Rolling bearing fault diagnosis method and system

Abstract

The application relates to a rolling bearing fault diagnosis method and system. Vibration signals of a rolling bearing to be detected are acquired, a plurality of decomposition algorithms are used for signal decomposition, a plurality of feature vectors under the plurality of decomposition methods are respectively obtained, the plurality of feature vectors are respectively input into a trained correlation vector machine model for fault diagnosis, a plurality of fault diagnosis outputs are obtained, the plurality of fault diagnosis outputs are fused, and a final diagnosis result is determined. The application adopts a plurality of decomposition methods to decompose the vibration signals, and fuses a plurality of fault diagnosis probability outputs obtained, so that fault features can be more effectively extracted, and the accuracy of fault diagnosis is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fault diagnosis, and particularly relates to a rolling bearing fault diagnosis method and system. BACKGROUND

[0002] As one of the commonly used components of mechanical equipment, rolling bearings are widely used in various important departments such as chemical industry, metallurgy, electric power, aviation, etc., and are also one of the most easily damaged components. Therefore, if a rolling bearing fails during the operation of the mechanical equipment, it may cause the entire production process to be interrupted, which not only affects the economic benefits of the enterprise, but also may endanger the personal safety of the workers, and even bring disastrous consequences.

[0003] At present, the vibration signal of the rolling bearing during operation is monitored, and the fault is diagnosed in time so as to take corresponding measures to avoid causing significant losses. Therefore, the rolling bearing fault diagnosis method based on the vibration signal has great significance.

[0004] In the prior art, the conventional fault diagnosis method for rolling bearings mostly only analyzes and extracts the features of the fault information obtained by a single sensor, and then judges whether there is a fault and the type of the fault. However, the biggest disadvantage of this kind of method is that the fault features cannot be effectively extracted, and the accuracy of fault diagnosis is low. SUMMARY

[0005] The purpose of the application is to provide a rolling bearing fault diagnosis method and system with high fault diagnosis accuracy.

[0006] To achieve the above purpose, the application provides the following solutions.

[0007] A rolling bearing fault diagnosis method comprises the following steps:

[0008] Step 1: obtaining the vibration signal of a rolling bearing to be detected;

[0009] Step 2: performing signal decomposition on the vibration signal by using multiple decomposition algorithms to obtain feature vectors under multiple decomposition methods respectively;

[0010] Step 3: obtaining a trained relevant vector machine model;

[0011] Step 4: inputting the multiple feature vectors into the trained relevant vector machine model respectively to perform fault diagnosis and obtain multiple fault diagnosis probability outputs;

[0012] Step 5: fusing the multiple fault diagnosis probability outputs to obtain a final diagnosis result.

[0013] Optionally, the multiple decomposition methods comprise a variational mode decomposition algorithm, an empirical mode decomposition algorithm and a wavelet packet decomposition algorithm.

[0014] Optionally, the step 2 specifically comprises:

[0015] The vibration signal is decomposed by using multiple decomposition algorithms to obtain multiple component signals decomposed by the decomposition methods respectively.

[0016] The fuzzy quotient of each component signal is calculated to obtain multiple feature vectors.

[0017] Optionally, the step 5 specifically comprises:

[0018] The fault diagnosis probability outputs are obtained.

[0019] The trust degree and the false degree between each fault diagnosis probability output are calculated.

[0020] The weight coefficient of each fault diagnosis probability output is determined according to the trust degree and the false degree.

[0021] The fault diagnosis probability outputs are weighted and averaged.

[0022] The weighted and averaged fault diagnosis probability outputs are combined by using the Dempster combination rule to obtain the fusion fault diagnosis result.

[0023] Optionally, the combination of the weighted and averaged fault diagnosis probability outputs by using the Dempster combination rule specifically comprises:

[0024] According to The weighted and averaged fault diagnosis probability outputs are combined, wherein m1(A1) is the fault diagnosis probability function of fault A1; m2(A2) is the fault diagnosis probability function of fault A2; m3(A3) is the fault diagnosis probability function of fault A3. m1, m2, and m3 are the fault diagnosis probability outputs.

[0025] Optionally, before the step 3, the method further comprises:

[0026] The historical vibration signal data and the historical fault type of the rolling bearing are obtained.

[0027] An untrained relevant vector machine model is obtained as an initial relevant vector machine model.

[0028] The initial relevant vector machine model is trained by taking the historical vibration signal data as input and the historical fault type as label to obtain a trained relevant vector machine model.

[0029] Optionally, the initial relevant vector machine model comprises a combination kernel function formula.

[0030] Optionally, the combined kernel function is calculated according to K(x,x')=w×K1(x,x')+(1-w)×K2(x,x'), where K1 is the Gaussian kernel function: K2 is a polynomial kernel function: K2(x,x')=(x T x'+1) d d is the polynomial order; w is the coefficient, and w∈(0,1). δ, d and w are all parameters of the combined kernel function.

[0031] Optionally, the combined kernel function parameters are combined kernel function parameters optimized by an optimization algorithm.

[0032] The present invention also provides a rolling bearing fault diagnosis system, comprising:

[0033] The signal acquisition module is used to acquire the vibration signal of the rolling bearing to be detected.

[0034] The feature calculation module is used to decompose the vibration signal according to multiple decomposition algorithms to obtain feature vectors under the multiple decomposition methods respectively.

[0035] The model acquisition module is used to acquire pre-trained relevant vector machine models;

[0036] The fault probability module is used to input the multiple feature vectors into the trained correlation vector machine model for fault diagnosis and obtain multiple fault diagnosis probability outputs.

[0037] The fault diagnosis module is used to fuse multiple fault diagnosis probability outputs and calculate the final diagnosis result.

[0038] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The present invention proposes a method for diagnosing rolling bearing faults. Based on the vibration signal obtained during the movement of the rolling bearing, the vibration signal is decomposed into different feature vectors through various decomposition methods. The feature vectors are then input into a trained vector machine model to obtain multiple fault diagnosis probabilities. These multiple fault diagnosis probabilities are fused to calculate the final diagnosis result. Because fault diagnosis is performed on the same vibration signal through multiple methods and multiple fault diagnosis probabilities are fused, fault features can be effectively extracted, improving the accuracy of fault diagnosis. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flow chart of a rolling bearing fault diagnosis method of the present application;

[0041] Figure 2 A rolling bearing fault diagnosis method of an embodiment of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0043] The purpose of the present application is to provide a rolling bearing fault diagnosis method and system

[0044] In order to make the above-mentioned purposes, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0045] As Figure 1 shown, the rolling bearing fault diagnosis method provided by the present application comprises the following steps:

[0046] Step 1: obtaining a vibration signal of a rolling bearing to be detected.

[0047] Step 2: using a plurality of decomposition algorithms to decompose the vibration signal to obtain a plurality of feature vectors under the plurality of decomposition methods respectively.

[0048] Step 3: obtaining a trained relevant vector machine model.

[0049] Step 4: inputting the plurality of feature vectors into the trained relevant vector machine model respectively for fault diagnosis to obtain a plurality of fault diagnosis probability outputs.

[0050] Step 5: fusing the plurality of fault diagnosis probability outputs to obtain a final diagnosis result.

[0051] In some embodiments, the plurality of decomposition algorithms can include a variational mode decomposition algorithm, an empirical mode decomposition algorithm and a wavelet packet decomposition algorithm.

[0052] The variational mode decomposition, abbreviated as VMD, is a completely non-recursive and adaptive signal decomposition method, which can effectively suppress the mode aliasing phenomenon and has good noise robustness. The empirical mode decomposition, abbreviated as EMD, is a time-frequency signal analysis method for signal processing, which has significant advantages in nonlinear and non-stationary signal processing. The wavelet packet decomposition can provide a more detailed analysis method for signals.

[0053] In some implementations, step 2 may specifically include:

[0054] The vibration signal is decomposed using multiple decomposition algorithms to obtain component signals after decomposition by multiple decomposition methods;

[0055] The fuzzy quotient of each component signal is calculated to obtain multiple feature vectors.

[0056] like Figure 2 The rolling bearing fault diagnosis method shown takes the use of variational mode decomposition algorithm to decompose the vibration signal of the rolling bearing under test and obtain the feature vector under the variational mode decomposition algorithm as an example. Specifically:

[0057] The vibration signal of the rolling bearing to be tested is decomposed using the variational mode decomposition algorithm to obtain the component signal f after decomposition by the variational mode decomposition algorithm. The component signal f contains multiple signal components f(n).

[0058] The fuzzy quotient of the component signal f is calculated to obtain the feature vector F. The steps for calculating the fuzzy quotient of the component signal f are as follows:

[0059] 1) Assuming the length of the component signal f is N, and the component signal f is represented as X = [x(1), x(2), ..., x(N)], construct a set of m-dimensional vectors from the original signal sequence data:

[0060] X m (i)=[x(i),x(i+1),...,x(i+m-1)]-u(i)

[0061] Where i = 1, 2, ..., N-m+1,

[0062] 2) Calculate vector X m (i) and X m The distance of (j) will be the distance to vector X. m (i) and X m The maximum value of the absolute value of the element difference corresponding to (j) is defined as in The calculation formula is as follows:

[0063]

[0064] Where i = 1, 2, ..., N-m+1, j = 1, 2, ..., N-m+1 and i ≠ j.

[0065] 3) Introduce fuzzy membership function to represent vector X m (i) and Xm (j) similarity:

[0066]

[0067] where r is a similarity tolerance parameter.

[0068] 4) Definition The function, the specific formula can be as follows:

[0069]

[0070]

[0071] 5) Increase the dimension, make m=m+1, repeat 1)-4), get The specific formula can be as follows:

[0072]

[0073] 6) The fuzzy quotient of the signal sequence is defined as follows:

[0074]

[0075] 7) According to the fuzzy quotient definition formula, the eigenvector of the component signal is calculated.

[0076] In some embodiments, before step 3, further comprising:

[0077] Obtain historical vibration signal data and historical fault types of the rolling bearing.

[0078] Obtain an untrained relevant vector machine model as an initial relevant vector machine model.

[0079] Train the initial relevant vector machine model with the historical vibration signal data as input and the historical fault types as labels to obtain a trained relevant vector machine model.

[0080] The above relevant vector machine model includes a combined kernel function, and the formula of the constructed combined kernel function can be as follows:

[0081] K(x,x')=w×K1(x,x')+(1-w)×K2(x,x')

[0082] Where K1 is a Gaussian kernel function: K2 is a polynomial kernel function: K2(x,x')=(x T x'+1) d , d is the polynomial order; w is a coefficient, and w∈(0,1).

[0083] In the combined kernel function formula, the value of the combined kernel function parameter will affect the fault diagnosis accuracy of the relevance vector machine. In order to make the fault diagnosis accuracy of the relevance vector machine optimal, the optimal value of the combined kernel function parameter is obtained.

[0084] Taking the need to optimize three combined kernel function parameters δ, d and w as an example, the sparrow search algorithm (SSA) is used to optimize and select the combined kernel function parameters, and the specific optimization steps are as follows:

[0085] 1) Set the sparrow search algorithm related parameters: population size M = 20, maximum iteration number T = 200, problem dimension dim value is related to the number of combined kernel function parameters to be optimized, here dim = 3, and the fitness function is the fault diagnosis error rate of mRVM.

[0086] 2) Initialize the sparrow population, randomly generate the initial position X of each sparrow, here X is a 3-dimensional vector, and each dimension of the vector element corresponds to a combined kernel function parameter.

[0087] 3) Calculate the fitness value of each sparrow in the initial population, and sort the fitness values from small to large, and mark the sparrow with the highest fitness value as the discoverer. The discoverer usually has a higher energy reserve and is responsible for searching the area with rich food in the whole population, and provides the area and direction for the joiners to forage. The joiners will always follow the discoverer in order to get better food. However, the identities of the discoverer and the joiner are dynamic, and as long as better food sources are found, each sparrow can become a discoverer. However, the proportion of the number of discoverers and joiners in the whole population is constant, that is, when one sparrow becomes a discoverer, another sparrow becomes a joiner.

[0088] 4) Update the position of the discoverer according to the formula , wherein t is the current iteration number; j = 1, 2, … d; is the value of the jth dimension of the ith sparrow in the tth iteration; T is the maximum iteration number; d is the polynomial order; α is a random number, and the value range is a ∈ (0, 1]; R2 is the alarm value, and the value range is R2 ∈ [0, 1]; ST is the safety threshold, and the value range is ST ∈ [0.5, 1.0]; Q is a random number of standard normal distribution.

[0089] 5) Update the joiner according to the formula , wherein X P is the optimal position of the current discoverer; X worst is the current global worst position; A is a 1 × d matrix, each element of which is randomly-1, and A + = A T (AA T ) -1.

[0090] 6) According to the formula Update the position X of the sparrow aware of the danger best is the current global optimal position; β is a step control parameter, and the value of β is a random number obeying a normal distribution with a mean of 0 and a variance of 1; H is a random number, and H ∈ [-1, 1]; f i is the fitness value of the current sparrow individual; f g is the fitness value of the current global optimal; f w is the fitness value of the current global worst; and ε is the minimum constant.

[0091] 7) Obtain the current optimal value.

[0092] 8) If the iteration does not reach the set maximum number of iterations, return to step 3 to continue iteration, otherwise, output the global optimal position and the optimal fitness value.

[0093] Determine the parameters of the combined kernel function of the relevant vector machine based on the output global optimal position, to obtain the combined kernel function parameters optimized by the optimization algorithm.

[0094] As Figure 2 shown, in some embodiments, step 4 specifically includes:

[0095] The feature vectors F1, F2, and F3 are respectively input into the trained relevant vector machine model with the combined kernel parameters of the mRVM optimized by the SSA, to obtain the mRVM fault diagnosis outputs m1, m2, and m3, respectively.

[0096] In some embodiments, step 5 specifically includes:

[0097] Obtain the fault diagnosis probability outputs.

[0098] Calculate the trust degree and the false degree between each of the fault diagnosis probability outputs.

[0099] Determine the weight coefficient of each of the fault diagnosis probability outputs according to the trust degree and the false degree.

[0100] Weighted average the fault diagnosis probability outputs.

[0101] Combine the weighted average fault diagnosis probability outputs by using the Dempster combination rule to obtain the fusion fault diagnosis result.

[0102] The trust degree between each of the fault diagnosis probability outputs is calculated, and the calculation formula is specifically as follows:

[0103]

[0104] wherein

[0105] sim(m1, m2) is calculated according to the generalized Jaccard coefficient between the fault diagnosis probability outputs, and the calculation formula is as follows:

[0106] The false degree between each fault diagnosis probability output is calculated, and the calculation formula is as follows:

[0107] wherein, K0 is a global conflict system, K i is a conflict coefficient between the remaining fault diagnosis probability outputs except m i The fault diagnosis probability outputs are divided into m1, m2, m3, and thus the false degree vector of the fault diagnosis probability outputs is F = [F(m1), F(m2), F(m3)] T .

[0108] wherein, the weight coefficient of each fault diagnosis probability output is determined according to the trust degree and the false degree, and the calculation formula is as follows:

[0109] w i = Crd i +1-F(m i )

[0110] wherein, the fault diagnosis probability outputs are weighted and averaged, and the specific steps are as follows:

[0111] According to the weight coefficients are normalized, the basic probability values of all evidences are weighted according to the normalized weight coefficients, and then the average value is calculated.

[0112] Specifically, the weighted and averaged fault diagnosis probability outputs are combined by using the Dempster combination rule to obtain the fusion fault diagnosis result, which specifically includes:

[0113] The weighted and averaged BPA is combined n-1 times by using the Dempster combination rule;

[0114] The Dempster combination is used for the three fault diagnosis probability outputs m1, m2 and m3, and the combination formula is as follows: wherein, m1(A1) is the fault diagnosis probability function of fault A1; m2(A2) is the fault diagnosis probability function of fault A2; m3(A3) is the fault diagnosis probability function of fault A3; ; m1, m2, m3 are the fault diagnosis probability outputs m1(A1), m2(A2), m3(A3) respectively.

[0115] The application further provides a rolling bearing fault diagnosis system, and specifically comprises:

[0116] A signal acquisition module is configured to acquire a rolling bearing vibration signal to be detected.

[0117] A feature calculation module is configured to decompose the vibration signal according to multiple decomposition algorithms to obtain feature vectors under multiple decomposition methods respectively.

[0118] A model acquisition module is configured to acquire a trained relevant vector machine model.

[0119] A fault probability module is configured to input the multiple feature vectors into the trained relevant vector machine model to perform fault diagnosis and obtain multiple fault diagnosis probability outputs.

[0120] A fault diagnosis module is configured to fuse the multiple fault diagnosis probability outputs to obtain a final diagnosis result.

[0121] In summary, the application has the following advantages: multiple decomposition algorithms are adopted to decompose the vibration signal of the rolling bearing to obtain feature vectors under corresponding decomposition methods respectively, each feature vector is input into a relevant vector machine model, multiple fault diagnosis probability outputs are fused to obtain a final diagnosis result. Since multiple decomposition algorithms are adopted, the same vibration signal can be subjected to multiple fault diagnoses, compared with the technical means of using a single decomposition algorithm in the traditional diagnosis method, the application can more effectively extract fault features and perform fault diagnosis. Moreover, the relevant vector machine model is trained by using historical data samples of various bearing states, and a combined kernel function formula is constructed in the relevant vector machine model, which can solve the problems of weak generalization ability and low fault diagnosis accuracy of a single kernel function.

[0122] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0123] The principles and implementation manners of the application are described by using specific examples in this paper, and the above embodiment description is only used to help understand the method and core idea of the application; meanwhile, for those skilled in the art, according to the idea of the application, the specific implementation manner and application range can be changed. In summary, the content of the specification should not be understood as a limitation of the application.

Claims

1. A method for diagnosing rolling bearing faults, characterized in that, include: Step 1: Acquire the vibration signal of the rolling bearing to be tested; Step 2: The vibration signal is decomposed using multiple decomposition algorithms to obtain feature vectors under each algorithm; the multiple decomposition algorithms include variational mode decomposition algorithm, empirical mode decomposition algorithm, and wavelet packet decomposition algorithm. Step 2 specifically includes: The vibration signal is decomposed using multiple decomposition algorithms to obtain component signals after decomposition by the multiple decomposition algorithms. Calculate the fuzzy quotient of each component signal to obtain multiple feature vectors; Step 3: Obtain the trained relevance vector machine model; Step 4: Input the multiple feature vectors into the trained correlation vector machine model to perform fault diagnosis and obtain multiple fault diagnosis probability outputs; Step 5: Merge multiple fault diagnosis probability outputs to calculate the final diagnosis result; Step 5 specifically includes: Obtain the fault diagnosis probability output; Calculate the confidence level and false positive level among the various fault diagnosis probability outputs; The weighting coefficients for each of the fault diagnosis probability outputs are determined based on the trust level and the falsehood level. The fault diagnosis probability output is weighted and averaged. The Dempster synthesis rule is used to combine the weighted average fault diagnosis probability outputs to obtain the fused fault diagnosis result. The combination of the weighted averaged fault diagnosis probability outputs using the Dempster synthesis rule specifically includes: according to The weighted average fault diagnosis probability outputs are combined, where m1(A1) is the fault diagnosis probability function of fault A1; m2(A2) is the fault diagnosis probability function of fault A2; and m3(A3) is the fault diagnosis probability function of fault A3. m1, m2, and m3 are the fault diagnosis probability outputs of m1(A1), m2(A2), and m3(A3), respectively. Before step 3, the following is also included: Acquire historical vibration signal data and historical fault types of rolling bearings; Obtain an untrained relevance vector machine model as the initial relevance vector machine model; Using the historical vibration signal data as input and the historical fault type as label, the initial correlation vector machine model is trained to obtain the trained correlation vector machine model. The initial correlation vector machine model includes a combined kernel function formula; The combined kernel function is calculated according to K(x,x')=w×K1(x,x')+(1-w)×K2(x,x'), where K1 is the Gaussian kernel function: K2 is a polynomial kernel function: K2(x,x')=(x T x'+1) d d is the order of the polynomial; w is the coefficient, and w∈(0,1). δ, d and w are all parameters of the combined kernel function.

2. The rolling bearing fault diagnosis method according to claim 1, characterized in that, The combined kernel function parameters are the combined kernel function parameters optimized by the optimization algorithm.

3. A rolling bearing fault diagnosis system for implementing the rolling bearing fault diagnosis method according to any one of claims 1-2, characterized in that, include: The signal acquisition module is used to acquire the vibration signal of the rolling bearing to be detected. The feature calculation module is used to decompose the vibration signal according to multiple decomposition algorithms, and obtain feature vectors under the multiple decomposition algorithms respectively. The model acquisition module is used to acquire pre-trained relevant vector machine models; The fault probability module is used to input the multiple feature vectors into the trained correlation vector machine model for fault diagnosis and obtain multiple fault diagnosis probability outputs. The fault diagnosis module is used to fuse multiple fault diagnosis probability outputs and calculate the final diagnosis result.

Images