Rotor circle measurement signal noise reduction method based on improved ASCSSA-SVMD-PE

Abstract

The invention provides a rotor circle measurement signal noise reduction method based on improved ASCSSA-SVMD-PE, and the method comprises the steps: carrying out the signal decomposition through employing improved continuous variational mode decomposition (SVMD) according to the effect that VMD is controlled by two hyper-parameters; an adaptive spiral Cauchy sparrow search algorithm is obtained by improving an optimization step length, a search strategy and the like of a sparrow algorithm, and SVMD decomposition parameters are intelligently optimized according to the algorithm to obtain an optimal penalty factor value; the SVMD signal decomposition method optimized based on the ASCSSA algorithm is applied to the noise reduction field in the rotor roundness measurement signal processing process, and the maximum mutual information coefficient is adopted as a target constraint function of SVMD optimized by the ASCSSA algorithm; according to the rotor roundness measurement signal noise reduction method, a PE method is adopted to carry out entropy arrangement on IMF components decomposed by SVMD, only the minimum entropy component is reserved as a reconstruction signal, a final noise reduction signal is obtained, the defects and the technical problems existing in an existing rotor roundness measurement signal noise reduction method are effectively solved through combined use of the methods, and the practicability and the reliability of a noise reduction algorithm are enhanced.

Description

Technical Field

[0001] The invention belongs to the field of rotor measurement of hydropower units, and in particular relates to a rotor circularity measurement signal noise reduction method based on improved ASCSA-SVMD-PE. Background Art

[0002] Rotor roundness measurement is an important step in the installation and maintenance of hydropower units in the field of water conservancy and hydropower. The purpose of rotor roundness measurement is to check the roundness of the magnetic poles through laser sensors to ensure that each magnetic pole is installed correctly. There is insulating paint on the surface of the rotor, and its thickness is inconsistent. In the process of measuring the signal, the sensor is affected by the jitter of the cantilever of the roundness measurement frame, resulting in inaccurate roundness measurement signals, affecting the accuracy of continuous measurement, and may cause deviations in the installation of the magnetic poles. If there is a deviation in the installation of the magnetic poles, the rotor roundness will not meet the construction requirements, and a relatively obvious air gap unevenness will occur during the operation of the hydropower unit. Therefore, the data processing problem of the rotor roundness measurement signal needs to be solved urgently.

[0003] The processing of circular measurement data has been widely studied, among which the multi-component adaptive decomposition method - variational mode decomposition (VMD) is widely used in the field of signal denoising. Since VMD itself does not have an endpoint effect and can adaptively separate two harmonic signals with similar frequencies in a composite signal, this method has been widely used in the field of denoising. However, the following problems still exist in the current field of signal denoising: 1) The decomposition effect of VMD is limited by the influence of two parameters, the number of modal components and the penalty factor. The optimal solution is limited by the influence of the two parameters, and the calculation is cumbersome; 2) The parameters are selected by empirical formulas, which often cannot reach the optimal solution. Therefore, an intelligent algorithm is needed to optimize the hyperparameters to find the optimal solution and solve the problem of the intelligent algorithm falling into the local optimum; 3) What kind of fitness objective constraint function can be used to retain the signal trend term to the greatest extent and separate the noise at the same time; 4) How to automatically screen out the appropriate IMF components through non-human methods to reduce the impact of human factors. Summary of the invention

[0004] The purpose of the present invention is to propose an intelligent algorithm for denoising the rotor roundness measurement signal of ASCSA-SVMD-PE, so as to solve the above-mentioned technical problems existing in the current field of signal denoising.

[0005] To achieve the above object, the present invention provides the following technical solution: a rotor roundness measurement signal noise reduction method based on improved ASCSA-SVMD-PE, comprising the following steps:

[0006] S1. Use TENT mapping to initialize the sparrow population, add adaptive weight coefficients to optimize the search flexibility of the explorer sparrow individual, use the Cauchy mutation operator to update the search step of the explorer sparrow individual, and use the spiral strategy to improve the search path of the follower sparrow individual to obtain the adaptive spiral Cauchy sparrow search algorithm ASCSSA;

[0007] S2. The maximum mutual information coefficient is used as the objective constraint function of ASCSSA optimization of SVMD;

[0008] S3. Use the ASCSSA algorithm to optimize the penalty factors of important hyperparameters of SVMD decomposition;

[0009] S4. Perform SVMD decomposition on the rotor circularity measurement signal according to the obtained optimal hyperparameters to obtain several IMF signal components;

[0010] S5. The PE method is used to entropy arrange the modal components of SVMD decomposition, and only the minimum entropy IMF component is retained, which is the final denoised signal.

[0011] Further, the SVMD decomposition in S4 comprises the following steps:

[0012] S4-1: Decompose the rotor roundness measurement signal f(c) into L-order IMF components u L (c) and the residual signal f r (c) Two parts:

[0013] f(c)=u L (c)+f r (c) Formula 4-1

[0014] In formula 4-1, the residual signal f r (c) Includes the unprocessed part and the sum of the modes obtained before the Lth order

[0015] During the decomposition process, the following constraints should be met to obtain the Lth IMF component:

[0016] S4-1.1: The Lth mode should be minimized with the following constraints:

[0017]

[0018] In formula 4-2, represents the partial derivative with respect to time c, δ(c) represents the Dirac function, and * represents the convolution operation; ω L It represents the center frequency of the Lth-order mode;

[0019] S4-1.2: in u L (c) Based on the effective component, the residual signal fr (c) should be minimized as much as possible. In order to ensure that this constraint can be stably implemented, a suitable filter should be selected with a frequency response of:

[0020]

[0021] In formula 4-3, is the balance parameter;

[0022] The constraints that should be established are:

[0023]

[0024] S4-2: According to the constraints, it is impossible to distinguish the L-order mode and the L-1 mode. Using the idea of ​​J2 constraint, a suitable filter is selected, and its frequency response is:

[0025]

[0026] From this we can establish the constraints:

[0027]

[0028] Therefore, the problem of extracting the modal component IMF can be considered as a minimization problem of the above three constraints:

[0029]

[0030] In formula 4-7, α is the penalty factor;

[0031] S4-3: In order to obtain better convergence, the combination of the penalty factor α and the Lagrangian multiplier λ is used to construct the Lagrangian augmented matrix:

[0032]

[0033] The SVMD in S4 finally obtains the L-order IMF component of the original signal through the above steps.

[0034] Furthermore, the maximum mutual information coefficient calculation process in S2 is as follows:

[0035] For two groups of data G with length o, G = {g1, g2, ..., g o},H={h1,h2,…,h0}, and its mutual information I(G,H) is:

[0036]

[0037] In formula 2-1, p(g,h) is the joint probability density of G and H, p(g), p(h) are the marginal probability densities of G and H; assuming D = {(g i,h i ),i=1,2,…,o}; define J as the u×v grid division, divide G into u segments, and divide H into v segments; B(o) is the upper limit of the grid division u×v, generally taken as o 0.6 ; Calculate each grid division J in the D interval u,v The mutual information value I(J u,v ) and normalize it, the function is:

[0038]

[0039] Then the maximum mutual information coefficient MI(G,H) of G and H is:

[0040] MI(G,H)=max{I'(J u,v ),uv<B(o)} Formula 2-3

[0041] Calculate the maximum mutual information coefficient between adjacent IMF components after SVMD decomposition and record it as KMI k ={k1,k2,…k b-1}, b is the number of IMF components; the maximum mutual information coefficient between each IMF component and the original signal is calculated and recorded as SMI s ={s1,s2,…,s b};

[0042] Then the objective function of ACSSA optimization SVMD decomposition, that is, the calculation formula of the sparrow fitness value in the next section can be expressed as:

[0043]

[0044] The lower the F(α) value is, the lower the similarity between the IMF components is under the current α, while the similarity between the IMF components and the original signal is high. That is, the decomposition process of the rotor circularity measurement signal separates the effective information and the noise information in different IMF components, while ensuring to the greatest extent that the separated effective information has a high degree of similarity with the original rotor circularity measurement signal.

[0045] Furthermore, the ASCSA algorithm in S3 optimizes the SVMD penalty factor α, and the related parameter settings include the following steps:

[0046] S3-1: Related parameter settings: The ACSSA algorithm optimizes SVMD in order to obtain the optimal decomposition scheme of the rotor circularity measurement signal and better separate the noise from the effective information in the rotor circularity measurement signal. Therefore, each SVMD decomposition scheme using different penalty factors α can be regarded as a sparrow, and a sparrow population consists of pop sparrows. After multiple signal noise reduction tests, the maximum number of iterations of the ACSSA intelligent algorithm is determined to be iter maxThe number of times is 20, the size of the sparrow population pop is 20, the explorer ratio is set to 0.2, the safety threshold value ST is 0.8, and the upper and lower limits of the SVMD penalty factor α are set to [500, 1500];

[0047] S3-2: The steps for optimizing the SVMD penalty factor α by ASCSSA are as follows:

[0048] S3-2.1: Introduce the TENT mapping principle to initialize the sparrow population, make the initial sparrow distribution more reasonable, and search for the best rotor circular measurement signal decomposition solution in a wider area;

[0049] Z1=rand(0,1) Formula 3-1

[0050]

[0051] After Bernoulli transformation, we get:

[0052]

[0053] rand(0,1) is to randomly generate a random value between 0 and 1; N is the size of the sparrow population pop multiplied by the dimension j. Since the present invention only optimizes the penalty factor α of SVMD, the dimension j is 1; mod1 is to round down, and then iterate through the function after Bernoulli transformation to generate a Z sequence of size pop; use the Z sequence, multiply the optimization interval length of α plus the optimization interval lower limit to generate a group of sparrow individuals:

[0054]

[0055] Calculate the fitness value of the flock after initialization:

[0056]

[0057] Select 20% of individuals with high fitness values ​​in the population as explorers, and the remaining 80% as followers;

[0058] S3-2.2: Improve the position update of the searcher sparrow individual; the SSA algorithm position update formula is as follows:

[0059] R2=rand(0,1) Formula 3-6

[0060]

[0061] In formula 3-7, p∈(0,1] is a uniform random number, R2 is the warning value for controlling the flight behavior of individual sparrows; ST is the safety threshold for the flight of individual explorers; if R2<ST, it means that the current position of the current individual sparrow is in a safe state, and the individual explorer can search in the sub-interval near the current penalty factor α; when R2≥ST, it is considered that the interval where the individual explorer is located is not safe, and it is necessary to search in other sub-intervals; Q is a random number that obeys the standard normal distribution;

[0062] For the update of the searcher's individual position, an adaptive weight coefficient ω(t) is introduced to improve the quality of the sparrow's individual position:

[0063]

[0064] In formula 3-8, t represents the current number of iterations:

[0065] The improved explorer position update formula is as follows:

[0066]

[0067] S3-2.3: Continue to improve the explorer's individual position update, introduce the Cauchy mutation strategy, to further improve the situation where the explorer is prone to fall into the local optimum, enrich the diversity of the population position, one-dimensional standard Cauchy distribution function:

[0068]

[0069] After further improvement, the formula for updating the explorer's individual position is as follows:

[0070]

[0071] In formula 3-11, Cauchy(0,1) is the standard Cauchy distribution in the interval (0,1); l is the step size control function; is the position of the optimal sparrow individual in t iterations;

[0072] S3-2-4: Improve the position update of followers; SSA algorithm follower position update formula:

[0073]

[0074] In formula 3-13, L∈[0,1] is a random number with uniform distribution: T is a random number with standard normal distribution; is the location of the optimal penalty factor α of the explorer in the t+1th iteration, and It indicates the worst penalty factor α position of the explorer in the tth iteration; A is a random number of 1 or -1; if This means that the current follower has not found the best penalty factor α, and the follower needs to move to a larger interval for optimization; if Then follow the explorer to search for the best result;

[0075] The variable spiral search strategy is introduced to optimize the search path of the follower and improve its global search ability. The improved individual position update formula of the follower sparrow is:

[0076]

[0077] In formula 3-14, Z changes as the number of iterations increases, and is composed of an exponential function based on e; the size and amplitude of the generated spiral are adjusted according to the cosine function; k is the coefficient of variation, which is selected as 5 here;

[0078] When a sparrow is in a dangerous position, whether it is an explorer or a follower, the position should be updated. The update formula is as follows:

[0079]

[0080] β is a random number that follows a standard normal distribution. K∈[-1,1] is a random number. ε is a minimum constant that prevents the denominator from being zero. F i represents the fitness value of the i-th sparrow individual, F g and F w are the best fitness value and the worst fitness value in the iteration process respectively.

[0081] According to the above steps, the best solution for SVMD decomposition of the rotor circularity measurement signal can be obtained, and the best IMF component can be obtained by performing SVMD decomposition according to the best solution.

[0082] Furthermore, in S5, the PE method is used to entropy arrange the IMF components decomposed by SVMD, and only the minimum entropy component is retained, that is, the final noise reduction signal, and the steps are:

[0083] S5-1: For the IMF signal of the rotor roundness measurement signal decomposed by SVMD, assume that one of the IMF sequences is M = {m1,m2,…,m s}, s is the signal length:

[0084] Construct a window of length edim for the sequence M:

[0085] W i ={m i ,m i+1 ,…m i+edim},i∈(1,s-(edim-1)) Formula 5-1

[0086] In formula 5-1, edim is the embedding dimension, which is 3 in the present invention;

[0087] S5-2: For each window W i Sort the elements in and record the sorted index arrangement;

[0088] π i =(π1,π2,…,π edim ) Formula 5-2

[0089] In formula 5-2, π i (j) represents the position of index j in the sorted window;

[0090] S5-3: Calculate the number of occurrences of each permutation index, convert the frequency of each permutation into probability, and calculate its permutation entropy using probability distribution;

[0091]

[0092] In formula 5-4, p k is the permutation probability, r k Indicates the number of times the kth permutation index occurs;

[0093] The PE method is used to entropy arrange the IMF components of SVMD decomposition, remove several high entropy components, and add the remaining signals to obtain the reconstructed signal f'(c). Since the maximum mutual information number is used as the constraint function, the noise signal is effectively separated, so only the minimum entropy component is retained, that is, the final rotor roundness measurement denoised signal.

[0094] Compared with the prior art, the present invention has the following beneficial effects:

[0095] The present invention proposes a rotor roundness measurement signal denoising method based on improved ASCSA-SVMD-PE. In view of the fact that VMD is controlled by two hyperparameters, an improved continuous variational mode decomposition (SVMD) is used to decompose the signal. Compared with traditional VMD, SVMD can adaptively adjust the number of modes without setting the optimized number of modes, and has a better decomposition effect.

[0096] By improving the optimal step length and search strategy of the sparrow algorithm, an adaptive spiral Cauchy sparrow search algorithm is obtained. The SVMD decomposition parameters are intelligently optimized using the adaptive spiral Cauchy sparrow search algorithm to obtain the optimal penalty factor value.

[0097] The SVMD signal decomposition method based on ACSSA algorithm optimization is applied to the field of noise reduction in the process of rotor roundness measurement signal processing. This method improves the speed of signal decomposition, reduces the modal aliasing phenomenon during decomposition, and has certain signal decomposition advantages. The maximum mutual information coefficient is used as the objective constraint function of ACSSA optimization SVMD, which can effectively separate the noise information in the rotor roundness measurement signal and can adapt to more complex systems.

[0098] The PE method is used to entropy arrange the IMF components decomposed by SVMD, and only the minimum entropy component is retained as the reconstructed signal, that is, the final denoised signal is obtained.

[0099] The combined use of the above methods effectively solves the shortcomings and technical problems of the existing rotor roundness measurement signal noise reduction methods, and enhances the practicality and reliability of the noise reduction algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 The overall flow chart of the rotor roundness measurement signal noise reduction of the present invention;

[0101] Figure 2 This is a schematic diagram of the test noise reduction effect of Example 1, including an original signal diagram, a noise-added signal diagram, and a noise reduction signal comparison diagram;

[0102] Figure 3 The iterative line graph of adding sinusoidal signals of different noise levels to the test groups in Example 1 to find the minimum fitness function value F;

[0103] Figure 4 The IMF component diagram obtained by SVMD decomposition of the sinusoidal signals with different noises added to the test group in Example 1;

[0104] Figure 5 This is a schematic diagram of the effect of using the method of the present invention to reduce the noise of the rotor roundness measurement signals of groups a, b, and c in Example 2, including an original signal diagram and a noise reduction signal diagram;

[0105] Figure 6 An iterative line graph for finding the minimum fitness function value F for the rotor roundness measurement signal of the test group in Example 2;

[0106] Figure 7 This is the IMF component diagram obtained by SVMD decomposition of the rotor circularity measurement signal of the test group in Example 2. DETAILED DESCRIPTION

[0107] In order to enable those skilled in the art to better understand the solution method of the present invention, the present invention is further described in detail below with reference to specific embodiments in conjunction with the accompanying drawings. However, this method is not limited to a specific example and can be applied to similar situations.

[0108] Example 1

[0109] See also Figures 1 to 4 Before the pre-processed rotor circular measurement original signal is subjected to noise reduction, a sine signal with a length of 300, an amplitude of 1, and a period of 100 is used to add Gaussian white noise with a signal-to-noise ratio of 5dB, 10dB, and 15dB, and the present invention is used to perform noise reduction processing on it. The effectiveness of the present invention is verified by comparing the determination coefficient, signal-to-noise ratio, and mean square error before and after noise reduction. The comparison diagram of the original signal, the noise-added signal, and the noise-reduced signal is shown in Table 1. The relevant parameter table of the test signal algorithm is shown in Table 2.

[0110] Table 1 Test signal algorithm related parameters

[0111]

[0112] Table 2 Test signal related evaluation index table

[0113]

[0114] After denoising by the algorithm, the signal-to-noise ratio of the denoised signal increased by a maximum of 12.82dB compared to the noisy signal. The determination coefficient of the denoised signal to the original signal increased by a maximum of 0.28 compared to the noisy signal to the original signal, and the determination coefficients after denoising were all greater than 0.98. The mean square error of the denoised signal to the original signal decreased by a maximum of 0.14 compared to the mean square error of the noisy signal to the original signal, and the mean square error of the denoised signal to the original signal was less than 0.01. This illustrates the effectiveness of the denoising of the present invention.

[0115] Example 2

[0116] See also Figures 5 to 7 , the comparison diagram of the effect of using the present invention to reduce the noise of the rotor roundness measurement signals of groups a, b, and c. The algorithm-related parameter table is shown in Table 3.

[0117] Table 3 Parameters of rotor roundness measurement signal algorithm

[0118]

[0119] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the above examples. Any changes, modifications, additions or substitutions made by those skilled in the art within the essential scope of the present invention shall also fall within the protection scope of the present invention. The present invention does not elaborate on some of the known technologies of those skilled in the art.

Claims

1. A rotor roundness measurement signal noise reduction method based on improved ASCSA-SVMD-PE, characterized in that: The following steps are involved: S1. Use TENT mapping to initialize the sparrow population, add adaptive weight coefficients to optimize the search flexibility of the explorer sparrow individual, use the Cauchy mutation operator to update the search step of the explorer sparrow individual, and use the spiral strategy to improve the search path of the follower sparrow individual to obtain the adaptive spiral Cauchy sparrow search algorithm ASCSSA; S2. The maximum mutual information coefficient is used as the objective constraint function of ASCSSA optimization of SVMD; S3. Use the ASCSSA algorithm to optimize the penalty factors of important hyperparameters of SVMD decomposition; S4. Perform SVMD decomposition on the rotor circularity measurement signal according to the obtained optimal hyperparameters to obtain several IMF signal components; S5. The PE method is used to entropy arrange the modal components of SVMD decomposition, and only the minimum entropy IMF component is retained, which is the final denoised signal.

2. The rotor roundness measurement signal noise reduction method based on improved ASCSA-SVMD-PE according to claim 1 is characterized in that: The SVMD decomposition in S4 includes the following steps: S4-1: Decompose the rotor roundness measurement signal f(c) into L-order IMF components u L (c) and the residual signal f r (c) Two parts: f(c)=u L (c)+f r (c) Formula 4-1 In formula 4-1, the residual signal f r (c) Includes the unprocessed part and the sum of the modes obtained before the Lth order During the decomposition process, the following constraints should be met to obtain the Lth IMF component: S4-1.1: The Lth mode should be minimized with the following constraints: In formula 4-2, represents the partial derivative with respect to time c, δ(c) represents the Dirac function, and * represents the convolution operation; ω L It represents the center frequency of the Lth-order mode; S4-1.2: in u L (c) Based on the effective component, the residual signal f r (c) should be minimized as much as possible. In order to ensure that this constraint can be stably implemented, a suitable filter should be selected with a frequency response of: In formula 4-3, is the balance parameter; The constraints that should be established are: S4-2: According to the constraints, it is impossible to distinguish the L-order mode and the L-1 mode. Using the idea of ​​J2 constraint, a suitable filter is selected, and its frequency response is: From this we can establish the constraints: Therefore, the problem of extracting the modal component IMF can be considered as a minimization problem of the above three constraints: In formula 4-7, α is the penalty factor; S4-3: In order to obtain better convergence, the combination of the penalty factor α and the Lagrangian multiplier λ is used to construct the Lagrangian augmented matrix: The SVMD in S4 finally obtains the L-order IMF component of the original signal through the above steps.

3. The rotor roundness measurement signal noise reduction method based on improved ASCSA-SVMD-PE according to claim 1 is characterized in that: The maximum mutual information coefficient calculation process in S2 is as follows: For two groups of data G with length o, G = {g1, g2, ..., g o },H={h1,h2,…,h0}, and its mutual information I(G,H) is: In formula 2-1, p(g,h) is the joint probability density of G and H, p(g), p(h) are the marginal probability densities of G and H; assuming D = {(g i ,h i ),i=1,2,…,o}; define J as the u×v grid division, divide G into u segments, and divide H into v segments; B(o) is the upper limit of the grid division u×v, generally taken as o 0.6 ; Calculate the J of each grid division in the D interval u,v The mutual information value I(J u,v ) and normalize it, the function is: Then the maximum mutual information coefficient MI(G,H) of G and H is: MI(G,H) = max{I'(J u,v ), uv < B(o)} Equation 2-3 Calculate the maximum mutual information coefficient between adjacent IMF components after SVMD decomposition and record it as KMI k ={k1,k2,…k b-1 }, b is the number of IMF components; the maximum mutual information coefficient between each IMF component and the original signal is calculated and recorded as SMI s ={s1,s2,…,s b }; Then the objective function of ACSSA optimization SVMD decomposition, that is, the calculation formula of the sparrow fitness value in the next section can be expressed as: The lower the F(α) value is, the lower the similarity between the IMF components is under the current α, while the similarity between the IMF components and the original signal is high. That is, the decomposition process of the rotor circularity measurement signal separates the effective information and the noise information in different IMF components, while ensuring to the greatest extent that the separated effective information has a high degree of similarity with the original rotor circularity measurement signal.

4. The rotor roundness measurement signal noise reduction method based on improved ASCSA-SVMD-PE according to claim 1 is characterized in that: The ASCSA algorithm in S3 optimizes the SVMD penalty factor α, and the related parameter settings include the following steps: S3-1: Related parameter settings: The ACSSA algorithm optimizes SVMD in order to obtain the optimal decomposition scheme of the rotor circularity measurement signal and better separate the noise from the effective information in the rotor circularity measurement signal. Therefore, each SVMD decomposition scheme using different penalty factors α can be regarded as a sparrow, and a sparrow population consists of pop sparrows. After multiple signal noise reduction tests, the maximum number of iterations of the ACSSA intelligent algorithm is determined to be iter max The number of times is 20, the size of the sparrow population pop is 20, the explorer ratio is set to 0.2, the safety threshold value ST is 0.8, and the upper and lower limits of the SVMD penalty factor α are set to [500, 1500]; S3-2: The steps for optimizing the SVMD penalty factor α by ASCSSA are as follows: S3-2.1: Introduce the TENT mapping principle to initialize the sparrow population, make the initial sparrow distribution more reasonable, and search for the best rotor circular measurement signal decomposition solution in a wider area; Z1=rand(0,1) Formula 3-1 After Bernoulli transformation, we get: rand(0,1) is to randomly generate a random value between 0 and 1; N is the size of the sparrow population pop multiplied by the dimension j. Since the present invention only optimizes the penalty factor α of SVMD, the dimension j is 1; mod1 is to round down, and then iterate through the function after Bernoulli transformation to generate a Z sequence of size pop; use the Z sequence, multiply the optimization interval length of α plus the optimization interval lower limit to generate a group of sparrow individuals: Calculate the fitness value of the flock after initialization: Select 20% of individuals with high fitness values ​​in the population as explorers, and the remaining 80% as followers; S3-2.2: Improve the position update of the searcher sparrow individual; the SSA algorithm position update formula is as follows: R2=rand(0,1) Formula 3-6 In formula 3-7, p∈(0,1] is a uniform random number, R2 is the warning value for controlling the flight behavior of individual sparrows; ST is the safety threshold for the flight of individual explorers; if R2<ST, it means that the current position of the individual sparrow is in a safe state, and the individual explorer can search in the sub-interval near the current penalty factor α; when R2≥ST, it is considered that the interval where the individual explorer is located is not safe, and it is necessary to search in other sub-intervals; Q is a random number that obeys the standard normal distribution; For the update of the searcher's individual position, an adaptive weight coefficient ω(t) is introduced to improve the quality of the sparrow's individual position: In formula 3-8, t represents the current number of iterations: The improved explorer position update formula is as follows: S3-2.3: Continue to improve the explorer's individual position update, introduce the Cauchy mutation strategy, to further improve the situation where the explorer is prone to fall into the local optimum, enrich the diversity of the population position, one-dimensional standard Cauchy distribution function: After further improvement, the formula for updating the explorer's individual position is as follows: In formula 3-11, Cauchy(0,1) is the standard Cauchy distribution in the interval (0,1); l is the step size control function; is the position of the optimal sparrow individual in t iterations; S3-2-4: Improve the position update of followers; SSA algorithm follower position update formula: In formula 3-13, L∈[0,1] is a random number with uniform distribution: T is a random number with standard normal distribution; is the location of the optimal penalty factor α of the explorer in the t+1th iteration, and It indicates the worst penalty factor α position of the explorer in the tth iteration; A is a random number of 1 or -1; if This means that the current follower has not found the best penalty factor α, and the follower needs to move to a larger interval for optimization; if Then follow the explorer to search for the best result; The variable spiral search strategy is introduced to optimize the search path of the follower and improve its global search ability. The improved individual position update formula of the follower sparrow is: In formula 3-14, Z changes as the number of iterations increases, and is composed of an exponential function based on e; the size and amplitude of the generated spiral are adjusted according to the cosine function; k is the coefficient of variation, which is selected as 5 here; When a sparrow is in a dangerous position, whether it is an explorer or a follower, the position should be updated. The update formula is as follows: β is a random number that follows a standard normal distribution; K∈[-1,1] is a random number; ε is a minimum constant to prevent the denominator from being zero; F i represents the fitness value of the i-th sparrow individual, F g and F w They are the best fitness value and the worst fitness value in the iteration process respectively; According to the above steps, the best solution for SVMD decomposition of the rotor circularity measurement signal can be obtained, and the best IMF component can be obtained by performing SVMD decomposition according to the best solution.

5. The intelligent algorithm for reducing the noise of rotor roundness measurement signal based on improved ASCSA-SVMD-PE according to claim 1 is characterized in that: In S5, the PE method is used to perform entropy arrangement on the IMF components decomposed by SVMD, and only the minimum entropy component is retained, that is, the final noise reduction signal, and the steps are as follows: S5-1: For the IMF signal of the rotor roundness measurement signal decomposed by SVMD, assume that one of the IMF sequences is M = {m1,m2,…,m s }, s is the signal length: Construct a window of length edim for the sequence M: W i = {m i , m i+1 , … m i+edim}, i ∈ (1, s - (edim - 1)) Equation 5-1 In formula 5-1, edim is the embedding dimension, which is 3 in the present invention; S5-2: For each window W i Sort the elements in and record the sorted index arrangement; p i =(π1,π2,…,π edim ) formula 5-2 In formula 5-2, π i (j) represents the position of index j in the sorted window; S5-3: Calculate the number of occurrences of each permutation index, convert the frequency of each permutation into probability, and calculate its permutation entropy using probability distribution; In formula 5-4, p k is the permutation probability, r k Indicates the number of times the kth permutation index occurs; The PE method is used to perform entropy arrangement on the IMF components decomposed by SVMD. Since the maximum mutual information number is used as the constraint function, the noise signal is effectively separated, so only the minimum entropy component is retained, that is, the final rotor roundness measurement noise reduction signal.

Images