Rotary machinery fault diagnosis method based on improved OCSSA-VMD-CNN-BILSTM

Abstract

The invention provides a rotating machine fault diagnosis method based on OCSSA-VMD-CNN-BILSTM. The rotating machine fault diagnosis method comprises the following steps: inputting a rotating machine vibration signal; obtaining a range of a vmd decomposition layer number K value through a center frequency method, and setting a range of a penalty factor alpha; and taking the minimum sample entropy as a fitness function, and performing parameter optimization on the VMD by using an improved sparrow optimization algorithm OCSSA to obtain a [K, alpha] optimal combination. According to the method, the bearing fault diagnosis model is built based on the CNN-BILSTM model, the multi-strategy fusion improved sparrow search algorithm (OCSSA) is utilized to optimize the VMD parameters, and when the model is subjected to hyper-parameter optimization, the OCSSA has extremely high convergence speed and extremely high test precision. According to the method, the optimal parameters of the vmd can be obtained without priori knowledge, the algorithm efficiency is improved, manual parameter screening is avoided, higher diagnosis precision and stability are achieved, and the generalization and intelligent level of a diagnosis model are greatly enhanced.

Description

Technical Field

[0001] The invention relates to the field of mechanical fault diagnosis, and in particular to a rotating machinery fault diagnosis method based on an improved OCSSA-VMD-CNN-BILSTM. Background Art

[0002] During the operation of mechanical equipment, rolling bearings are responsible for the spatial fixation of the rotor. They must not only withstand the influence of alternating stress, but also bear most of the static and dynamic loads generated by the mechanism. In addition, the working environment faced by rolling bearings is often very harsh, including high speed, high temperature, high pressure, heavy load and speed change. When mechanical equipment runs for a long time in such a working environment, the failure rate of rolling bearings is much higher than that of other components. It causes huge economic losses to producers and society. Therefore, it is of great practical significance to propose a feasible rolling bearing fault diagnosis method.

[0003] Many scholars have conducted extensive research on signal denoising. Among them, the variational mode decomposition method (VMD) is widely used in mechanical fault diagnosis because it can effectively eliminate modal aliasing and can better avoid endpoint effects. However, the effect of the VMD method is not stable. Many parameters set manually based on subjective factors will affect the final effect. Among them, the more important parameters are the decomposition layer K and the penalty factor α. These two parameters affect the decomposition effect of VMD most significantly. When the K value is too small, the "under-decomposition" phenomenon will occur. If the α value is too small, the modal aliasing phenomenon will occur. Vice versa, when the K value is too large, the "over-decomposition" phenomenon will occur. If the α value is too large, the false component phenomenon will appear.

[0004] In order to reduce noise interference, the current mainstream fault diagnosis method uses intelligent optimization algorithms to optimize VMD parameters combined with various IMF component screening criteria. However, when using the optimization algorithm, there are still problems such as unreasonable parameter range settings, incorrect screening of IMF components, removal of some useful signals, and excessive noise reduction. Therefore, setting appropriate IMF component parameter ranges and screening criteria when optimizing VMD parameters is an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide an improved OCSSA-VMD-CNN-BILSTM method for rotating machinery fault diagnosis, so as to improve the problem that early weak defects of rotating machinery are easily concealed and early defect diagnosis is difficult in a noisy environment, thereby providing a scientific basis for early fault diagnosis of rotating machinery in a strong noise environment and providing reliable support for protecting people's lives and property safety.

[0006] To achieve the above object, the present invention proposes the following technical solution: an improved OCSSA-VMD-CNN-BILSTM rotating machinery fault diagnosis method, comprising the following steps:

[0007] Step 1, inputting a rotating machinery vibration signal;

[0008] Step 2: Get the range of the vmd decomposition layer K value by the center frequency method, and set the range of the penalty factor α;

[0009] Step 3, using the minimum sample entropy as the fitness function, the improved sparrow optimization algorithm OCSSA is used to optimize the parameters of VMD to obtain the optimal combination of [K, α];

[0010] Among them, the improved sparrow optimization algorithm OCSSA includes defining sample entropy and improvement strategy, and the sample entropy is defined as: Among them B m (r) is the probability that two sequences match m points under similarity tolerance r, and A m (r) is the probability that two sequences match m+1 points;

[0011] The improvement strategy includes adopting the improved Tent chaotic map, initializing the diversity of the population, and introducing a random variable rand(0, 1) / N on the original Tent chaotic map expression to solve the problem that the Tent chaotic sequence falls into small periodic points and unstable periodic points during iteration. The improved Tent chaotic map expression is as follows:

[0012]

[0013] Where i represents the number of mappings, Xi represents the function value of the i-th mapping. rand(0, 1) represents a random number;

[0014] The global exploration strategy of the Osprey optimization algorithm in the first stage is used to replace the explorer position update formula of the original sparrow algorithm. The expression is as follows:

[0015]

[0016] Where: 1, is the new position of the i-th osprey in the first stage, and X i,j For individuals; SF i，j is a random number between [0, 1], I i,j is a random number in the set {1, 2};

[0017] The Cauchy mutation strategy is used to replace the follower position update formula of the original sparrow algorithm. The expression is as follows:

[0018]

[0019] Among them, cauchy(0,1) is the standard Cauchy distribution. ⊕ means multiplication.

[0020] Step 4: Based on the optimal combination of [K, α], perform VMD signal decomposition on the vibration signal to obtain several IMF components;

[0021] Step 5: Filter out the IMF components with the most fault characteristic signal components according to the correlation coefficient and kurtosis index of each IMF component, and then reconstruct the IMF components;

[0022] Step 6: Input the filtered data features into the CNN-BILSTM network for fault diagnosis.

[0023] Furthermore, in the present invention, the specific operations in step 2 are as follows: first, the penalty factor α is set to a default value of 1000, and VMD decomposition is performed when K is a positive integer from 2 to 10, to obtain IMF components and calculate each center frequency; secondly, the center frequency error between the last IMF components is calculated, and the K value in which the center frequency error finally appears continuously within a range of less than 0.01 is determined as the optimal K value range; finally, the search range setting of the penalty factor α is determined.

[0024] Furthermore, in the present invention, the center frequency formula of the K modes is:

[0025]

[0026] Where P(f) is the power spectral density function and f is the frequency variable.

[0027] Furthermore, in the present invention, in step 3, sparrow individuals are randomly selected according to the set ratio of alerters, and the alerter positions are updated, and the formula is as follows:

[0028]

[0029] in, represents the position of the best discoverer in the n+1 generation; x” represents the position of the global worst individual in the current n generation; A + Represents a 1xDim matrix, each cell in the matrix is ​​randomly assigned -1 and 1, and A + =A(AA T ) -1 ;

[0030] According to the greedy rule, the optimal fitness value and the optimal position are updated; if the maximum number of iterations is not reached, the iteration is returned and re-expanded; if the maximum number of iterations is reached, the optimal weight and the optimal threshold are output.

[0031] Furthermore, in the present invention, in step 4, nine time domain indicators of the optimal IMF component are calculated, including mean, variance, peak, kurtosis, effective value, peak factor, pulse factor, waveform factor, margin factor and constructed eigenvector.

[0032] Beneficial effects: The technical solution of this application has the following technical effects:

[0033] 1. The present invention innovatively uses the improved Tent mapping to initialize the population, replaces the explorer position update formula of the original sparrow algorithm with the global exploration strategy of the Osprey optimization algorithm in the first stage, and replaces the follower position update formula of the original sparrow algorithm with the Cauchy mutation strategy. Combining multiple strategies, a fusion improved sparrow search algorithm is proposed. The present invention uses OCSSA to optimize VMD parameters. When optimizing model hyperparameters, OCSSA has extremely fast convergence speed and extremely high test accuracy. This method can obtain the optimal parameters of vmd without prior knowledge, improves algorithm efficiency, avoids manual parameter screening, and has higher diagnostic accuracy and stability, greatly enhancing the generalization and intelligence level of the diagnostic model.

[0034] 2. The present invention builds a bearing fault diagnosis model based on the CNN-BILSTM model, and optimizes the VMD parameters by using the multi-strategy fusion improved sparrow search algorithm (Osprey-Cauchy-Sparrow Search Algorithm, OCSSA), avoiding the subjective error introduced by manual screening parameters and having higher diagnostic accuracy and stability, and further optimizing the hyperparameters of the model. Among them, the present invention aims at the uneven population distribution and poor population diversity of the sparrow search algorithm at the time of initialization; the position update method of the discoverer has certain limitations, and it is impossible to balance the global search ability and the local mining ability, and it is easy to have the problem of suboptimal solution. And when the warning value is less than the set safety threshold, the search ability of the discoverer will slowly decrease with the increase of the number of iterations of the algorithm, resulting in insufficient spatial search area of ​​the algorithm and low convergence efficiency of the algorithm; the worst and best sparrow individuals lack communication feedback and are prone to fall into the problem of local optimal solution.

[0035] It should be appreciated that all combinations of the foregoing concepts, as well as additional concepts described in greater detail below, may be considered to be part of the inventive subject matter of the present disclosure, provided such concepts are not mutually inconsistent.

[0036] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in various figures may be represented by the same reference numeral. For clarity, not every component is labeled in every figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings, in which:

[0038] Figure 1 It is a flow chart of the bearing fault diagnosis method based on OCSSA-VMD-CNN-BILSTM of the present invention;

[0039] Figure 2 This is the overall algorithm flow chart of OCSSA;

[0040] Figure 3 It is the network structure information of CNN-BILSTM.

[0041] Figure 4 This is the test set result diagram of the example OCSSA-VMD-CNN-BILSTM

[0042] Figure 5 This is the confusion matrix diagram of the OCSSA-VMD-CNN-BILSTM example. DETAILED DESCRIPTION

[0043] In order to better understand the technical content of the present invention, specific embodiments are cited and described as follows in conjunction with the accompanying drawings. Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily defined to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed in the present invention are not limited to any implementation. In addition, some aspects disclosed in the present invention can be used alone, or in any appropriate combination with other aspects disclosed in the present invention.

[0044] Example 1

[0045] A bearing fault diagnosis method based on OCSSA-VMD-CNN-BILSTM includes the following processes:

[0046] First, the bearing vibration signal data of a machine tool factory in Nanjing is obtained. The sampling frequency is 12k, the sampling interval is 1min, and each sampling is 1.8s. A set of normal data and three sets of fault data (inner ring, ball, outer ring) are collected, and they are divided into training set and test set in a ratio of 7:3.

[0047] The range of the decomposition layer K value obtained by the center frequency method is 2 to 10, and the range of the penalty factor α is set to (200, 1000). The vibration signal is decomposed by VMD in turn to obtain the center frequency value of each IMF component, and finally the K value with a center frequency error within 0.01 is selected;

[0048] Taking the minimum sample entropy as the fitness function, the sample entropy is defined as:

[0049]

[0050] Among them B m (t) is the probability that two sequences match m points under similarity tolerance r, and A m (r) is the probability that two sequences match m+1 points;

[0051] Initialize the parameters of the sparrow optimization algorithm, including the number of sparrow populations, the maximum number of iterations, the ratio of discoverers to alerters, use the improved Tent mapping to initialize the coordinates of the sparrow population, and calculate the optimal fitness and the corresponding sparrow individuals;

[0052] The improved sparrow optimization algorithm has the following steps: initializing the parameters of the sparrow optimization algorithm, including the number of sparrow populations, the maximum number of iterations, the ratio of discoverers to alerters, initializing the coordinates of the sparrow population using the improved Tent mapping, and calculating the optimal fitness and the corresponding sparrow individuals. The formula is as follows:

[0053]

[0054] According to the set proportion of discoverers, sparrow individuals are selected, and the global exploration formula of the osprey optimization algorithm in the first stage is used as the discoverer's position update formula. This method can correct the sparrow algorithm's excessive reliance on the update method of the previous generation of sparrow positions, and compensate for this shortcoming by randomly detecting and attacking the position of one of the foods. The formula is as follows:

[0055]

[0056] Where: 1, is the new position of the i-th osprey in the first stage, and x i，j For individuals; SF i,j is a random number between [0,1], I i,j is a random number from the set {1,2}.

[0057] All remaining individuals are regarded as followers, and the Cauchy mutation strategy is used as the update formula of the follower position. Similar to the standard normal distribution, the Cauchy distribution is a continuous probability distribution with a small value at the origin, a relatively flat and long end, and a slow approach to zero rate, so it can produce greater disturbances than the normal distribution. Therefore, using Cauchy mutation to perturb individuals in the sparrow position update can expand the search scale of the sparrow algorithm, thereby improving the algorithm's ability to jump out of the local optimum. The formula is as follows:

[0058]

[0059] Randomly select sparrow individuals according to the set proportion of sentinels and update the sentinel positions. The formula is as follows;

[0060]

[0061] in, represents the position of the best discoverer in the n+1 generation; x” represents the position of the global worst individual in the current n generation; A + Represents a 1xDim matrix, each cell in the matrix is ​​randomly assigned -1 and 1, and A + =A(AA T ) -1 .

[0062] According to the greedy rule, update the best fitness value and the best position; if the maximum number of iterations is not reached, return to re-expand the iteration; if the maximum number of iterations is reached, output the optimal weight and optimal threshold;

[0063] According to the optimal weight and optimal threshold, i.e. the optimal combination of [K, α], the optimal K and α values ​​are back-substituted to extract the optimal IMF component corresponding to the minimum sample entropy;

[0064] Extract the feature vector. Calculate the 9 time domain indicators of this optimal IMF component, including: mean, variance, peak, kurtosis, effective value, peak factor, pulse factor, waveform factor, margin factor, and construct the feature vector. Then use these 9 indicators to construct the feature vector of each sample. After extracting the features, each state has 120 samples, a total of ten states, and a 1200*9 column matrix is ​​obtained, and then each row of data is labeled. 1-10 represents different fault types. Divide the first 90 groups of each state into training sets and the last 30 groups into test sets.

[0065] The selected data features are input into the CNN-BILSTM network for fault diagnosis. First, CNN is used to extract time-frequency domain features and fuse them, and then the BiLSTM network is used to learn sequence features to complete the multi-level classification task. The number of training times is set to 150, the learning rate is 0.01, and the regularization parameter is 0.001. After 100 training times, the learning rate is adjusted, and the learning rate adjustment factor is 0.01.

[0066] The VMD used in the present invention is an adaptive, completely non-recursive modal variation and signal processing method. It uses iterative search for the optimal solution of the variational model to determine the center frequency and bandwidth of each component, and is a completely non-recursive model. The model searches for a set of modal components and their respective center frequencies, and each mode is smooth after demodulation to baseband, has strong robustness, and has a strong ability to avoid modal aliasing. The author uses the proposed OCSSA optimization algorithm to optimize the parameter optimization of VMD.

[0067] The specific implementation of OCSSA-VMD is as follows: find the optimal penalty factor and modal component of VMD; find the index value of its minimum fitness, substitute the two optimized parameters and the index value of the minimum fitness back into VMD; find the optimal IMF component for each fault type, and use the criterion of minimum sample entropy to extract the optimal IMF component.

[0068] The CNN-BiLSTM model proposed in this paper can capture features from time series, model dependencies, and process variable-length sequences, and has the advantages of reducing dimensions and flexibility. These features enable the model to better understand the patterns and changes of time series and make accurate identification and classification.

[0069] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. An improved OCSSA-VMD-CNN-BILSTM method for rotating machinery fault diagnosis, Features: The following steps are involved: Step 1, inputting a rotating machinery vibration signal; Step 2: Get the range of the number of vmd decomposition layers K by the center frequency method, and set the range of the penalty factor α; Step 3, using the minimum sample entropy as the fitness function, the improved sparrow optimization algorithm OCSSA is used to optimize the parameters of VMD to obtain the optimal combination of [K, α]; Among them, the improved sparrow optimization algorithm OCSSA includes defining sample entropy and improvement strategy, and the sample entropy is defined as: Among them B m (r) is the probability that two sequences match m points under similarity tolerance r, and A m (r) is the probability that two sequences match m+1 points; The improvement strategy includes adopting the improved Tent chaotic map, initializing the diversity of the population, and introducing a random variable rand(0, 1) / N on the original Tent chaotic map expression to solve the problem that the Tent chaotic sequence falls into small periodic points and unstable periodic points during iteration. The improved Tent chaotic map expression is as follows: Where i represents the number of mappings, Xi represents the function value of the i-th mapping. rand(0, 1) represents a random number; The global exploration strategy of the Osprey optimization algorithm in the first stage is used to replace the explorer position update formula of the original sparrow algorithm. The expression is as follows: Where: 1, is the new position of the i-th osprey in the first stage, and x i，j For individuals; SF i，j is a random number between [0, 1], I i，j is a random number in the set {1, 2}; The Cauchy mutation strategy is used to replace the follower position update formula of the original sparrow algorithm. The expression is as follows: Among them, cauchy(0,1) is the standard Cauchy distribution. It means multiplication; Step 4: Based on the optimal combination of [K, α], perform VMD signal decomposition on the vibration signal to obtain several IMF components; Step 5: Filter out the IMF components with the most fault characteristic signal components according to the correlation coefficient and kurtosis index of each IMF component, and then reconstruct the IMF components; Step 6: Input the filtered data features into the CNN-BILSTM network for fault diagnosis.

2. According to the improved OCSSA-VMD-CNN-BILSTM rotating machinery fault diagnosis method of claim 1, Features: The specific operations in step 2 are as follows: first, the penalty factor α is set to a default value of 1000, and VMD decomposition is performed when K is a positive integer from 2 to 10, to obtain IMF components and calculate each center frequency; second, the center frequency error between the last IMF components is calculated, and the K value with a center frequency error within a range of 0.01 that appears continuously is determined as the optimal K value range; finally, the search range setting of the penalty factor α is determined.

3. According to the improved OCSSA-VMD-CNN-BILSTM rotating machinery fault diagnosis method of claim 1, Features: The modal number K formula is: Where P(f) is the power spectral density function, and f is the frequency variable; Among them, fmax and fmin are the highest frequency and the lowest frequency of the signal spectrum, respectively, indicating rounding up.

4. According to the improved OCSSA-VMD-CNN-BILSTM rotating machinery fault diagnosis method of claim 1, Features: In step 3, sparrow individuals are randomly selected according to the set proportion of guards, and the positions of the guards are updated. The formula is as follows; in, represents the position of the best discoverer in the n+1 generation; x” represents the position of the global worst individual in the current n generation; A + Represents a 1xDim matrix, each cell in the matrix is ​​randomly assigned -1 and 1, and A + =A(AA T ) -1 ; According to the greedy rule, the optimal fitness value and the optimal position are updated; if the maximum number of iterations is not reached, the iteration is returned and re-expanded; if the maximum number of iterations is reached, the optimal weight and the optimal threshold are output.

5. According to the improved OCSSA-VMD-CNN-BILSTM rotating machinery fault diagnosis method of claim 1, Features: In step 4, nine time domain indicators of the optimal IMF component are calculated, including mean, variance, peak, kurtosis, effective value, peak factor, pulse factor, waveform factor, margin factor and constructed eigenvector.

Images