A method for extracting fault features of motor bearings

Abstract

The present invention discloses a method for extracting features of motor bearing faults, comprising the following steps: introducing an inverse tangent non-convex penalty function to define a regularization term in second-order total variation denoising; optimizing the regularization parameter λ and the convexity parameter a in NCTVD by using the BAS algorithm through the fitness function CK_I, limiting the two parameters within a certain region and solving the optimal parameter combination to ensure the strictly convex nature of the convex optimization problem; then, using a minimum optimization algorithm to solve the generated new convex optimization problem to achieve noise reduction and feature enhancement of vibration signals; applying a TEO demodulation method to perform Teager energy spectrum analysis on the denoised signal to achieve feature extraction of motor bearing faults. The method proposed by the present invention mainly solves the problem that the fault features of motor fault bearings are submerged by strong background noise and the fault information representation is not obvious, and can effectively extract fault features in motor bearing vibration signals, having certain engineering application value.

Description

Technical Field

[0001] The invention belongs to the technical field of mechanical fault diagnosis and signal processing, and in particular relates to a method for extracting fault features of a motor bearing. Background Art

[0002] Rolling bearings are one of the important components of rotating machinery, playing a supporting and guiding role in the operation of equipment. Rolling bearings have low friction during operation, are easy to install and use, and can improve the efficiency of machinery. Their operating status often affects the status of the equipment and even affects the normal operation of the entire system and even the entire production process. The operating conditions of motor rolling bearings are very complex, and their vibration signals are easily affected by noise and interference from other excitation sources, resulting in non-stationary and nonlinear characteristics of the vibration signals. Therefore, preprocessing the vibration signals of motor rolling bearings, thereby reducing noise and effectively extracting fault features, is of great significance for fault diagnosis of motor bearings.

[0003] Since there are various interferences in the vibration signal, and the background noise can easily increase the difficulty of extracting fault information. Therefore, an effective signal analysis method is needed to extract and distinguish the fault features and noise contained in the original vibration signal, and ensure that it has high accuracy. Total Variation Denoising (TVD) is a denoising method based on the optimal problem. It was first used to remove the noise components in the image. Its basic idea is to use a regularization term as the prior knowledge of the original signal to constrain the structural distribution of the reconstructed signal. It is essentially a sparse representation method of the signal. However, the key to the TVD algorithm is the selection of the regularization parameter λ, which will directly affect the performance of denoising. In order to better induce the sparsity of the signal and estimate the signal value, a non-convex penalty function is added to the TVD algorithm, and the minimum optimization algorithm (MM) is adopted to solve the non-convex regularized total variation denoising (NCTVD). The regularization parameter λ and the convexity parameter a in NCTVD directly affect the denoising performance. Therefore, finding a set of optimal parameter combinations is the key to NCTVD.

[0004] Although rolling bearings are relatively mature in terms of design, materials and manufacturing, their life span is very discrete and uncertain. That is, a batch of bearings produced with the same materials, processing technology and production equipment still have very different life spans even under the same external conditions of load, speed and lubrication. This characteristic of bearings makes some bearings still work normally even though they have greatly exceeded their design life, while some bearings have various faults far before reaching their design life. Therefore, the traditional practice of regular maintenance of bearings according to the design life is bound to cause the following situations: "insufficient maintenance" and "excessive maintenance" problems. At the same time, the working environment of rolling bearings is mostly relatively harsh. Long-term high-load work greatly increases the probability of failure of the component, and it is easy to fail. Due to its connectivity and transmission, any small failure may trigger a chain reaction. Once a failure occurs, it will greatly affect the overall performance of the mechanical equipment, resulting in huge economic losses or even irreversible catastrophic consequences.

[0005] Therefore, in order to solve the above problems, this paper proposes a motor bearing fault feature extraction method. Summary of the invention

[0006] In order to solve the above technical problems, the present invention designs a motor bearing fault feature extraction method to solve the problem that the fault features of the vibration signal of the motor bearing under a strong noise background are difficult to extract, so as to achieve the purpose of suppressing the interference of strong background noise and characterizing the fault information of the motor bearing, and provide a new solution for motor bearing fault feature extraction.

[0007] In order to achieve the above technical effects, the present invention is implemented by the following technical solutions: a motor bearing fault feature extraction method of NCTVD-BAS-TEO, the specific steps are as follows:

[0008] Step 1: Introduce the arctangent penalty function into TVD to enhance the algorithm-induced sparsity feature, define a new non-convex total variation denoising NCTVD, collect vibration signals for NCTVD processing and solve them with the MM algorithm;

[0009] Step 2: Initialize the spatial position and parameters of BAS, define the value range of regularization parameter λ and convexity parameter a, and execute the BAS optimization algorithm. By evaluating the fitness function CK_I, the optimal parameter combination [a,λ] is obtained.

[0010] Step 3: The optimal parameter combination is input into NCTVD to obtain the optimal denoising signal, and then TEO is introduced for demodulation processing to obtain the Teager energy spectrum, extract the fault characteristic frequency, and then realize the diagnosis of the fault type.

[0011] Furthermore, in Step 1, the specific steps of constructing non-convex total variation denoising are:

[0012] Step 1.1: Collect the vibration signal x(n) of the motor fault bearing;

[0013] Step 1.2: Find the first-order total variation of the vibration signal x(n);

[0014]

[0015] Where Dx represents the first-order differential, ||·|| p (p≥1) is the lp norm, where the matrix D is defined as follows:

[0016]

[0017] The tridiagonal matrix DDT is represented as:

[0018]

[0019] ||·|| p is the p-order norm, and ||x|| p It can be expressed as:

[0020]

[0021] When p = 1 or p = 2, the first-order and second-order norms of the signal can be expressed as:

[0022] ||x|| 1 =(|x 1 |+|x 2 |+...+|x N |)

[0023]

[0024] The minimum optimization objective function is:

[0025]

[0026] In the formula represents the signal fidelity term, λ>0 is the regularization parameter, and its purpose is to find the signal x to minimize the objective function:

[0027]

[0028] Step 1.3: Calculate the arctan-NCTVD of the vibration signal x(n);

[0029] The non-convex regularization term is the arctangent penalty function (arctan):

[0030]

[0031] Set λ||Dx|| 1 After being transformed into a non-convex regularization term, equation (8) becomes the following:

[0032]

[0033] Step 1.4: Use the MM algorithm to solve the above equation, and the result can be expressed as:

[0034]

[0035] Step 1.5: Convexity condition;

[0036] If the non-convex regularization term of the inverse tangent penalty function remains convex, the condition is:

[0037]

[0038] Furthermore, in Step 2, the specific steps of optimizing the BAS algorithm are as follows:

[0039] Step 2.1: BAS initializes the position and parameters of the longicorn beetle;

[0040] The BAS algorithm mainly relies on the longhorn beetle to continuously compare the odor concentrations of the left and right antennae. Before calculating the odor concentrations of the two antennae, a series of preparations need to be performed. The position of the longhorn beetle in the D-dimensional space is X = (x 1 ,x 2 ,...,x n ), the positions of the left and right antennae of the longicorn beetle are defined as the model shown in the following formula:

[0041]

[0042] In the above formula, l represents the distance between the centroid of the longhorn beetle and its antennae, It indicates that the random unit vector guides the random orientation of the longhorn beetle in the solution space, and it needs to be normalized:

[0043]

[0044] Step2.2: Construct fitness function KC_I;

[0045] Weighted kurtosis (KC_I) is an indicator for measuring the impact characteristics of the signal. It is often used in bearing fault diagnosis because it retains the characteristics of kurtosis and also includes the characteristics of correlation functions. Its calculation formula is:

[0046]

[0047]

[0048] KC_I=K_I×|C_I| r

[0049] r is an adjustable positive real number used to control the fidelity of the output signal. r is generally set to 2.

[0050] Step 2.3: Find the optimal parameter combination;

[0051] The difference in odor concentration perceived by the left and right antennae is defined as the fitness function in the algorithm. The larger the odor concentration, the higher the fitness. The beetle constantly calculates the size of the fitness function to continue moving forward and moves towards the position with higher fitness. The beetle determines the next position according to the following formula:

[0052]

[0053] In the formula, t represents the current number of iterations; f(.) represents the fitness function; δ t It represents the exploration step size at the tth iteration, and the sign() function is the sign function.

[0054] After the iteration conditions are met, the optimal parameter combination [abest, λbest] is output.

[0055] Furthermore, in Step 3, the specific steps of extracting fault features are:

[0056] Step 3.1: The optimal parameter combination is input into NCTVD to obtain the optimal denoised signal;

[0057] Step 3.2: Introduce TEO for demodulation to obtain the Teager energy spectrum.

[0058] Step 3.3: Use the motor fault bearing parameters collected before the experiment to calculate the theoretical fault characteristic frequency, and compare it with the experimental fault characteristic frequency extracted from the Teager energy spectrum to achieve fault diagnosis.

[0059] The beneficial effects of the present invention are:

[0060] Firstly, the inverse tangent non-convex penalty function is introduced to define the regularization term in the second-order total variation denoising; secondly, the BAS algorithm is used to optimize the regularization parameter λ and convexity parameter a in NCTVD through the fitness function CK_I, and the two parameters are restricted to a certain region and the optimal parameter combination is solved to ensure the strict convex nature of the convex optimization problem; then, the minimum optimization algorithm is used to solve the new convex optimization problem generated to achieve denoising and feature enhancement of the vibration signal; finally, the TEO demodulation method is applied to perform Teager energy spectrum analysis on the denoised signal to achieve feature extraction of the motor bearing fault; the method proposed in this paper mainly solves the problem that the fault characteristics of the motor faulty bearing are submerged by the strong background noise and the fault information representation is not obvious, and can effectively extract the fault characteristics in the motor bearing vibration signal, with high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0062] Figure 1 is a flow chart of the method of the present invention;

[0063] Figure 2 It is the Bearing Data Center Test Bench of Case Western Reserve University in the United States;

[0064] Figure 3 It is the time domain waveform of the vibration signal of the inner ring fault bearing;

[0065] Figure 4 It is the frequency domain waveform of the vibration signal of the inner ring fault bearing;

[0066] Figure 5 The inner ring vibration signal passes through the two-dimensional space search trajectory diagram of BAS optimization;

[0067] Figure 6 It is the fitness function change curve of the inner ring vibration signal after BAS optimization;

[0068] Figure 7 It is the spectrum of the inner ring vibration signal after NCTVD-BAS-TEO filtering;

[0069] Figure 8 It is the time domain waveform of the vibration signal of the outer ring fault bearing;

[0070] Fig. 9 It is the frequency domain waveform of the vibration signal of the outer ring fault bearing;

[0071] Fig.10 It is the two-dimensional space search trajectory diagram of the outer ring vibration signal after BAS optimization;

[0072] Fig.11 It is the fitness function change curve of the outer ring vibration signal after BAS optimization;

[0073] Fig.12 It is the spectrum of the outer ring vibration signal after NCTVD-BAS-TEO filtering. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] Example 1

[0076] See also Figures 1 to 12 As shown, a motor bearing fault feature extraction method based on NCTVD-BAS-TEO is shown, and the specific steps are as follows:

[0077] Step 1: Introduce the arctangent penalty function into TVD to enhance the algorithm-induced sparsity feature, define a new non-convex total variation denoising NCTVD, collect vibration signals for NCTVD processing and solve them with the MM algorithm;

[0078] Step 2: Initialize the spatial position and parameters of BAS, define the value range of regularization parameter λ and convexity parameter a, and execute the BAS optimization algorithm. By evaluating the fitness function CK_I, the optimal parameter combination [a,λ] is obtained.

[0079] Step 3: The optimal parameter combination is input into NCTVD to obtain the optimal denoising signal, and then TEO is introduced for demodulation processing to obtain the Teager energy spectrum, extract the fault characteristic frequency, and then realize the diagnosis of the fault type.

[0080] Furthermore, in Step 1, the specific steps of constructing non-convex total variation denoising are:

[0081] Step 1.1: Collect the vibration signal x(n) of the motor fault bearing;

[0082] Step 1.2: Find the first-order total variation of the vibration signal x(n);

[0083]

[0084] Where Dx represents the first-order differential, ||·|| p (p≥1) is the lp norm, where the matrix D is defined as follows:

[0085]

[0086] The tridiagonal matrix DDT is represented as:

[0087]

[0088] ||·|| p is the p-order norm, and ||x|| p It can be expressed as:

[0089]

[0090] When p = 1 or p = 2, the first-order and second-order norms of the signal can be expressed as:

[0091] ||x|| 1 =(|x 1 |+|x 2 |+...+|x N |)

[0092]

[0093] The minimum optimization objective function is:

[0094]

[0095] In the formula represents the signal fidelity term, λ>0 is the regularization parameter, and its purpose is to find the signal x to minimize the objective function:

[0096]

[0097] Step 1.3: Calculate the arctan-NCTVD of the vibration signal x(n);

[0098] The non-convex regularization term is the arctangent penalty function (arctan):

[0099]

[0100] Set λ||Dx|| 1 After being transformed into a non-convex regularization term, equation (8) becomes the following:

[0101]

[0102] Step 1.4: Use the MM algorithm to solve the above equation, and the result can be expressed as:

[0103]

[0104] Step 1.5: Convexity condition;

[0105] If the non-convex regularization term of the inverse tangent penalty function remains convex, the condition is:

[0106]

[0107] Furthermore, in Step 2, the specific steps of optimizing the BAS algorithm are as follows:

[0108] Step 2.1: BAS initializes the position and parameters of the longicorn beetle;

[0109] The BAS algorithm mainly relies on the longhorn beetle to continuously compare the odor concentrations of the left and right antennae. Before calculating the odor concentrations of the two antennae, a series of preparations need to be performed. The position of the longhorn beetle in the D-dimensional space is X = (x 1 ,x 2 ,…,x n ), the positions of the left and right antennae of the longicorn beetle are defined as the model shown in the following formula:

[0110]

[0111] In the above formula, l represents the distance between the centroid of the longhorn beetle and its antennae, It indicates that the random unit vector guides the random orientation of the longhorn beetle in the solution space, and it needs to be normalized:

[0112]

[0113] Step2.2: Construct fitness function KC_I;

[0114] Weighted kurtosis (KC_I) is an indicator for measuring the impact characteristics of the signal. It is often used in bearing fault diagnosis because it retains the characteristics of kurtosis and also includes the characteristics of correlation functions. Its calculation formula is:

[0115]

[0116]

[0117] KC_I=K_I×|C_I| r

[0118] r is an adjustable positive real number used to control the fidelity of the output signal. r is generally set to 2.

[0119] Step 2.3: Find the optimal parameter combination;

[0120] The difference in odor concentration perceived by the left and right antennae is defined as the fitness function in the algorithm. The larger the odor concentration, the higher the fitness. The beetle constantly calculates the size of the fitness function to continue moving forward and moves towards the position with higher fitness. The beetle determines the next position according to the following formula:

[0121]

[0122] In the formula, t represents the current number of iterations; f(.) represents the fitness function; δ t It represents the exploration step size at the tth iteration, and the sign() function is the sign function.

[0123] After the iteration conditions are met, the optimal parameter combination [abest, λbest] is output.

[0124] Furthermore, in Step 3, the specific steps of extracting fault features are:

[0125] Step 3.1: The optimal parameter combination is input into NCTVD to obtain the optimal denoised signal;

[0126] Step 3.2: Introduce TEO for demodulation to obtain the Teager energy spectrum.

[0127] Step 3.3: Use the motor fault bearing parameters collected before the experiment to calculate the theoretical fault characteristic frequency, and compare it with the experimental fault characteristic frequency extracted from the Teager energy spectrum to achieve fault diagnosis.

[0128] Example 2

[0129] Firstly, the inverse tangent non-convex penalty function is introduced to define the regularization term in the second-order total variation denoising; secondly, the BAS algorithm is used to optimize the regularization parameter λ and convexity parameter a in NCTVD through the fitness function CK_I, and the two parameters are restricted to a certain region and the optimal parameter combination is solved to ensure the strict convex nature of the convex optimization problem; then, the minimum optimization algorithm is used to solve the new convex optimization problem generated to achieve denoising and feature enhancement of the vibration signal; finally, the TEO demodulation method is applied to perform Teager energy spectrum analysis on the denoised signal to achieve feature extraction of the motor bearing fault; the method proposed in this paper mainly solves the problem that the fault characteristics of the motor faulty bearing are submerged by the strong background noise and the fault information representation is not obvious, and can effectively extract the fault characteristics in the motor bearing vibration signal, with high engineering application value.

[0130] Example 3

[0131] The embodiment uses the bearing data of Case Western Reserve University (CWRU) in the United States. The CWRU bearing data experimental platform is as follows: Figure 2 As shown. The experimental platform includes a 1492W motor, a torque sensor, a power meter and electronic control equipment. In the experiment, an acceleration sensor is used to collect vibration signals, and the sensor is placed on the motor housing by using a magnetic base. The rolling bearing model used is 6205-2RSJEMSKF. The test speed is 1797r / min, and the signal sampling frequency is 12kHz. The vibration signals of the outer ring and inner ring faults of the rolling bearing are selected for experimental analysis. The theoretical fault frequency of the outer ring of the rolling bearing BPFO = 107.305Hz, and the theoretical fault frequency of the inner ring BPFI = 162.185Hz are calculated.

[0132] The specific steps are:

[0133] Step 1: Normalize the inner ring vibration signal. Figure 3 The time domain waveform of the bearing inner ring fault signal after standardization. Figure 4 The frequency spectrum can only identify the frequency 161.1HZ which is very close to BPFI, but cannot identify the frequency components of the characteristic frequency of the bearing fault. The time domain waveform and frequency domain waveform of the bearing inner ring fault signal are shown in the figure below. Figure 3 , Figure 4 shown.

[0134] Step 2: Figure 3 The time domain signal is processed by NCTVD and BAS, and the range of the regularization parameter λ and the convexity parameter a is selected according to the TVD parameter determination method. The two-dimensional space search trajectory diagram of the inner ring fault BAS algorithm optimization process is shown in Figure 5 As shown in the figure, the relationship between the fitness function value and the number of iterations is as follows Figure 6 As shown in the figure, the fitness function of the BAS algorithm reaches its maximum value after 70 iterations. At this time, the optimal combination of the regularization parameter λ and the convexity parameter a is [0.015, 10].

[0135] Step 3: Right Figure 3 The time domain signal is processed by NCTVD-BAS-TEO based on the optimal combination of regularization parameter λ and convexity parameter a, and the envelope spectrum of the inner circle signal is obtained as follows: Figure 7 As shown in the figure, there is an obvious peak at the theoretical value of the inner ring fault frequency, and there is also an obvious impact at 2 to 6 times the frequency. It can be judged that the inner ring fault has been clearly highlighted from the background noise and can be clearly distinguished.

[0136] Step 4: Standardize the outer ring vibration signal. Figure 8 The time domain waveform of the bearing outer ring fault signal after standardization. Fig. 9Only the frequency 108.4Hz, which is very close to BPFO, can be identified in the spectrum, and the frequency components of the characteristic frequency of the bearing fault cannot be identified. The time domain waveform and frequency domain waveform of the bearing outer ring fault signal are shown in the figure. Figure 8 , 9 shown.

[0137] Step 5: Right Figure 8 The time domain signal is processed by NCTVD and BAS, and the range of the regularization parameter λ and the convexity parameter a are selected according to the TVD parameter determination method. The two-dimensional space search trajectory diagram of the BAS algorithm optimization process for the outer ring fault is shown in Fig.10 As shown in the figure, the relationship between the fitness function value and the number of iterations is as follows Fig.11 As shown in the figure, the fitness function of the BAS algorithm reaches its maximum value after 70 iterations. At this time, the optimal combination of the regularization parameter λ and the convexity parameter a is [0.015, 10].

[0138] Step 6: Right Figure 8 The time domain signal is processed by NCTVD-BAS-TEO based on the optimal combination of regularization parameter λ and convexity parameter a, and the envelope spectrum of the outer ring signal is obtained as follows: Fig.12 As shown in the figure, there is an obvious peak at the theoretical value of the outer ring fault frequency, and there is also an obvious impact at 2 to 9 times the frequency. It can be judged that the outer ring fault has been clearly highlighted from the background noise and can be clearly distinguished.

[0139] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0140] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for extracting motor bearing fault features, characterized in that: The following steps are involved: Step 1: Introduce the arctangent penalty function into the total variation denoising TVD to enhance the algorithm-induced sparsity feature, define a new non-convex total variation denoising NCTVD, collect vibration signals for NCTVD processing and solve them using the MM algorithm; Step 2: Initialize the spatial position and parameters of BAS, define the value range of regularization parameter λ and convexity parameter a, and execute BAS optimization algorithm. By evaluating the fitness function CK_I, the optimal parameter combination [a,λ] is obtained; Step 3: The optimal parameter combination is input into NCTVD to obtain the optimal denoising signal, and then TEO is introduced for demodulation processing to obtain the Teager energy spectrum, extract the fault characteristic frequency, and then realize the diagnosis of the fault type.

2. A motor bearing fault feature extraction method according to claim 1, characterized in that: The specific steps of constructing non-convex total variation denoising in Step 1 are: Step 1.1: Collect the vibration signal x(n) of the motor fault bearing; Step 1.2: Find the first-order total variation of the vibration signal x(n); ； Where Dx represents the first-order differential, (p≥1) is the lp norm, where the matrix D is defined as follows: ； The tridiagonal matrix DDT is represented as: ； is the p-order norm, and It is expressed as: ； When p=1 or p=2, the first-order and second-order norms of the signal are expressed as: ； ； The minimum optimization objective function is: ； In the formula represents the signal fidelity term, >0 is the regularization parameter, the purpose of which is to find the signal x to minimize the objective function: ； Step 1.3: Calculate the arctan-NCTVD of the vibration signal x(n); The non-convex regularization term is the arctangent penalty function (arctan): ； Will After being transformed into a non-convex regularization term, the above formula becomes the following: ； Step 1.4: Use the MM algorithm to solve the above equation, and the result is expressed as: ； Step 1.5: Convexity condition; If the non-convex regularization term of the inverse tangent penalty function remains convex, the condition is: 。 3. A motor bearing fault feature extraction method according to claim 1, characterized in that: The specific steps of the BAS algorithm optimization in Step 2 are: Step 2.1: BAS initializes the position and parameters of the longicorn; The BAS algorithm is based on the beetle's continuous comparison of the odor concentrations of its left and right antennae. Before calculating the odor concentrations of its two antennae, a series of preparations are required. The position of the beetle in the D-dimensional space is , the positions of the left and right antennae of the longicorn beetle are defined as the model shown in the following formula: ； In the above formula, l represents the distance between the centroid of the longhorn beetle and its antennae, It indicates that the random unit vector guides the random orientation of the longhorn beetle in the solution space, and it needs to be normalized: ； Step2.2: Construct fitness function KC_I; Weighted kurtosis (KC_I) is an indicator for measuring the impact characteristics of the signal. It is often used in bearing fault diagnosis because it retains the characteristics of kurtosis and also includes the characteristics of correlation functions. Its calculation formula is: ； ； ； r is an adjustable positive real number used to control the fidelity of the output signal. r is generally set to 2; Step 2.3: Find the optimal parameter combination; The difference in odor concentration perceived by the left and right antennae is defined as the fitness function in the algorithm. The larger the odor concentration, the higher the fitness. The beetle constantly calculates the size of the fitness function to continue moving forward and moves towards the position with higher fitness. The beetle determines the next position according to the following formula: ； In the formula, t represents the current number of iterations; f( ) represents the fitness function; It represents the exploration step size at the tth iteration, and the sign() function is the sign function; After the iteration conditions are met, the optimal parameter combination [abest, λbest] is output.

4. A motor bearing fault feature extraction method according to claim 1, characterized in that: The specific steps of extracting fault features in Step 3 are as follows: Step 3.1: The optimal parameter combination is input into NCTVD to obtain the optimal denoised signal; Step 3.2: Introduce TEO for demodulation to obtain the Teager energy spectrum; Step 3.3: Use the motor fault bearing parameters collected before the experiment to calculate the theoretical fault characteristic frequency, and compare it with the experimental fault characteristic frequency extracted from the Teager energy spectrum to achieve fault diagnosis.

5. An application of the motor bearing fault feature extraction method as described in any one of claims 1 to 4 in the field of mechanical fault diagnosis and signal processing technology.

Images