A method for fan blade fault diagnosis based on soundprint signals

Abstract

The invention discloses a fan blade fault diagnosis method based on soundprint signals, comprising the following steps: collecting fan blade soundprint signals and filtering them; decomposing the filtered soundprint signals into a plurality of intrinsic mode functions by a fully integrated empirical mode decomposition method; converting the intrinsic mode functions into a two-dimensional mirror snowflake graph and normalizing them; optimizing the internal penalty parameters and kernel parameters of a support vector machine by an improved red-billed blue magpie optimization algorithm to obtain an IRBMO-SVM model; inputting the normalized two-dimensional mirror snowflake graph into the IRBMO-SVM model for fault diagnosis; the invention adopts the improved red-billed blue magpie optimization algorithm to optimize the support vector machine, thereby improving the calculation efficiency of the support vector machine and further improving the fault diagnosis accuracy of the fan blade.

Description

Technical Field

[0001] The present invention relates to the technical field of fan blade fault diagnosis, and in particular to a fan blade fault diagnosis method based on voiceprint signals. Background Art

[0002] With the development and promotion of new energy technologies, wind energy as a clean energy source has received more and more attention. At the same time, wind turbine blades are the main equipment of wind power generation systems, and their health status directly affects the power generation efficiency. The wind turbine blades rotate under the action of wind energy, and the energy is converted into electrical energy through transmission. Natural wind is random and volatile, and is mixed with impurities such as sand, ice and snow. These factors can cause ice, cracking, and light transmission on the blades, which in turn affects the efficient and safe operation of the wind turbine.

[0003] The development of wind turbine blade fault diagnosis technology has gone through several stages. From the initial simple monitoring and inspection, to the later model-based methods, the application of traditional machine learning and other technologies, to the current development of intelligent and networked directions, the technology has been continuously improved and perfected. Especially in recent years, with the rapid development of the wind power industry and the popularization of intelligent technology, wind turbine blade fault diagnosis technology has made significant progress. At present, wind turbine blade fault diagnosis technology has formed a situation where multiple methods coexist. Among them, computer vision technology, acoustic emission technology, infrared thermal imaging technology, vibration analysis technology and ultrasonic technology are the most widely used methods. Among them, the detection method based on machine vision is to obtain the information of the image sequence of different positions and perspectives of the target object. It puts forward high requirements on imaging equipment and data processing, especially in complex backgrounds, requiring a lot of calculations, and the accuracy depends on image processing technology. Therefore, it is of great significance to develop wind turbine blade fault diagnosis based on voiceprint signals.

[0004] The penalty factor determines the degree of penalty for misclassified samples of the support vector machine (SVM). It can coordinate the classification interval and the sample misclassification rate, take into account the confidence level and empirical risk of the support vector machine (SVM), and make the fan blade fault diagnosis model take into account both its generalization and accuracy. When the penalty factor is large, the support vector machine (SVM) will increase the penalty for misclassification, requiring the algorithm to meet all constraints, and not allowing misclassification, which reduces the fitting error, but will lead to weak generalization ability of the support vector machine (SVM). Higher classification standards will also greatly extend the training time of the support vector machine (SVM); when the penalty factor is small, the support vector machine (SVM) will not pay too much attention to the penalty for sample misclassification, which greatly improves the generalization ability, but the fitting error increases, which will bring greater empirical risk. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a fan blade fault diagnosis method based on soundprint signals, which aims to solve the problems mentioned in the background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for diagnosing fan blade faults based on soundprint signals, comprising the following steps:

[0007] Step S1: collecting the fan blade soundprint signal and performing filtering processing;

[0008] Step S2: decomposing the filtered voiceprint signal into multiple intrinsic mode functions by using a fully integrated empirical mode decomposition method;

[0009] Step S3: converting the intrinsic mode function into a two-dimensional mirror snowflake graph and performing normalization processing;

[0010] Step S4: optimizing the internal penalty parameters and kernel parameters of the support vector machine (SVM) by improving the red-billed blue magpie optimization algorithm to obtain an IRBMO-SVM model; the improved red-billed blue magpie optimization algorithm adopts the initial population of the original red-billed blue magpie optimization algorithm using the improved Circle chaos mapping, and introduces the lens imaging reverse learning mechanism to perturb the red-billed blue magpie population;

[0011] Step S5: inputting the normalized two-dimensional mirror snowflake image into the IRBMO-SVM model for fault diagnosis;

[0012] The specific process of step S4 is:

[0013] Step S4.1: using improved Circle chaos mapping to generate the initial population;

[0014] Step S4.2: Evaluate the fitness of each individual red-billed blue magpie;

[0015] Step S4.3: simulating prey-seeking behavior, where prey-seeking behavior is divided into small groups and groups;

[0016] Step S4.4: simulating attacking prey;

[0017] Step S4.5: simulating prey storage behavior;

[0018] Step S4.6: In the later stage of iteration, when the optimal value falls into a local optimal solution, a lens imaging reverse learning mechanism is introduced on the basis of the original red-billed blue magpie optimization algorithm to disturb the red-billed blue magpie population;

[0019] Step S4.7: Determine whether the current iteration has reached the preset maximum number of times. If not, continue the iteration. Otherwise, stop the iteration and output the optimal red-billed blue magpie individual, which is the optimal value of the penalty parameter and kernel parameter of the support vector machine SVM.

[0020] Furthermore, the initial population using improved Circle chaotic mapping is expressed as:

[0021] ;

[0022] ;

[0023] ;

[0024] In the formula, represents the red-billed blue magpie population; represents the value of the i-th red-billed blue magpie individual in the j-th dimension, i∈1,2,…,n, j∈1,2,…,d; n represents the population size of the red-billed blue magpie; d represents the problem dimension; and Respectively represent the upper and lower bounds of the problem; Represents a random number between 0 and 1; Indicates the current iteration number;

[0025] Each red-billed blue magpie individual represents a set of penalty parameters C and kernel parameters of the support vector machine SVM Parameter solution of .

[0026] Furthermore, in step S4.2, the fitness function It is expressed as:

[0027] ;

[0028] In the formula, is the relaxation factor.

[0029] Furthermore, in step S4.3, when the red-billed blue magpie population explores prey in small groups, it is expressed as:

[0030] ;

[0031] In the formula, represents the updated position of the i-th red-billed blue magpie individual; It means randomly selecting 2 to 5 red-billed blue magpie individuals from the red-billed blue magpie population; represents the position of the mth randomly selected red-billed blue magpie individual; represents the position of the i-th red-billed blue magpie individual; Indicates the position of the most recently selected red-billed blue magpie individual in the current iteration;

[0032] When a population of blue magpies explores prey as a group, it is expressed as:

[0033] ;

[0034] In the formula, represents 10 to n red-billed blue magpie individuals randomly selected from the red-billed blue magpie population.

[0035] Furthermore, the simulated attack on prey behavior is expressed as:

[0036] ;

[0037] ;

[0038] In the formula, Indicates the location of prey; Represents a constant that changes as the number of iterations changes; is the maximum number of iterations; Represents random numbers used to generate a standard normal distribution.

[0039] Furthermore, the simulated prey storage behavior is expressed as:

[0040] ;

[0041] In the formula, and They respectively represent the fitness values ​​of the i-th red-billed blue magpie before and after the individual position is updated.

[0042] Furthermore, the lens imaging reverse learning mechanism is introduced on the basis of the original red-billed blue magpie optimization algorithm to disturb the red-billed blue magpie population, which is expressed as:

[0043] ;

[0044] ;

[0045] In the formula, Indicates the individual position of the red-billed blue magpie after reverse learning of lens imaging; Individual positions of red-billed blue magpies before reverse learning for lens imaging; Represents a constant that changes with the number of iterations.

[0046] Furthermore, the specific process of step S2 is as follows:

[0047] Step S2.1: The voiceprint signal after filtering Add K times of Gaussian white noise with a mean of 0 to construct a sequence to be decomposed for N experiments , expressed as:

[0048] ;

[0049] ;

[0050] In the formula, is the weight coefficient of Gaussian white noise; is the Gaussian white noise added for the i-th time; The first individual data;

[0051] Step S2.2: Perform empirical mode decomposition EMD to obtain the first eigenmode function and the residual signal , expressed as:

[0052] ;

[0053] ;

[0054] In the formula, represents the i-th intrinsic mode function IMF component obtained in the first-order decomposition;

[0055] Step S2.3: Determine the residual signal Is it a monotonic function? If not, repeat steps S2.1-S2.2 until the residual signal is a monotonic function, end, get N and N residual signals , expressed as:

[0056] ;

[0057] In the formula, represents the i-th eigenmode function.

[0058] Furthermore, the specific process of step S3 is as follows:

[0059] Step S3.1: converting the intrinsic mode function into a two-dimensional mirror snowflake graph by using a symmetric point mode formula;

[0060] Step S3.2: normalize the two-dimensional mirror snowflake image.

[0061] Compared with the existing technology, the present invention has the following beneficial effects: the present invention adopts the fully integrated empirical mode decomposition method CEEMDAN to decompose the voiceprint signal, uses the symmetrical point pattern to convert the one-dimensional signal into a two-dimensional snowflake graph, thereby improving the intuitiveness of the data, and adopts the improved red-billed blue magpie optimization algorithm to optimize the support vector machine SVM, thereby improving the computational efficiency of the support vector machine SVM, thereby improving the fault diagnosis accuracy of the fan blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0063] like Figure 1 As shown, the present invention provides a technical solution: a fan blade fault diagnosis method based on a soundprint signal, comprising the following steps:

[0064] Step S1: Use a sound collection device to collect the soundprint signal of the fan blade, and filter the collected soundprint signal through AUKF adaptive filtering to reduce the interference of external soundprint signals on the soundprint signal of the fan blade.

[0065] Step S2: Decompose the filtered voiceprint signal into multiple intrinsic mode functions (IMFs) through the fully integrated empirical mode decomposition method CEEMDAN.

[0066] Step S3: Convert the intrinsic mode function (IMF) of the one-dimensional data into a two-dimensional mirror snowflake graph and perform normalization processing.

[0067] Step S4: By improving the red-billed blue magpie optimization algorithm, the internal penalty parameters and kernel parameters of the support vector machine (SVM) are optimized to obtain the IRBMO-SVM model.

[0068] Step S5: Input the normalized two-dimensional mirror snowflake image into the IRBMO-SVM model for fault diagnosis.

[0069] The specific process of decomposing the filtered voiceprint signal into multiple intrinsic mode functions (IMFs) through CEEMDAN is as follows:

[0070] Step S2.1: The voiceprint signal after filtering Add K times of Gaussian white noise with a mean of 0 to construct a sequence to be decomposed for N experiments , expressed as:

[0071] (1);

[0072] (2);

[0073] In the formula, is the weight coefficient of Gaussian white noise; is the Gaussian white noise added for the i-th time; The first data.

[0074] Step S2.2: Perform empirical mode decomposition EMD to obtain the first eigenmode function and the residual signal , expressed as:

[0075] (3);

[0076] (4);

[0077] In the formula, represents the i-th intrinsic mode function IMF component obtained in the first-order decomposition.

[0078] Step S2.3: Determine the residual signal Is it a monotonic function? If not, repeat steps S2.1-S2.2 until the residual signal is a monotonic function, end, get N and N residual signals , expressed as:

[0079] (5);

[0080] In the formula, represents the i-th eigenmode function.

[0081] The specific process of converting the intrinsic mode function (IMF) of one-dimensional data into a two-dimensional mirror snowflake graph and performing normalization is as follows:

[0082] Step S3.1: Convert the intrinsic mode function (IMF) of the one-dimensional data into a two-dimensional mirror snowflake graph through the symmetric point pattern formula, expressed as:

[0083] (6);

[0084] (7);

[0085] (8);

[0086] In the formula, represents the polar radius; Indicates the counter-clockwise rotation angle of polar coordinates; Indicates the clockwise rotation angle of polar coordinates; represents the amplitude of the voiceprint signal at time p; Indicates time The amplitude of the voiceprint signal at is the time interval coefficient; Indicates the minimum value of the voiceprint signal amplitude; Indicates the maximum value of the voiceprint signal amplitude; represents the rotation angle of the mirror symmetry plane; Represents the angular magnification factor.

[0087] Step S3.2: normalize the two-dimensional mirror snowflake graph, expressed as:

[0088] (9);

[0089] In the formula, Represents a normalized two-dimensional mirrored snowflake graph; is a two-dimensional mirror snowflake image of the original input; is the maximum value of the two-dimensional mirror snowflake graph; is the minimum value of the two-dimensional mirror snowflake graph.

[0090] The improved Circle chaotic mapping is used to improve the problem of uneven initial population distribution of the original red-billed blue magpie optimization algorithm. Since the original red-billed blue magpie optimization algorithm is prone to fall into local optimal values, the lens imaging reverse learning mechanism is introduced to disturb the red-billed blue magpie population.

[0091] Among them, the internal penalty parameter C and kernel parameter of the support vector machine SVM are improved by improving the red-billed blue magpie optimization algorithm. The specific process of optimizing and obtaining the IRBMO-SVM model is as follows:

[0092] Step S4.1: Using the improved initial population of Circle chaotic mapping to solve the uneven distribution problem caused by random initialization; considering that the original Circle chaotic mapping has a relatively dense value between [0.2, 0.6] and its distribution is not uniform, the original Circle chaotic mapping is improved to make it more uniform;

[0093] (10);

[0094] (11);

[0095] (12);

[0096] In the formula, represents the red-billed blue magpie population; represents the value of the i-th red-billed blue magpie individual in the j-th dimension, i∈1,2,…,n, j∈1,2,…,d; n represents the population size of the red-billed blue magpie; d represents the problem dimension; and Respectively represent the upper and lower bounds of the problem; Represents a random number between 0 and 1; Indicates the current iteration number.

[0097] Among them, each red-billed blue magpie individual represents a set of penalty parameters C and kernel parameters of the support vector machine SVM. Parameter solution of .

[0098] Step S4.2: Evaluate the fitness of each individual red-billed blue magpie. The fitness of each individual red-billed blue magpie is the penalty parameter C and kernel parameter of each support vector machine SVM. Performance; Fitness function It is expressed as:

[0099] (13);

[0100] In the formula, is the relaxation factor.

[0101] Step S4.3: simulating prey-seeking behavior;

[0102] In the process of searching for prey, red-billed blue magpie populations usually act in small groups (2 to 5) or groups (more than 10) to improve search efficiency; they use a variety of means, such as jumping on the ground, walking or searching for prey resources in trees; this variety of means enables red-billed blue magpie populations to adopt a variety of hunting strategies according to environmental conditions and available resources to ensure sufficient prey supply. When red-billed blue magpie populations explore prey in small groups, it is expressed as:

[0103] (14);

[0104] In the formula, represents the updated position of the i-th red-billed blue magpie individual; It means randomly selecting 2 to 5 red-billed blue magpie individuals from the red-billed blue magpie population; represents the position of the mth randomly selected red-billed blue magpie individual; represents the position of the i-th red-billed blue magpie individual; Indicates the position of the most recently selected red-billed blue magpie individual in the current iteration.

[0105] When a population of blue magpies explores prey as a group, it is expressed as:

[0106] (15);

[0107] In the formula, represents 10 to n red-billed blue magpie individuals randomly selected from the red-billed blue magpie population.

[0108] Step S4.4: Simulate the attack behavior of prey, expressed as:

[0109] (16);

[0110] (17);

[0111] In the formula, Indicates the location of prey; Represents a constant that changes as the number of iterations changes; Indicates the maximum number of iterations; Represents random numbers used to generate a standard normal distribution (mean 0, standard deviation 1).

[0112] Step S4.5: Simulate the prey storage behavior, expressed as:

[0113] (18);

[0114] In the formula, and They respectively represent the fitness values ​​of the i-th red-billed blue magpie before and after the individual position is updated.

[0115] Step S4.6: In the later stage of iteration, when the optimal value falls into the local optimal solution, the lens imaging reverse learning mechanism is introduced on the basis of the original red-billed blue magpie optimization algorithm to disturb the red-billed blue magpie population so that it jumps out of the local optimal solution, which is expressed as:

[0116] (19);

[0117] (20);

[0118] In the formula, Indicates the individual position of the red-billed blue magpie after reverse learning of lens imaging; Individual positions of red-billed blue magpies before reverse learning for lens imaging; Represents a constant that changes with the number of iterations.

[0119] Step S4.7: Determine whether the current iteration has reached the preset maximum number of times. If not, continue the iteration; otherwise, stop the iteration and output the optimal red-billed blue magpie individual, which is the penalty parameter C and kernel parameter of the support vector machine SVM. Optimal value.

[0120] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing fan blade faults based on voiceprint signals, characterized in that: The steps include: Step S1: collecting the fan blade soundprint signal and performing filtering processing; Step S2: decomposing the filtered voiceprint signal into multiple intrinsic mode functions by using a fully integrated empirical mode decomposition method; Step S3: converting the intrinsic mode function into a two-dimensional mirror snowflake graph and performing normalization processing; Step S4: optimizing the internal penalty parameters and kernel parameters of the support vector machine (SVM) by improving the red-billed blue magpie optimization algorithm to obtain an IRBMO-SVM model; the improved red-billed blue magpie optimization algorithm adopts the initial population of the original red-billed blue magpie optimization algorithm using the improved Circle chaos mapping, and introduces the lens imaging reverse learning mechanism to perturb the red-billed blue magpie population; Step S5: inputting the normalized two-dimensional mirror snowflake image into the IRBMO-SVM model for fault diagnosis; The specific process of step S4 is: Step S4.1: using improved Circle chaos mapping to generate the initial population; Step S4.2: Evaluate the fitness of each individual red-billed blue magpie; Step S4.3: simulating prey-seeking behavior, where prey-seeking behavior is divided into small groups and groups; Step S4.4: simulating attacking prey; Step S4.5: simulating prey storage behavior; Step S4.6: In the later stage of iteration, when the optimal value falls into a local optimal solution, a lens imaging reverse learning mechanism is introduced on the basis of the original red-billed blue magpie optimization algorithm to disturb the red-billed blue magpie population; Step S4.7: Determine whether the current iteration has reached the preset maximum number of times. If not, continue the iteration. Otherwise, stop the iteration and output the optimal red-billed blue magpie individual, which is the optimal value of the penalty parameter and kernel parameter of the support vector machine SVM. The improved Circle chaotic map initial population is expressed as: ； ； ； In the formula, represents the red-billed blue magpie population; represents the value of the i-th red-billed blue magpie individual in the j-th dimension, i∈1,2,…,n, j∈1,2,…,d; n represents the population size of the red-billed blue magpie; d represents the problem dimension; and Respectively represent the upper and lower bounds of the problem; Represents a random number between 0 and 1; Indicates the current iteration number; Each red-billed blue magpie individual represents a set of penalty parameters C and kernel parameters of the support vector machine SVM Parameter solution of ; In step S4.2, the fitness function It is expressed as: ； In the formula, is the relaxation factor; In step S4.3, when the red-billed blue magpie population explores prey in small groups, it is expressed as: ； In the formula, represents the updated position of the i-th red-billed blue magpie individual; It means randomly selecting 2 to 5 red-billed blue magpie individuals from the red-billed blue magpie population; represents the position of the mth randomly selected red-billed blue magpie individual; represents the position of the i-th red-billed blue magpie individual; Indicates the position of the most recently selected red-billed blue magpie individual in the current iteration; When a population of blue magpies explores prey as a group, it is expressed as: ； In the formula, represents 10 to n red-billed blue magpie individuals randomly selected from the red-billed blue magpie population; The simulated attack on prey behavior is expressed as: ； ； In the formula, Indicates the location of prey; Represents a constant that changes as the number of iterations changes; is the maximum number of iterations; Represents random numbers used to generate standard normal distribution; The simulated prey storage behavior is expressed as: ； In the formula, and They represent the fitness values ​​of the i-th red-billed blue magpie before and after the individual position is updated; The lens imaging reverse learning mechanism is introduced on the basis of the original red-billed blue magpie optimization algorithm to disturb the red-billed blue magpie population, which is expressed as: ； ； In the formula, Indicates the individual position of the red-billed blue magpie after reverse learning of lens imaging; Individual positions of red-billed blue magpies before reverse learning for lens imaging; represents a constant that changes with the number of iterations; The specific process of step S2 is: Step S2.1: The voiceprint signal after filtering Add K times of Gaussian white noise with a mean of 0 to construct a sequence to be decomposed for N experiments , expressed as: ； ； In the formula, is the weight coefficient of Gaussian white noise; is the Gaussian white noise added for the i-th time; The first individual data; Step S2.2: Perform empirical mode decomposition EMD to obtain the first eigenmode function and the residual signal , expressed as: ； ； In the formula, represents the i-th intrinsic mode function IMF component obtained in the first-order decomposition; Step S2.3: Determine the residual signal Is it a monotonic function? If not, repeat steps S2.1-S2.2 until the residual signal is a monotonic function, end, get N and N residual signals , expressed as: ； In the formula, represents the i-th eigenmode function.

2. A fan blade fault diagnosis method based on soundprint signals according to claim 1, characterized in that: The specific process of step S3 is: Step S3.1: converting the intrinsic mode function into a two-dimensional mirror snowflake graph by using a symmetric point mode formula; Step S3.2: normalize the two-dimensional mirror snowflake image.

Images