Equipment fault identification method based on evidence neural network

Abstract

The invention discloses an equipment fault identification method based on an evidence neural network, and relates to the technical field of equipment detection. Comprising the steps of obtaining equipment vibration data based on a vibration sensor, preprocessing the equipment vibration data, extracting features of a frequency domain, a time domain and a time-frequency domain of the vibration data as a training data set, constructing an evidence neural network comprising an input layer, a hidden layer and an output layer, inputting the training data set into the evidence neural network, and optimizing network parameters. And obtaining a trained evidence neural network, and inputting the vibration data to be detected into the trained evidence neural network to identify an equipment fault. According to the invention, the position and the type of the fault of the equipment can be accurately positioned and judged.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment detection, and particularly to a method for identifying equipment faults based on an evidence neural network. Background Art

[0002] There are a large number of special equipment and devices for chemical industry in traditional chemical enterprises. When a device fails, due to the complexity of the use environment and technological process, it is very time-consuming and has a low accuracy rate to locate the fault and determine the type of fault by traditional manual methods. At the same time, there is also a certain degree of danger in the process of manually checking for faults on-site.

[0003] Therefore, providing a method for identifying equipment faults based on an evidence neural network, which can automatically identify equipment faults, is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, the present invention provides a method for identifying equipment faults based on an evidence neural network, which can accurately locate and judge the position and type of equipment faults.

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

[0006] A method for identifying equipment faults based on an evidence neural network includes the following steps:

[0007] Data acquisition step: Obtain equipment vibration data based on vibration sensors;

[0008] Data preprocessing step: Preprocess the equipment vibration data, and extract the features in the frequency domain, time domain, and time-frequency domain of the vibration data as the training data set;

[0009] Network construction step: Construct an evidence neural network including an input layer, a hidden layer, and an output layer;

[0010] Network training step: Input the training data set into the evidence neural network, optimize the network parameters, and obtain a trained evidence neural network;

[0011] Fault identification step: Input the vibration data to be detected into the trained evidence neural network to identify equipment faults.

[0012] Optionally, in the network construction step, the input layer of the evidence neural network uses the K-nearest neighbor algorithm to cluster the training data set to obtain different categories, and then obtain the center points of different categories.

[0013] Optionally, in the network construction step, the evidence neural network is divided into an L1 hidden layer and an L2 hidden layer.

[0014] Optionally, the expression of the L1 hidden layer includes n activation functions si , the expression is:

[0015]

[0016] where τ i and σ i are the self - learning parameters of the evidence neural network, i = 1,..., n, and d i represents the Euclidean distance between the sample x and the model center p i , and the expression is: d i = ||x - p i ||.

[0017] Optionally, the L2 hidden layer includes the basic probability distributions BPA m corresponding to n models i , and for each m i in it, the probability distribution corresponding to the subset in it is expressed as:

[0018]

[0019] where {Class i} represents the set of fault classes, and m i ({Class j ) is the basic probability of the determined class being {Class i}, and are the self - learning parameters of the evidence neural network, k = 1,..., m, and m i M+1 represents the total set probability, Ω C = {Class 1 , Class 2 ,... Class M}, representing the total set, and m i (Ω C ) is the basic probability of the total set Ω C .

[0020] Optionally, in the network training step, optimizing the network parameters includes initializing the adjustable parameters of the model and then performing iterative optimization.

[0021] As can be seen from the above technical solutions, compared with the prior art, the present invention provides a device fault recognition method based on an evidence neural network, which has the following beneficial effects: 1) The present invention adopts a neural network architecture based on the evidence theory, which is not only more comprehensive in information fusion and information representation, but also has the advantages of the neural network in parameter fitting; 2) The present invention has high robustness in the case of incorrect sample labels, effectively improving the learning ability of the depth information in the fault samples in the case of small samples; 3) The present invention can accurately locate and judge the position and type of the device fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0023] Figure 1 Flowchart of a device fault recognition method based on an evidence neural network disclosed by the present invention;

[0024] Figure 2 Principle diagram of the evidence neural network disclosed by the present invention;

[0025] Figure 3 Structure diagram of the evidence neural network disclosed by the present invention;

[0026] Figure 4 Parameter optimization flowchart of the evidence neural network disclosed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] In this application, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0029] Referring to Figure 1 as shown, the present invention discloses a device fault identification method based on an evidence neural network, comprising the following steps:

[0030] Data acquisition step: obtaining device vibration data based on vibration sensors;

[0031] Data preprocessing step: preprocessing the device vibration data and extracting features in the frequency domain, time domain and time-frequency domain of the vibration data as a training data set;

[0032] Network construction step: constructing an evidence neural network including an input layer, a hidden layer and an output layer;

[0033] Network training step: inputting the training data set into the evidence neural network, optimizing network parameters, and obtaining a trained evidence neural network;

[0034] Fault identification step: inputting the vibration data to be detected into the trained evidence neural network to identify device faults.

[0035] Further, in the data acquisition step, vibration sensors are arranged at different positions of the device to obtain device vibration data.

[0036] Further, referring to Figure 2 as shown, the principle of the evidence neural network is to automatically cluster multiple model center points based on the training data obtained by vibration sensors using the K-nearest neighbor algorithm, realizing the process of model modeling. Then, the Euclidean distance between the sample and the model center is calculated, and each fault category {Class i} and the universal set Ω C = {Class 1 , Class 2 ,... Class M} basic probability distribution BPA, then, the conjunctive rule is used to fuse the BPAs of all subsets, and finally, the category corresponding to the largest value in the fused BPA is used as the final predicted fault classification result.

[0037] Further, referring to Figure 3 As shown, in the network construction step, the input layer of the evidential neural network uses the K-nearest neighbor algorithm to cluster the training data set to obtain the model center point. The input layer is the training sample x, which includes q different features.

[0038] Further, the evidential neural network in the network construction step is divided into an L1 hidden layer and an L2 hidden layer.

[0039] Further, the expression of the L1 hidden layer includes n activation functions s i , and the expression is:

[0040]

[0041] Among them, τ i and σ i are the self-learning parameters of the evidential neural network, i = 1,..., n, d i represents the Euclidean distance between the sample x and the model center p i , and the expression is: d i = ||x - p i ||.

[0042] Further, the L2 hidden layer includes the basic probability distribution BPA m corresponding to n models i , and the probability distribution corresponding to each subset in each m i is expressed as:

[0043]

[0044] Among them, {Class i} represents the set of fault categories, and m i ({Class j}) is the basic probability that the determined category is {Class i}, and are the self-learning parameters of the evidential neural network, k = 1,..., m, m i M+1 represents the total probability of the universal set, Ω C = {Class 1 , Class 2 ,... Class M}, represents the universal set, and m i (Ω C ) is the universal set Ω CThe basic probability.

[0045] Furthermore, the output layer contains n nodes θ i , and the final output result is θ n , and each node θ in the output layer k has the following expression:

[0046]

[0047] where θ 1 = m 1 . Therefore, the BPA result given by the final output layer is

[0048] Furthermore, referring to Figure 4 as shown, optimizing the network parameters in the network training step includes iterative optimization after initializing the tunable parameters of the network.

[0049] Specifically, initializing the tunable parameters of the network means initializing the four tunable parameters p i , τ i σ i in the network, and the iterative optimization needs to be based on the loss function Loss as the objective function.

[0050] The definition of the loss function Loss is as follows:

[0051]

[0052] where λ D and λ M are given parameters, is the divergence, m T (X) is the actual probability corresponding to X, m P (X) is the predicted probability corresponding to X, is the Deng entropy, |X| is the cardinality of X, m(X) is the predicted probability corresponding to X, and H max = log(3 |Ω| - 2 |Ω| ) is the maximum value of the Deng entropy, and |Ω| is the cardinality of the universal set.

[0053] Specifically, in order to obtain the optimal network parameters, after substituting the loss function Loss into the entire evidence neural network, the partial derivatives of the four tunable parameters are calculated to obtain Finally, under the condition of determining the maximum value Er s of the loss function and the maximum number of iterations t s , after iterating multiple times, the optimal network parameters are obtained, and the optimal network parameters are input into the evidence network to obtain the trained evidence network, and then the sample to be detected is identified.

[0054] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for a system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and reference can be made to the corresponding part of the method embodiment for relevant content. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0055] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for equipment fault identification based on evidence neural network, characterized in that: The following steps are involved: Data acquisition step: obtaining equipment vibration data based on the vibration sensor; Data preprocessing step: preprocess the equipment vibration data and extract the characteristics of the vibration data in frequency domain, time domain and time-frequency domain as training data sets; Network construction steps: construct an evidence neural network including input layer, hidden layer and output layer; Network training steps: input the training data set into the evidence neural network, optimize the network parameters, and obtain the trained evidence neural network; Fault identification step: Input the vibration data to be detected into the trained evidence neural network to identify equipment faults.

2. The method for equipment fault identification based on evidential neural network according to claim 1 is characterized in that: In the network construction step, the input layer of the evidence neural network uses the K nearest neighbor algorithm to cluster the training data set to obtain different categories, and then obtain the center points of different categories.

3. The method for equipment fault identification based on evidential neural network according to claim 1 is characterized in that: In the network construction step, the evidence neural network is divided into L1 hidden layer and L2 hidden layer.

4. The method for identifying equipment faults based on an evidential neural network according to claim 3 is characterized in that: The L1 hidden layer expression includes n activation functions s i , the expression is: Among them, τ i and σ i is the self-learning parameter of the evidence neural network, d i Represents the sample x and the model center p i The Euclidean distance between them is expressed as: d i =||xp i ||.

5. The method for equipment fault identification based on evidential neural network according to claim 3 is characterized in that: The L2 hidden layer includes the basic probability distribution BPA m corresponding to n models i , where each m i The probability distribution expression corresponding to the subset in is: Among them, {Class i } represents the fault category set, m i ({Class j }) to determine the category as {Class i The basic probability of and is the self-learning parameter of the evidence neural network, k=1,...,m,m i M+1 Expressed as the full set probability, Ω C ={Class1,Class2,...Class M }, represents the full set, m i (Ω C ) is the complete set Ω C The basic probability of .

6. The method for equipment fault identification based on evidential neural network according to claim 1 is characterized in that: Optimizing network parameters in the network training step involves iterative optimization after initializing the model's adjustable parameters.

Images