residual matrix are determined according to the plurality of groups of eigenmode function components and residuals.

[0031] The intrinsic mode function component mean and the residual mean are calculated column by column for the intrinsic mode function component matrix and the residual matrix respectively.

[0032] Reconstructed data is obtained according to the intrinsic mode function component mean and the residual mean.

[0033] The step of obtaining reconstructed data according to the intrinsic mode function component mean and the residual mean specifically includes:

[0034] according to Get the reconstructed data, where C iis the mean value corresponding to each column in the eigenmode function component matrix, i∈[1,w], w is the number of eigenmode function components, v is the number of random white noise components added, R is the residual mean, r iw is the residual of the i-th column in the residual matrix.

[0035] The step of constructing a back propagation neural network based on the reconstructed data, correcting the weights and thresholds of the back propagation neural network by using the crossover probability and the mutation probability, and determining the hybrid network specifically includes:

[0036] The number of nodes in the input layer, hidden layer and output layer of the back propagation neural network is determined according to the reconstructed data, and the back propagation neural network is constructed.

[0037] The population size and evolution times of the particle swarm, the initial position and initial velocity of each particle, the inertia weight and the acceleration factor are set. Each particle represents a set of weights and thresholds of the back propagation neural network.

[0038] According to the inertia weight and the acceleration factor, the initial velocity and initial position of each particle are updated until a preset population size is reached, and the optimal velocity and optimal position of each particle are obtained.

[0039] The optimal position of the particle is changed according to the crossover probability and mutation probability.

[0040] According to the optimal position of the changed particle, the individual optimal fitness and the population optimal fitness of the particle swarm are determined to obtain the optimal particle swarm.

[0041] The weights and thresholds of the back propagation neural network are modified according to the optimal particle swarm until a preset number of evolutions is reached, thereby determining a hybrid network.

[0042] The method of correcting the fault diagnosis probability of each sensor according to the evidence reliability factor, determining the evidence probability, and predicting the motor fault based on the evidence probability specifically includes:

[0043] according to m ij ′Determine the probability of evidence, where m ij ′=β i m ij , m ij is the fault diagnosis probability of each sensor in the jth case of the i-th evidence, i∈[1,L′], j∈[1,N], L′ is the number of independent evidences, N′ is the number of evaluation levels, m ij ′ is the optimized fault diagnosis probability of each sensor, and Θ is the completely uncertain situation in the fault situation.

[0044] A motor fault diagnosis system based on multi-sensor data fusion, the system comprising:

[0045] The initial signal data acquisition module is used to obtain the initial signal data of each sensor.

[0046] The signal denoising module is used to perform denoising on the initial signal data through singular value decomposition.

[0047] The reconstructed data acquisition module is used to extract the intrinsic mode function components of the initial signal data and obtain reconstructed data according to the intrinsic mode function components.

[0048] The hybrid network determination module is used to construct a back propagation neural network based on the reconstructed data, introduce the crossover probability and mutation probability in the genetic algorithm to correct the weights and thresholds of the back propagation neural network, and determine the hybrid network.

[0049] The module for obtaining the fault diagnosis probability of each sensor is used to obtain the fault diagnosis probability of each sensor according to the hybrid network.

[0050] The evidence reliability factor determination module is used to determine the evidence reliability factor according to the inherent reliability coefficients of the sensors and the inherent reliability coefficient of the hybrid network.

[0051] The motor fault prediction module is used to correct the fault diagnosis probability of each sensor according to the evidence reliability factor, determine the evidence probability, and predict the motor fault based on the evidence probability.

[0052] A computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the steps of the above method.

[0053] The embodiments of the present invention have the following beneficial effects:

[0054] The present invention is based on the initial signal data provided by multiple sensors to improve the accuracy and robustness of fault diagnosis. By extracting intrinsic mode function components through the empirical mode decomposition method, a complex signal can be decomposed into a series of intrinsic mode function components with different frequencies and amplitudes, which helps to reveal the inherent local characteristics of the signal. In combination with the intrinsic mode function extraction process, it helps to remove the noise components in the signal, improve the signal quality, and make subsequent analysis more accurate. The reconstructed data based on the intrinsic mode function components is closer to the true form of the signal, which helps to more accurately reflect the operating status of the equipment. Furthermore, the back propagation neural network constructed based on the reconstructed data has a strong nonlinear mapping ability, can handle complex nonlinear relationships, and is suitable for scenarios such as fault diagnosis. Combining the crossover probability and mutation probability to correct the weights and thresholds of the neural network can optimize the network structure, improve the generalization ability and diagnostic accuracy of the network, and obtain a hybrid network. The hybrid network can output the fault diagnosis probability of each sensor, providing a quantitative basis for subsequent fault prediction and decision-making. Furthermore, by comprehensively considering the reliability coefficients of sensors and hybrid networks, reasonable weights are assigned to different sensors and diagnostic results by determining the evidence reliability factor, thereby improving the accuracy and credibility of the overall diagnosis. The fault diagnosis probability of each sensor is corrected according to the evidence reliability factor, and the evidence probability is determined. Motor faults are predicted based on the evidence probability, potential faults are discovered in advance, and strong support is provided for equipment maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] in:

[0057] Figure 1 A flow chart of an embodiment of a motor fault diagnosis method based on multi-sensor data fusion provided by the present invention;

[0058] Figure 2 A flow chart of another embodiment of the motor fault diagnosis method based on multi-sensor data fusion provided by the present invention;

[0059] Figure 3 A flow chart of another embodiment of the motor fault diagnosis method based on multi-sensor data fusion provided by the present invention;

[0060] Figure 4 A flow chart of an embodiment of a motor fault diagnosis system based on multi-sensor data fusion provided by the present invention;

[0061] Figure 5 This is a schematic structural diagram of an embodiment of the medium provided by the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0063] like Figure 1 As shown, Figure 1 A motor fault diagnosis method based on multi-sensor data fusion is provided in accordance with an embodiment of the present invention. The motor fault diagnosis method based on multi-sensor data fusion comprises:

[0064] S101: Acquire initial signal data of each sensor.

[0065] S102: Performing noise reduction processing on the initial signal data by singular value decomposition.

[0066] Exemplarily, the initial signal data is collected by a plurality of data collection sensors and stored, and further, the initial signal data is subjected to noise reduction processing by singular value decomposition.

[0067] Specifically, for any initial signal data containing noise X(N)={x1,x2,…,x N}, construct the m×n order Hankel matrix through phase space reconstruction:

[0068]

[0069] Among them, x1, x2, …, x N is the initial signal data, m is the number of rows of the Hankel matrix, n is the number of columns of the Hankel matrix, D m×n is the signal subspace without noise interference, W m×n is the noise signal subspace;

[0070] The number of rows and columns of the Hankel matrix is ​​determined according to the number of signal samples of the initial signal data, as shown in the following formula:

[0071]

[0072] Wherein, N is the number of signal samples, N=m+n-1.

[0073] Furthermore, the constructed Hankel matrix is ​​subjected to singular value decomposition, as shown below:

[0074]

[0075] Among them, U is an orthogonal matrix, V is the transpose of the orthogonal matrix, and Σ is a diagonal matrix. At this time:

[0076]

[0077] S=diag(σ1,σ2,…,σ r )(1≤r≤m);

[0078] Where r is the rank of the Hankel matrix H, σ i is the i-th singular value of the matrix H, and σ i ≥σ i+1 (i=1,2,…,r-1).

[0079] At the same time, determine the degree of change b between adjacent singular values i , as shown below:

[0080]

[0081] Among them, a i is the i-th singular value, a max is the maximum value among the singular values, a min is the minimum value among the singular values. If the degree of change between the current singular value and the next singular value changes suddenly, it indicates that this point is the dividing point between the noise-free interference signal and the noise signal. All singular values ​​before the dividing point are retained, and the singular values ​​after the dividing point are set to 0. The new Hankel matrix is ​​obtained according to the inverse process of singular value decomposition. The new Hankel matrix is ​​converted into the initial signal data after noise reduction according to the inverse process of phase space reconstruction.

[0082] S103: Extracting intrinsic mode function components of the initial signal data after noise reduction, and obtaining reconstructed data according to the intrinsic mode function components.

[0083] For example, random white noise {n1(t), n2(t), ..., n v (t)}, then the initial signal data after adding random noise becomes a set of signal groups {x1(t),x2(t),…,x v (t)}, the relationship between the signal containing random white noise and the initial signal data after noise reduction without white noise is as follows:

[0084]

[0085] Among them, n v is the vth random white noise, xv is the initial signal data after denoising after the vth white noise is added, and v is the number of components of random white noise added.

[0086] Furthermore, for the signal group {x1(t),x2(t),…,x v (t)} for each signal x i (t)(i=1,2,…,v) are subjected to EMD decomposition respectively to obtain v groups of intrinsic mode function components {[c 11 ,c 12 ,…,c 1w ],[c 21 ,c 22 ,…,c 2w ],…[c v1 ,c v2 ,…,c vm ]}(denoted as C) and v group margin [r 1w ,r 2w ,…,r vw ] T Denoted as (R), the eigenmode function component matrix is ​​shown as follows:

[0087]

[0088] Where v is the number of rows of the intrinsic mode function component matrix, that is, the number of components of the added random white noise, and w is the number of columns of the intrinsic mode function component matrix, that is, the number of intrinsic mode function components.

[0089] The residual matrix is ​​shown as follows:

[0090]

[0091] Where v is the number of rows in the residual matrix, which is the same as the number of rows in the intrinsic mode function component matrix, and w is the number of columns in the residual matrix, which is the same as the number of columns in the intrinsic mode function component matrix.

[0092] Furthermore, the intrinsic mode function component matrix and the residual matrix are respectively calculated column by column, and the reconstructed data is obtained according to the intrinsic mode function component mean and the residual mean. Specifically, the matrix C is calculated column by column to obtain the mean value [C1, C2, ..., C w ] is shown in the following formula:

[0093]

[0094] The mean R of the m group margins is as follows:

[0095]

[0096] The reconstructed data x'(t) is shown in the following formula:

[0097]

[0098] S104: Construct a back propagation neural network based on the reconstructed data, introduce the crossover probability and mutation probability in the genetic algorithm to modify the weights and thresholds of the back propagation neural network, and determine the hybrid network.

[0099] Exemplarily, the number of nodes in the input layer, hidden layer and output layer of the back-propagation neural network is determined based on the reconstructed data, and the back-propagation neural network is constructed; the population size and number of evolutions of the particle swarm, the initial position and initial velocity of each particle, the inertia weight and the acceleration factor are set, and each particle represents a set of weights and thresholds of the back-propagation neural network; according to the inertia weight and the acceleration factor, the initial velocity and initial position of each particle are updated until the preset population size is reached, and the optimal velocity and optimal position of each particle are obtained; the optimal position of the particle is changed according to the crossover probability and the mutation probability; according to the optimal position of the changed particle, the individual optimal fitness and the population optimal fitness of the particle swarm are determined to obtain the optimal particle swarm; according to the optimal particle swarm, the weights and thresholds of the back-propagation neural network are corrected until the preset number of evolutions is reached, and the hybrid network is determined.

[0100] Specifically, define the inertia weight ω(k):

[0101] ω(k)=ω start -(ω start -ω end )×k / T max ;

[0102] Among them, ω start is the initial inertia weight, ω end is the inertia weight of the maximum number of iterations, k is the current number of iterations, T max is the maximum number of iterations.

[0103] Define the particle acceleration factors c1 and c2:

[0104] c1=c2=2.05.

[0105] S105: Obtaining the fault diagnosis probability of each sensor according to the hybrid network.

[0106] Exemplarily, the sensor data to be diagnosed is input into the trained hybrid network, and the fault diagnosis probability of each sensor is calculated through the forward propagation process of the hybrid network.

[0107] S106: Determine the evidence reliability factor according to the inherent reliability coefficient of each sensor and the inherent reliability coefficient of the hybrid network.

[0108] For example, the calculation of the evidence reliability factor is mainly related to two factors: the inherent reliability coefficient g of each sensor itself and the reliability coefficient s of the hybrid network.

[0109] It should be noted that the inherent reliability coefficient of each sensor is generally determined by experience and is not limited here. The reliability coefficient of the hybrid network generally refers to the reliability of the fault diagnosis probability assignment of each sensor calculated after the data is processed by the hybrid network. In the calculation of this method, it is mainly related to the accuracy of the hybrid network, which is denoted as tru. The relationship between the evidence reliability factor and the two reliability coefficients is defined as follows:

[0110] β i =g i ×s i ;

[0111]

[0112] Among them, β i is the evidence reliability factor, g i is the inherent reliability coefficient of the multi-sensor, s i is the reliability coefficient of the hybrid network, e is the base of the natural logarithm, tru i is the credibility of the hybrid network used, which mainly refers to the accuracy in this method, where 0≤g i ≤1,0≤tru i ≤1.

[0113] S107: Correct the fault diagnosis probability of each sensor according to the evidence reliability factor, determine the evidence probability, and predict the motor fault based on the evidence probability.

[0114] For example, suppose that an evaluation result is composed of L′ pieces of independent evidence m i (i=1,2,…,L′) evaluation, the recognition framework is composed of N′ evaluation levels θ n (n=1,2,…,N′), and redistribute the probability of each fault state of the original evidence according to the evidence reliability factor, and correct the fault diagnosis probability of each sensor. The principle of defining the evidence reliability factor to optimize the basic probability assignment is shown in the following formula:

[0115] m ij ′=β i m ij (i=1,2,…,L′,j=1,2,…,N′);

[0116] Among them, m ijis the fault diagnosis probability of each sensor in the jth case of the i-th evidence, i∈[1,L′], j∈[1,N], L′ is the number of independent evidences, N′ is the number of evaluation levels, m ij ′ is the optimized fault diagnosis probability of each sensor, and Θ is the completely uncertain situation in the fault situation. The global uncertainty probability in the allocated evidence, that is, the evidence probability, is shown as follows:

[0117]

[0118] Among them, m iΘ ' is the probability of evidence.

[0119] Motor fault prediction is performed using evidence probability. However, when the evidence reliability factor is higher, the corresponding probability assignment of the fusion result is more concentrated, and the prediction result is more accurate. When the evidence reliability factor becomes lower, the corresponding peak probability of the fault state becomes lower, the global uncertainty increases, and the prediction result is less accurate.

[0120] As can be seen from the above description, the present invention improves the accuracy and robustness of fault diagnosis based on the initial signal data provided by multiple sensors. By extracting the intrinsic mode function components through the empirical mode decomposition method, a complex signal can be decomposed into a series of intrinsic mode function components with different frequencies and amplitudes, which helps to reveal the inherent local characteristics of the signal. In combination with the intrinsic mode function extraction process, it helps to remove the noise components in the signal, improve the signal quality, and make subsequent analysis more accurate. The reconstructed data based on the intrinsic mode function components is closer to the true form of the signal, which helps to more accurately reflect the operating status of the equipment. Furthermore, the back propagation neural network constructed based on the reconstructed data has a strong nonlinear mapping ability, can handle complex nonlinear relationships, and is suitable for scenarios such as fault diagnosis. Combining the crossover probability and mutation probability to correct the weights and thresholds of the neural network can optimize the network structure, improve the generalization ability and diagnostic accuracy of the network, and obtain a hybrid network. The hybrid network can output the fault diagnosis probability of each sensor, providing a quantitative basis for subsequent fault prediction and decision-making. Furthermore, by comprehensively considering the reliability coefficients of sensors and hybrid networks, reasonable weights are assigned to different sensors and diagnostic results by determining the evidence reliability factor, thereby improving the accuracy and credibility of the overall diagnosis. The fault diagnosis probability of each sensor is corrected according to the evidence reliability factor, and the evidence probability is determined. Motor faults are predicted based on the evidence probability, potential faults are discovered in advance, and strong support is provided for equipment maintenance.

[0121] like Figure 2 As shown, Figure 2 A motor fault diagnosis method based on multi-sensor data fusion, comprising:

[0122] S201: Acquire initial signal data of each sensor.

[0123] S202: Determine the Hankel matrix of the initial signal data through phase space reconstruction, and determine the number of rows and columns of the Hankel matrix according to the number of signal samples of the initial signal data.

[0124] For example, specifically, for any initial signal data containing noise X(N)={x1, x2, ..., x N}, construct the m×n order Hankel matrix through phase space reconstruction:

[0125]

[0126] Among them, x1, x2, …, x N is the initial signal data, m is the number of rows of the Hankel matrix, n is the number of columns of the Hankel matrix, D m×n is the signal subspace without noise interference, W m×n is the noise signal subspace;

[0127] The number of rows and columns of the Hankel matrix is ​​determined according to the number of signal samples of the initial signal data, as shown in the following formula:

[0128] According to the basic inequality principle, when the number of rows and columns is equal or closest, the product of the two is the largest, which satisfies the condition that the product of the number of rows and columns of the Hankel matrix is ​​the largest, thereby enabling the signal and noise to be fully separated. The maximum product of the number of rows and columns depends mainly on the number of signal samples used. The number of rows and columns of the Hankel matrix can be determined based on the number of signal sampling points, that is,

[0129]

[0130] Wherein, N is the number of signal samples, n=N+1-m.

[0131] S203: Perform singular value decomposition on the Hankel matrix to obtain singular values, and determine the degree of change between adjacent singular values.

[0132] For example, the constructed Hankel matrix is ​​subjected to singular value decomposition, as shown in the following formula:

[0133]

[0134] Among them, U is an orthogonal matrix, V is the transpose of the orthogonal matrix, and Σ is a diagonal matrix. At this time:

[0135]

[0136] S=diag(σ1,σ2,…,σr )(1≤r≤m);

[0137] Where r is the rank of the Hankel matrix H, σ i is the i-th singular value of the matrix H, and σ i ≥σ i+1 (i=1,2,…,r-1).

[0138] At the same time, determine the degree of change b between adjacent singular values i , as shown below:

[0139]

[0140] Among them, a i is the i-th singular value, a max is the maximum value among the singular values, a min is the minimum value among the singular values.

[0141] S204: Determine if a sudden change occurs between the current singular value and the next singular value, and the singular value is a dividing point.

[0142] S205: retain all singular values ​​before the demarcation point, set the remaining singular values ​​to zero, and obtain a new Hankel matrix according to the inverse process of singular value decomposition.

[0143] S206: According to the inverse process of phase space reconstruction, the new Hankel matrix is ​​converted into the initial signal data after noise reduction.

[0144] Exemplarily, if the degree of change between the current singular value and the next singular value changes suddenly, it indicates that the point is the dividing point between the noise-free interference signal and the noise signal. All singular values ​​before the dividing point are retained, and the singular values ​​after the dividing point are set to 0. Then, a new Hankel matrix is ​​obtained according to the inverse process of singular value decomposition, and the new Hankel matrix is ​​converted into the initial signal data after noise reduction according to the inverse process of phase space reconstruction.

[0145] S207: adding random white noise with zero mean and equal variance to the initial signal data after noise reduction to obtain a signal group after adding white noise.

[0146] For example, random white noise {n1(t), n2(t), ..., n v (t)}, then the initial signal data after adding random noise becomes a set of signal groups {x1(t),x2(t),…,x v (t)}, the relationship between the signal containing random white noise and the initial signal data after noise reduction without white noise is as follows:

[0147]

[0148] Among them, n v is the vth random white noise, x v is the initial signal data after denoising after the vth white noise is added, and v is the number of components of random white noise added.

[0149] S208: Performing empirical mode decomposition on each signal in the signal group after adding white noise, and extracting several groups of intrinsic mode function components and residuals.

[0150] For example, after adding white noise to the signal group {x1(t),x2(t),…,x v (t)} for each signal x i (t)(i=1,2,…,v) are subjected to EMD decomposition respectively to obtain v groups of intrinsic mode function components {[c 11 ,c 12 ,…,c 1w ],[c 21 ,c 22 ,…,c 2w ],…[c v1 ,c v2 ,…,c vm ]}(denoted as C) and v group margin [r 1w ,r 2w ,…,r vw ] T Denoted as (R).

[0151] S209: Determine an intrinsic mode function component matrix and a residual matrix according to the plurality of groups of intrinsic mode function components and residuals.

[0152] For example, the intrinsic mode function component matrix is ​​shown as follows:

[0153]

[0154] Where v is the number of rows of the intrinsic mode function component matrix, that is, the number of components of the added random white noise, and w is the number of columns of the intrinsic mode function component matrix, that is, the number of intrinsic mode function components.

[0155] The residual matrix is ​​shown as follows:

[0156]

[0157] Where v is the number of rows in the residual matrix, which is the same as the number of rows in the intrinsic mode function component matrix, and w is the number of columns in the residual matrix, which is the same as the number of columns in the intrinsic mode function component matrix.

[0158] S210: Calculating the intrinsic mode function component mean and the residual mean for the intrinsic mode function component matrix and the residual matrix column by column.

[0159] For example, the intrinsic mode function component matrix and the residual matrix are respectively calculated column by column to obtain the intrinsic mode function component mean and the residual mean, and the reconstructed data is obtained according to the intrinsic mode function component mean and the residual mean. Specifically, the matrix C is calculated column by column to obtain the mean value [C1, C2, ..., C w ] is shown in the following formula:

[0160]

[0161] The mean R of the m group margins is as follows:

[0162]

[0163] S211: Obtain reconstructed data according to the intrinsic mode function component mean and the residual mean.

[0164] For example, the reconstructed data x'(t) is shown in the following formula:

[0165]

[0166] S212: Construct a back propagation neural network based on the reconstructed data, introduce the crossover probability and mutation probability in the genetic algorithm to modify the weights and thresholds of the back propagation neural network, and determine the hybrid network.

[0167] S213: Obtaining the fault diagnosis probability of each sensor according to the hybrid network.

[0168] S214: Determine an evidence reliability factor based on the inherent reliability coefficient of each sensor and the inherent reliability coefficient of the hybrid network.

[0169] S215: Correct the fault diagnosis probability of each sensor according to the evidence reliability factor, determine the evidence probability, and predict the motor fault based on the evidence probability.

[0170] It should be noted that steps S212-S15 are Figure 1 This has been discussed in detail in the implementation scenario shown and will not be repeated here.

[0171] From the above description, it can be seen that the present invention adds enough different groups of random white noise to the original signal data multiple times and then performs empirical mode decomposition to obtain a group of intrinsic mode function components. Then, using the characteristic that the mean of random white noise is zero, the overall average value of each group of intrinsic mode function components obtained by all empirical mode decompositions is calculated to eliminate the influence of the added random white noise. The addition of random white noise provides a relatively consistent reference scale for the empirical mode decomposition method to ensure the continuity of each modal function in the time domain, thereby reducing the problem of modal aliasing. And when multiple evidences are synthesized, the credibility coefficients between each evidence are taken into account to improve the accuracy and credibility of the overall diagnosis.

[0172] like Figure 3 As shown, Figure 3 A motor fault diagnosis method based on multi-sensor data fusion, comprising:

[0173] S301: Determine the number of nodes in the input layer, hidden layer, and output layer of the back-propagation neural network according to the reconstructed data, and construct the back-propagation neural network.

[0174] Exemplarily, the number of nodes in the input layer, hidden layer, and output layer of the back-propagation neural network, as well as the algorithm used for training, are determined based on existing data to construct a complete back-propagation neural network.

[0175] S302: Set the population size and evolution times of the particle swarm, the initial position and initial velocity of each particle, the inertia weight and the acceleration factor. Each particle represents a set of weights and thresholds of the back propagation neural network.

[0176] Exemplarily, the population size and evolution times of the particle swarm, the initial position and initial velocity of each particle, the inertia weight, and the acceleration factor are set.

[0177] Specifically, the inertia weight is defined as follows:

[0178] ω(k)=ω start -(ω start -ω end )×k / T max ;

[0179] Among them, ω start is the initial inertia weight, ω end is the inertia weight of the maximum number of iterations, k is the current number of iterations, T max is the maximum number of iterations.

[0180] Define the particle's acceleration factors c1 and c2, c1 = c2 = 2.05.

[0181] S303: Update the initial velocity and initial position of each particle according to the inertia weight and acceleration factor until the preset population size is reached, and obtain the optimal velocity and optimal position of each particle.

[0182] S304: Changing the optimal position of the particle according to the crossover probability and mutation probability in the genetic algorithm.

[0183] S305: According to the optimal position of the changed particle, the individual optimal fitness and the population optimal fitness of the particle swarm are determined to obtain the optimal particle swarm.

[0184] S306: Modify the weights and thresholds of the back propagation neural network according to the optimal particle swarm until a preset number of evolutions is reached, and determine the hybrid network.

[0185] For example, an improved particle swarm optimization algorithm that introduces crossover probability and mutation probability into a genetic algorithm is combined with a back propagation neural network to obtain a hybrid network, which is used to perform training and judgment for motor fault diagnosis.

[0186] Verified by the data set, the results obtained by the improved hybrid network prediction are relatively stable, with the mean square error stable at around 0.04, an average of 0.0414, and a running time varying around 0.5S, with an average of 0.5351s. Compared with the original back propagation neural network, the mean square error is more random, with an average of 0.1004, and the running time is mostly greater than 1s, with an average of 1.4637s. In comparison, the improved PSO_BP network has better reliability.

[0187] From the above description, it can be seen that in the hybrid network, the purpose of using the particle swarm optimization algorithm to optimize the back propagation neural network is to better correct the weights and thresholds of the back propagation neural network, optimize the network structure, prevent the back propagation neural network from falling into local minimum points, and improve the overall operation efficiency of the hybrid network.

[0188] like Figure 4 As shown, Figure 4 A motor fault diagnosis system 10 based on multi-sensor data fusion is provided in accordance with an embodiment of the present invention. The system comprises:

[0189] The initial signal data acquisition module 11 is used to acquire the initial signal data of each sensor.

[0190] The signal denoising module 12 is configured to perform denoising processing on the initial signal data by singular value decomposition.

[0191] The reconstructed data acquisition module 13 is used to extract the intrinsic mode function components of the initial signal data and acquire the reconstructed data according to the intrinsic mode function components.

[0192] The hybrid network determination module 14 is used to construct a back propagation neural network based on the reconstructed data, introduce the crossover probability and mutation probability in the genetic algorithm to modify the weights and thresholds of the back propagation neural network, and determine the hybrid network.

[0193] The module 15 for obtaining the fault diagnosis probability of each sensor is used to obtain the fault diagnosis probability of each sensor according to the hybrid network.

[0194] The evidence reliability factor determination module 16 is used to determine the evidence reliability factor according to the inherent reliability coefficient of each sensor and the inherent reliability coefficient of the hybrid network.

[0195] The motor fault prediction module 17 is used to correct the fault diagnosis probability of each sensor according to the evidence reliability factor, determine the evidence probability, and predict the motor fault based on the evidence probability.

[0196] Exemplarily, in the initial signal data acquisition module 11, initial signal data from each sensor is acquired. In the signal denoising module 12, a Hankel matrix of the initial signal data is determined through phase space reconstruction, and the number of rows and columns of the Hankel matrix is ​​determined based on the number of signal samples of the initial signal data. Singular value decomposition is performed on the Hankel matrix to obtain singular values, and the degree of change between adjacent singular values ​​is determined. When the degree of change between the current singular value and the next singular value suddenly changes, the singular value is determined to be a demarcation point. All singular values ​​before the demarcation point are retained, and the remaining singular values ​​are set to zero. A new Hankel matrix is ​​obtained according to the inverse process of the singular value decomposition. The new Hankel matrix is ​​converted into the denoised initial signal data according to the inverse process of the phase space reconstruction. Furthermore, in the reconstructed data acquisition module 13, random white noise with zero mean and equal variance is added to the initial signal data after denoising to obtain a signal group after adding white noise; empirical mode decomposition is performed on each signal in the signal group after adding white noise to extract several groups of intrinsic mode function components and residuals; the intrinsic mode function component matrix and the residual matrix are determined based on the several groups of intrinsic mode function components and residuals; the intrinsic mode function component mean and the residual mean are calculated column by column for the intrinsic mode function component matrix and the residual matrix respectively; and the reconstructed data is obtained based on the intrinsic mode function component mean and the residual mean. Furthermore, in the hybrid network determination module 14, the number of nodes in the input, hidden, and output layers of the back-propagation neural network is determined based on the reconstructed data, and the back-propagation neural network is constructed. The particle swarm population size and number of evolutions, the initial position and initial velocity of each particle, the inertia weight, and the acceleration factor are set. Each particle represents a set of weights and thresholds for the back-propagation neural network. Based on the inertia weight and acceleration factor, the initial velocity and initial position of each particle are updated until the preset population size is reached, thereby obtaining the optimal velocity and optimal position of each particle. The optimal position of the particle is changed based on the crossover probability and mutation probability. Based on the changed optimal position of the particle, the optimal fitness of the individual particles and the optimal fitness of the population are determined to obtain the optimal particle swarm. Based on the optimal particle swarm, the weights and thresholds of the back-propagation neural network are modified until the preset number of evolutions is reached, thereby determining the hybrid network. Furthermore, in the sensor fault diagnosis probability acquisition module 15, the fault diagnosis probability of each sensor is obtained based on the hybrid network. Finally, in the evidence reliability factor determination module 16, the evidence reliability factor is determined based on the inherent reliability coefficient of each sensor and the inherent reliability coefficient of the hybrid network.

[0197] like Figure 5 As shown, Figure 5 The structure diagram of an embodiment of the medium provided by the present invention. The medium 20 stores at least one computer program 21, which is executed by the processor 22 to implement the following Figure 1 、 Figure 2 ,and Figure 3In one embodiment, the storage medium 20 may be a memory chip, a hard disk, a mobile hard disk, a USB flash drive, an optical disk, or other readable and writable storage tools, or a server.

[0198] The foregoing description of specific embodiments of this specification is intended to be construed as a description of other embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0199] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer-readable storage medium embodiments are described briefly because they are generally similar to the method embodiments. For relevant portions, refer to the description of the method embodiments.

[0200] The apparatus, device, non-volatile computer-readable storage medium and method provided in the embodiments of this specification correspond to each other. Therefore, the apparatus, device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, device, and non-volatile computer storage medium will not be repeated here.

[0201] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0202] For the convenience of description, when describing the above device, various units are divided into functions and described separately. Of course, when implementing this specification, the functions of each unit can be implemented in the same one or more software and / or hardware. It should be understood by those skilled in the art that this specification embodiment can be provided as a method, system, or computer program product. Therefore, this specification embodiment can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification embodiment can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0203] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0204] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0205] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0206] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0207] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0208] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0209] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0210] This specification may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0211] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are described briefly because they are generally similar to the method embodiments. For relevant parts, refer to the description of the method embodiments.

[0212] The above disclosure is merely a preferred embodiment of the present invention and certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A motor fault diagnosis method based on multi-sensor data fusion, characterized in that: The method comprises: Obtaining initial signal data from each sensor; Perform noise reduction on the initial signal data through singular value decomposition; Extracting intrinsic mode function components of the initial signal data after noise reduction, and obtaining reconstructed data according to the intrinsic mode function components; Constructing a back propagation neural network based on the reconstructed data, introducing the crossover probability and mutation probability in the genetic algorithm to modify the weights and thresholds of the back propagation neural network, and determining a hybrid network; Obtaining the fault diagnosis probability of each sensor according to the hybrid network; determining an evidence reliability factor according to the inherent reliability coefficients of the sensors and the inherent reliability coefficient of the hybrid network; Correcting the fault diagnosis probability of each sensor according to the evidence reliability factor, determining the evidence probability, and predicting the motor fault based on the evidence probability; The denoising process of the initial signal data by singular value decomposition specifically includes: determining a Hankel matrix of the initial signal data by phase space reconstruction, and determining the number of rows and columns of the Hankel matrix according to the number of signal samples of the initial signal data; performing singular value decomposition on the Hankel matrix to obtain singular values, and determining a degree of variation between adjacent singular values; Determining that the degree of change between the current singular value and the next singular value suddenly changes, then the singular value is a demarcation point; retaining all singular values ​​before the demarcation point, setting the remaining singular values ​​to zero, and obtaining a new Hankel matrix according to the inverse process of the singular value decomposition; Converting the new Hankel matrix into the initial signal data after noise reduction according to the inverse process of the phase space reconstruction; The extracting the intrinsic mode function components of the initial signal data after noise reduction and obtaining the reconstructed data according to the intrinsic mode function components specifically includes: Adding random white noise with zero mean and equal variance to the initial signal data after noise reduction to obtain a signal group after adding white noise; Performing empirical mode decomposition on each signal in the signal group after adding white noise, and extracting several groups of intrinsic mode function components and residuals; Determining an intrinsic mode function component matrix and a residual matrix according to the plurality of groups of intrinsic mode function components and residuals; Calculating the intrinsic mode function component mean and the residual mean for the intrinsic mode function component matrix and the residual matrix column by column respectively; Obtaining reconstructed data according to the intrinsic mode function component mean and the residual mean; The method of correcting the fault diagnosis probability of each sensor according to the evidence reliability factor, determining the evidence probability, and predicting the motor fault based on the evidence probability specifically includes: according to Determine the probability of evidence, where , For the The first piece of evidence The fault diagnosis probability of each sensor in the case of ], j∈[1, N], is the number of independent evidences, is the number of evaluation levels, is the fault diagnosis probability of each sensor after optimization, For the completely uncertain situation in the fault situation; The relationship between the evidence reliability factor and the reliability coefficient is shown in the following formula: ； ； in, is the evidence reliability factor, is the inherent reliability coefficient of the multi-sensor, is the reliability coefficient of the hybrid network, e is the base of the natural logarithm, is the credibility of the adopted hybrid network, .

2. The motor fault diagnosis method based on multi-sensor data fusion according to claim 1, characterized in that: Determining the Hankel matrix of the initial signal through phase space reconstruction, and determining the number of rows and columns of the Hankel matrix according to the number of signal samples of the initial signal, specifically includes: pass Determine the Hankel matrix of the initial signal data, where is the initial signal data, m is the number of rows of the Hankel matrix, n is the number of columns of the Hankel matrix, is the signal subspace without noise interference, is the noise signal subspace; according to Determine the number of rows of the Hankel matrix, where is the number of signal samples; according to Determine the number of columns of the Hankel matrix.

3. The motor fault diagnosis method based on multi-sensor data fusion according to claim 2 is characterized in that: The step of performing singular value decomposition on the Hankel matrix to obtain singular values ​​and determining a degree of change between each singular value and the next singular value specifically includes: according to Perform singular value decomposition on the Hankel matrix, where m is the number of rows of the Hankel matrix and n is the number of columns of the Hankel matrix. are all orthogonal matrices, is a diagonal matrix; according to Get the singular values, where , is the rank of the Hankel matrix, r∈[1,m], is the first singular values, and , i∈[1,r-1]; according to Determine how much each singular value varies from the next, where For the singular values, is the maximum value among the singular values, is the minimum value among the singular values.

4. The motor fault diagnosis method based on multi-sensor data fusion according to claim 3 is characterized in that: The step of obtaining reconstructed data according to the intrinsic mode function component mean and the residual mean specifically includes: according to Get the reconstructed data, where is the mean value corresponding to each column in the eigenmode function component matrix, i∈[1,w], w is the number of eigenmode function components, , v is the number of components of random white noise added, is the residual mean, , is the residual of the i-th column in the residual matrix.

5. The motor fault diagnosis method based on multi-sensor data fusion according to claim 4 is characterized in that: The back propagation neural network is constructed based on the reconstructed data, and the weights and thresholds of the back propagation neural network are corrected by the crossover probability and the mutation probability to determine the hybrid network, specifically including: Determining the number of nodes in the input layer, hidden layer, and output layer of the back-propagation neural network according to the reconstructed data, and constructing the back-propagation neural network; Set the particle swarm population size and evolution times, each particle's initial position and initial velocity, inertia weight, and acceleration factor. Each particle represents a set of weights and thresholds for the back-propagation neural network. According to the inertia weight and acceleration factor, the initial velocity and initial position of each particle are updated until a preset population size is reached, and the optimal velocity and optimal position of each particle are obtained; Change the optimal position of particles according to the crossover probability and mutation probability; According to the optimal position of the changed particle, determining the individual optimal fitness and the population optimal fitness of the particle swarm, and obtaining the optimal particle swarm; The weights and thresholds of the back propagation neural network are modified according to the optimal particle swarm until a preset number of evolutions is reached, thereby determining a hybrid network.

6. A motor fault diagnosis system based on multi-sensor data fusion, characterized in that: The system comprises: An initial signal data acquisition module is used to acquire the initial signal data of each sensor; A signal denoising module is used to perform denoising on the initial signal data through singular value decomposition; a reconstructed data acquisition module, configured to extract intrinsic mode function components of the initial signal data and acquire reconstructed data according to the intrinsic mode function components; A hybrid network determination module is used to construct a back propagation neural network based on the reconstructed data, introduce the crossover probability and mutation probability in the genetic algorithm to modify the weights and thresholds of the back propagation neural network, and determine the hybrid network; A module for obtaining the fault diagnosis probability of each sensor, configured to obtain the fault diagnosis probability of each sensor according to the hybrid network; an evidence reliability factor determination module, configured to determine an evidence reliability factor based on the inherent reliability coefficients of the sensors and the inherent reliability coefficient of the hybrid network; a motor fault prediction module, configured to modify the fault diagnosis probability of each sensor according to the evidence reliability factor, determine the evidence probability, and perform motor fault prediction based on the evidence probability; The denoising process of the initial signal data by singular value decomposition specifically includes: determining a Hankel matrix of the initial signal data by phase space reconstruction, and determining the number of rows and columns of the Hankel matrix according to the number of signal samples of the initial signal data; performing singular value decomposition on the Hankel matrix to obtain singular values, and determining a degree of variation between adjacent singular values; Determining that the degree of change between the current singular value and the next singular value suddenly changes, then the singular value is a demarcation point; retaining all singular values ​​before the demarcation point, setting the remaining singular values ​​to zero, and obtaining a new Hankel matrix according to the inverse process of the singular value decomposition; The extracting the intrinsic mode function components of the initial signal data after noise reduction and obtaining the reconstructed data according to the intrinsic mode function components specifically includes: Adding random white noise with zero mean and equal variance to the initial signal data after noise reduction to obtain a signal group after adding white noise; Performing empirical mode decomposition on each signal in the signal group after adding white noise, and extracting several groups of intrinsic mode function components and residuals; Determining an intrinsic mode function component matrix and a residual matrix according to the plurality of groups of intrinsic mode function components and residuals; Calculating the intrinsic mode function component mean and the residual mean for the intrinsic mode function component matrix and the residual matrix column by column respectively; Obtaining reconstructed data according to the intrinsic mode function component mean and the residual mean; The method of correcting the fault diagnosis probability of each sensor according to the evidence reliability factor, determining the evidence probability, and predicting the motor fault based on the evidence probability specifically includes: according to Determine the probability of evidence, where , For the The first piece of evidence The fault diagnosis probability of each sensor in the case of ], j∈[1, N], is the number of independent evidences, is the number of evaluation levels, is the fault diagnosis probability of each sensor after optimization, For the completely uncertain situation in the fault situation; The relationship between the evidence reliability factor and the reliability coefficient is shown in the following formula: ； ； in, is the evidence reliability factor, is the inherent reliability coefficient of the multi-sensor, is the reliability coefficient of the hybrid network, e is the base of the natural logarithm, is the credibility of the adopted hybrid network, .

7. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 5.

Images