Gear early fault diagnosis method based on PSO-MCKD

Abstract

The invention belongs to the technical field of electric vehicle speed reducers, and relates to a PSO-MCKD-based gear early fault diagnosis method, which comprises the following steps: firstly, adaptively optimizing parameters L and T in an MCKD algorithm by applying PSO for an acquired gear early fault signal containing a relatively weak gear early fault signal; based on the MCKD algorithm after parameter optimization, impact component enhancement is carried out on the early fault signal; thirdly, analyzing the enhanced signal by using an envelope spectrum analysis method, and comparing the enhanced signal with a theoretical fault feature; and finally, outputting a diagnosis result. According to the method, the problem that the parameters L and T of the MCKD cannot be optimal at the same time is solved, adaptive optimization is carried out on the parameters of the MCKD by adopting PSO, the problem that early fault features of the gear cannot be accurately extracted is solved, noise reduction is carried out on an original signal by using the optimized MCKD, and envelope spectrum demodulation is carried out to extract the fault features, so that the accuracy of the fault feature extraction is improved. And the efficiency and accuracy of early weak fault feature diagnosis of the gear are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of electric vehicle reducers, and in particular relates to a gear early fault diagnosis method based on PSO-MCKD. Background Art

[0002] Gears are one of the key components in electric vehicle reducers. At present, new energy vehicles are developing towards high speed and high torque. The working environment of gears is becoming increasingly harsh, and it is difficult to avoid failures. If early weak faults are not identified in time, the working life of the entire system will be affected. However, since the impact caused by early gear failures is relatively weak, it is difficult to accurately extract it from the collected original vibration signal. Therefore, finding an effective way to reduce the noise of the original vibration signal and enhance the weak fault characteristics of the gear is the top priority for early gear fault diagnosis. Summary of the invention

[0003] The purpose of the present invention is to introduce a particle swarm optimization algorithm (PSO) and provide a gear early fault diagnosis method based on PSO-MCKD to solve the problem that the parameters L and T of MCKD cannot reach the optimal state at the same time and cannot accurately extract the early fault characteristics of gears; this method can find the optimal model parameters for MCKD to extract early gear faults.

[0004] The objective of the present invention is achieved through the following technical solutions:

[0005] A gear early fault diagnosis method based on PSO-MCKD includes the following steps:

[0006] A. Use a vibration acceleration sensor to collect vibration data of the reducer throughout its life cycle to obtain the original vibration acceleration data;

[0007] B. Determine the optimization range of MCKD parameters L and T, initialize PSO parameters, and start optimizing parameters L and T, updating the speed and position of particles until the maximum number of iterations is reached or the optimal parameter combination is obtained;

[0008] C. Assign the optimal MCKD parameter combination (L, T) suitable for gear early fault vibration signal processing obtained by PSO adaptive optimization to MCKD, and obtain the processed output signal;

[0009] D. Perform envelope spectrum demodulation on the processed fault signal, extract the fault characteristics and compare them with the calculated theoretical fault frequency, and analyze the diagnosis results.

[0010] Furthermore, the vibration acceleration signal described is a radial vibration signal of the shaft; wherein the signal is obtained by acquiring Y channel data of a three-axis vibration acceleration sensor.

[0011] Furthermore, when using MCKD for processing, the value of the deconvolution period T is not fixed, but a range centered on the theoretical value of T. When only one theoretical value of T is used to process the signal by MCKD, the processing result may be inaccurate. The range of the deconvolution period T can be calculated as follows:

[0012]

[0013] In the formula, f g is the vibration signal sampling frequency; f pit is the fluctuation range of fault characteristic frequency. The optimization range of filter length L is 100~1000.

[0014] Furthermore, PSO is a global random adaptive optimization algorithm that updates itself and the group's optimal position during the search. The group can search for the best position through the mutual cooperation of each particle. If there is a particle group with N particles in the D-dimensional space, then X i and V i Respectively represent the position vector and velocity vector of the i-th particle; use p i and g i They represent the local optimal position of the ith particle in the space and the global optimal position of the entire group respectively; the ith particle can update its speed and position through the following formula.

[0015]

[0016] Where i = 1, 2, ..., N; c1 and c2 are the individual factor and group factor in the optimization process respectively; r1 and r2 are random numbers in the range of 0 to 1; w is the inertia weight; u is the number of iterations;

[0017] c1 means the acceleration weight of the particle to find the best position of the individual based on its own experience, and c2 means the acceleration weight of the particle to find the best position of the group based on the experience of other particles in the group; c1 and c2 are 0 to 2, c1 is 0 and it is easy to fall into the local optimum, and c2 is 0, which will cause the algorithm to converge slowly; the inertia weight w represents the ability of the particle to explore the space, and takes values ​​between 0.4 and 2; the number of iterations u is 10 to 30.

[0018] Furthermore, we need to determine the particle swarm size N and the particle position and speed limit X max and V max , and the position vector X i and the velocity vector V i Cannot exceed X max With V maxThe particle swarm size N is generally between 20 and 200. The more complex the problem, the higher the value of N should be. However, if N is too large, it will waste computing resources, and if it is too small, it will fall into the local optimum. The maximum speed V max Setting it too large will lead to going beyond the optimal position, while setting it too small will also lead to falling into the local optimum. When initializing parameters, the balance between the calculation accuracy and calculation efficiency of actual engineering applications should be considered. Parameters including population size, particle optimization range, maximum number of iterations, optimization speed and error range should be set according to working conditions.

[0019] Furthermore, to determine whether the result of PSO’s optimization of MCKD adaptive parameters is the optimal solution, it is necessary to determine a fitness function and use the output value of the function as the basis for judgment; from the vibration signal impact theory, it can be seen that the more the number of periodic characteristic impacts appearing in the vibration signal, the greater the impact peak of the characteristic frequency after the signal is analyzed by envelope spectrum; considering the periodicity and intensity of the signal, the dimensionless envelope spectrum peak factor E is used sp As a fitness function, this indicator can be understood as the ratio of the maximum value and the effective value of the envelope spectrum peak within the target frequency range. The calculation formula is as follows:

[0020]

[0021] Where f(n) (n=1,2,3,…,N) is the envelope spectrum peak value within the fault frequency range.

[0022] If the parameters of L and T obtained after PSO adaptive optimization are accurate, the processed signal will contain obvious periodic pulse components, and the fitness function value E sp The larger the value, the larger the fitness function value E sp The maximum value of is used to determine whether the optimal result is obtained.

[0023] Furthermore, the output signal after MCKD processing using the optimal parameters is a time domain signal after noise reduction and feature enhancement.

[0024] Furthermore, the described gear early fault characteristic frequencies specifically include gear meshing frequency, gear rotation frequency, gear fault frequency and related side frequencies.

[0025] Furthermore, the envelope spectrum transform of the experimental time domain signal can be defined as:

[0026]

[0027] In the formula, t is time and τ is a certain moment.

[0028] Furthermore, after the envelope spectrum analysis, what is obtained is the envelope spectrum diagram of the time domain signal. It can be determined whether the characteristic frequencies of early gear failure, including the gear meshing frequency, gear rotation frequency, gear failure frequency and the characteristic amplitudes of related side frequencies and their respective integer multiples, can be observed on the spectrum diagram. If so, the gear fault information is determined according to the characteristic category. If not, normal information is output, wherein the diagnosis result includes fault prompt information and the normal prompt information.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The MCKD method is able to extract periodic pulses from low signal-to-noise ratio signals and applied to early gear fault diagnosis. The method used can effectively reduce the background noise and enhance the characteristics of the experimental signal, overcoming the shortcoming that the early fault characteristics cannot be accurately and quickly identified from the vibration signal due to the weak characteristics;

[0031] 2. The adaptive optimization algorithm PSO is introduced to perform global adaptive optimization on the main parameters of MCKD, overcoming the problem of poor or distorted noise reduction and signal enhancement effects of the MCKD algorithm caused by artificially selected parameters, thus improving signal processing efficiency;

[0032] 3. By utilizing the characteristic of envelope spectrum analysis that is sensitive to the impact components in the signal, the time domain signal after noise reduction and enhancement is processed to make the extraction of fault characteristic frequency more significant and improve the diagnostic accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 It is a flowchart of the steps of the gear early fault diagnosis method based on PSO-MCKD of the present invention;

[0035] Figure 2 Iterative curve for optimizing MCKD for PSO;

[0036] Figure 3-Figure 4 This is the time domain image before and after MCKD treatment;

[0037] Figure 5-Figure 6 This is the envelope spectrum after envelope spectrum analysis of the original data before and after MCKD. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with embodiments:

[0039] The present invention provides a gear early fault diagnosis method based on PSO-MCKD. For the collected gear early fault signal containing relatively weak gear early fault, the parameters L and T in the MCKD algorithm are first optimized by PSO adaptively; then the impact component of the early fault signal is strengthened based on the MCKD algorithm after the optimized parameters; then, the enhanced signal is analyzed by the envelope spectrum analysis method and compared with the theoretical fault characteristics; finally, the diagnosis result is output. The present invention overcomes the problem that the parameters L and T of MCKD cannot reach the optimum at the same time, adopts PSO to adaptively optimize the parameters of MCKD, overcomes the problem that the gear early fault characteristics cannot be accurately extracted, uses the optimized MCKD to reduce the noise of the original signal and performs envelope spectrum demodulation to extract the fault characteristics, and improves the efficiency and accuracy of the diagnosis of the weak early fault characteristics of the gear.

[0040] Step 1: Use a vibration acceleration sensor to collect vibration data of the reducer throughout its life cycle to obtain original vibration acceleration data;

[0041] Step 2: Determine the optimization range of MCKD parameters L and T, initialize PSO parameters, and start optimizing parameters L and T, updating the speed and position of particles until the maximum number of iterations is reached or the optimal parameter combination is obtained;

[0042] Step 3: assign the optimal MCKD parameter combination (L, T) suitable for gear early fault vibration signal processing obtained by PSO adaptive optimization to MCKD, and obtain the processed output signal;

[0043] Step 4: Perform envelope spectrum demodulation on the processed fault signal, extract the fault characteristics and compare them with the calculated theoretical fault frequency, and analyze the diagnosis results.

[0044] Optionally, the vibration acceleration signal described is a radial vibration signal of the shaft; wherein the signal is obtained by acquiring Y channel data of a three-axis vibration acceleration sensor.

[0045] Optionally, when using MCKD for processing, the value of the deconvolution period T is not fixed, but a range centered on the theoretical value of T. When only one theoretical value of T is used to process the signal by MCKD, the processing result may be inaccurate. The range of the deconvolution period T can be calculated as follows:

[0046]

[0047] In the formula, f g is the vibration signal sampling frequency; f pitis the fluctuation range of fault characteristic frequency. The shorter the filter length L is, the worse the filtering effect is, but the faster the calculation speed is. Therefore, under the premise of ensuring the calculation speed, a larger L value should be selected as much as possible. Generally, the optimization range of L is 100 to 1000.

[0048] Optionally, PSO is a global random adaptive optimization algorithm. When searching, the optimal position of itself and the group is updated. The group can search for the best position through the cooperation of each particle. If there is a particle group with N particles in the D-dimensional space, X can be used. i and V i denote the position vector and velocity vector of the ith particle respectively. i and g i They represent the local optimal position of the ith particle in the space and the global optimal position of the entire group respectively. The velocity and position of the ith particle can be updated by the following formula.

[0049]

[0050] Where i=1,2,…,N; c1 and c2 are the individual factor and group factor in the optimization process respectively; r1 and r2 are random numbers in the range of 0 to 1; w is the inertia weight; and u is the number of iterations.

[0051] c1 is the acceleration weight of the particle to find the best position for the individual based on its own experience, and c2 is the acceleration weight of the particle to find the best position for the group based on the experience of other particles in the group. c1 and c2 are generally set to 0-2. c1 is 0, which is easy to fall into the local optimum, and c2 is 0, which will cause the algorithm to converge slowly. The inertia weight w represents the ability of the particle to explore the space, preventing the particle from falling into the local optimum, and takes a value between 0.4 and 2. If the number of iterations u is too small, it will be unstable, but if it is too large, it will waste computing resources. It is generally 10-30.

[0052] Also determine the particle group size N and the particle position and speed limit X max and V max , and the position vector X i and the velocity vector V i Cannot exceed X max With V max The particle swarm size N is generally between 20 and 200. The more complex the problem, the higher the value of N should be. However, if N is too large, it will waste computing resources, and if it is too small, it will fall into the local optimum. The maximum speed V max Setting it too large will lead to going beyond the optimal position, while setting it too small will also lead to falling into the local optimum. When initializing parameters, the balance between the calculation accuracy and calculation efficiency of actual engineering applications should be considered. According to the working conditions, parameters including population size, particle optimization range, maximum number of iterations, optimization speed and error range should be set.

[0053] Optionally, to determine whether the result of PSO's optimization of MCKD adaptive parameters is the optimal solution, it is necessary to determine a fitness function and use the output value of the function as the basis for judgment. According to the vibration signal impact theory, the more the number of periodic characteristic impacts appearing in the vibration signal, the greater the impact peak of the characteristic frequency after the signal is analyzed by envelope spectrum. At the same time, considering the periodicity and intensity of the signal, the dimensionless envelope spectrum peak factor E is used. sp As a fitness function, this indicator can be understood as the ratio of the maximum value and the effective value of the envelope spectrum peak within the target frequency range. The calculation formula is as follows:

[0054]

[0055] Where f(n) (n=1,2,3,…,N) is the envelope spectrum peak value within the fault frequency range.

[0056] If the parameters of L and T obtained after PSO adaptive optimization are accurate, the processed signal will contain obvious periodic pulse components, and the fitness function value E sp The larger the value, the larger the fitness function value E sp The maximum value of is used to determine whether the optimal result is obtained.

[0057] Optionally, the signal outputted after MCKD processing using optimal parameters is a time domain signal after noise reduction and feature enhancement.

[0058] Optionally, the described gear early fault characteristic frequencies specifically include gear meshing frequency, gear rotation frequency, gear fault frequency and related side frequencies.

[0059] Optionally, envelope spectrum analysis is more sensitive to the impact in the signal, and the amplitude of the fault characteristic frequency of the signal in the envelope spectrum is higher. At the same time, envelope spectrum analysis can eliminate part of the signal interference, making the extraction effect of the fault characteristic frequency more significant. Therefore, the envelope spectrum is used in the analysis of the experimental signal below, and the envelope spectrum transformation of the experimental time domain signal can be defined as:

[0060]

[0061] In the formula, t is time and τ is a certain moment.

[0062] Optionally, after the envelope spectrum analysis, what is obtained is an envelope spectrum diagram of the time domain signal. It can be determined whether the characteristic frequencies of early gear failure, including the gear meshing frequency, gear rotation frequency, gear failure frequency and characteristic amplitudes of related side frequencies and their respective integer multiples, can be observed on the spectrum diagram. If so, the gear fault information is determined according to the characteristic category. If not, normal information is output, wherein the diagnosis result includes fault prompt information and the normal prompt information.

[0063] Example 1

[0064] A gear early fault diagnosis method based on PSO-MCKD, the specific process is as follows Figure 1 As shown, the specific steps include the following steps.

[0065] 1. Use vibration signal acquisition equipment to collect and save vibration acceleration signals in real time. The sensor uses a piezoelectric sensor. Due to the complex internal structure of the reducer and the multiple transmission paths of vibration signals, the sensor should be installed in a position with high box rigidity and low damping.

[0066] 2. Use PSO to adaptively optimize the parameters L and T of MCKD and determine the best parameter combination. The MCKD method improves the ability to reduce noise and extract shocks within the fault cycle range. At the same time, the definition of correlation kurtosis is proposed and used as the termination condition of the optimal inverse filter selection process. This method can be applied to gear fault diagnosis.

[0067] Assume that the fault signal when the gear transmission fails is:

[0068] y=gx+η

[0069] Where y is the vibration signal of the gear fault experiment; x is the periodic impact component of the fault; g is the signal transfer function of the transmission path in the gear transmission system; η is the noise.

[0070] The formula for the associated kurtosis is defined as follows:

[0071]

[0072] Where N is the number of input signal sampling points; T is the deconvolution period; y n is the vibration signal after zero mean processing; y n -m t nm T The vibration signal at the time.

[0073] The goal of MCKD is to find the filter f that maximizes the correlation kurtosis of the collected gear fault signal y, so as to extract the fault impulse component x from the collected fault signal y, that is:

[0074]

[0075] In the formula, are the filter coefficients and L is the length of the filter.

[0076] The filter coefficient f corresponding to the maximum correlation kurtosis k , can be obtained through CK M (T) for f kTake the derivative and get:

[0077]

[0078] In the formula, k = 1, 2, ..., L. So it can be sorted into a matrix form:

[0079]

[0080] According to the above formula, the implementation steps of the MCKD algorithm can be summarized as follows:

[0081] (1) Determine the deconvolution period T, the displacement number M and the filter length L, and initialize the filter f;

[0082] (2) Calculate X0, X0 T and X mT ;

[0083] (3) Calculate the filter f, and use the filter f to filter the signal x and output y;

[0084] (4) Update the filter coefficients and determine whether the conditions are met or the number of iterations is reached based on the change in the correlation kurtosis before and after filtering ΔCKM(T). Otherwise, return to step (3).

[0085] The displacement number M can be calculated by an empirical formula, and the selection of the filter length L and the deconvolution period T requires an optimization algorithm to perform adaptive optimization selection so that they can achieve the optimal value at the same time.

[0086] When the rotating machinery runs stably at a certain input speed, relevant parameters such as the rotation frequency of each rotating component, the meshing frequency of the gears, and the corresponding fault characteristic frequency can be known. However, under actual working conditions, due to factors such as slight fluctuations in the input speed, various errors, wear and even failures caused by long-term operation, the frequency of concern is not completely consistent with the theoretical value. Therefore, when using MCKD for processing, the value of the deconvolution period T is not fixed, but a range centered on the theoretical value of T. When only one theoretical value of T is used to perform MCKD processing on the signal, the processing result may not be accurate. The range of the deconvolution period T can be calculated as follows:

[0087]

[0088] In the formula, f g is the vibration signal sampling frequency; f pit is the fault characteristic frequency fluctuation range.

[0089] According to CK MFrom the calculation formula of (T), it can be seen that the larger the value of M, the more pulses there are after deconvolution, and the stronger the fault feature extraction capability is. However, if the value of M is too large, data distortion will occur. The value of M is selected based on the number of sampling points a:

[0090]

[0091] In MCKD, there is an interaction between the two key parameters, L and T. Therefore, the key points of MCKD parameter selection are as follows: First, find the optimal parameter T within the given parameter range. Second, find the matching parameter L while optimizing parameter T. In this way, both parameters can be optimized at the same time, and the application effect of the MCKD algorithm will be the best.

[0092] 3. PSO can overcome the influence of subjective factors, search for the optimal parameters L and T in a given search space, and improve the efficiency and accuracy of fault diagnosis. However, it is also necessary to determine a fitness function and use the output value of the function as the basis for judgment. According to the vibration signal impact theory, the more the number of periodic characteristic impacts appearing in the vibration signal, the greater the impact peak of the characteristic frequency after the signal is analyzed by envelope spectrum. At the same time, considering the periodicity and intensity of the signal, the dimensionless envelope spectrum peak factor E is used. sp As a fitness function, this indicator can be understood as the ratio of the maximum value and the effective value of the envelope spectrum peak within the target frequency range. The calculation formula is as follows:

[0093]

[0094] Where f(n) (n=1,2,3,…,N) is the envelope spectrum peak value within the fault frequency range.

[0095] If the parameters of L and T obtained after PSO adaptive optimization are accurate, the processed signal will contain obvious periodic pulse components, and the fitness function value E sp The larger the value, the larger the fitness function value E sp The maximum value of is used to determine whether the optimal result is obtained.

[0096] So Figure 1 As shown in the figure, the specific implementation process of early gear fault diagnosis based on PSO-MCKD is as follows:

[0097] (1) Input the original vibration signal. Determine the value of the MCKD parameter M and the optimal range of L and T;

[0098] (2) Initialize PSO parameters and select the envelope spectrum peak factor as the fitness function. When initializing parameters, the balance between the calculation accuracy and calculation efficiency of actual engineering applications should be considered. According to the working conditions, parameters such as population size, particle optimization range, maximum number of iterations, optimization speed and error range should be set;

[0099] (3) PSO starts to search for the best result and compares and evaluates it with the fitness function value, and updates the particle speed and position accordingly until the maximum number of iterations is reached or the optimal parameter combination is obtained;

[0100] (4) The optimal MCKD parameter combination (L, T) suitable for gear vibration signal processing obtained by PSO adaptive optimization is assigned to MCKD, and the processed output signal is obtained;

[0101] (5) Perform envelope spectrum demodulation on the processed fault time domain signal, extract fault features and analyze them;

[0102] (6) Based on the extracted characteristic frequency information, the theoretical fault characteristics of the gear are compared to diagnose the early fault type of the gear.

[0103] 4. The particle swarm optimization algorithm parameters of this embodiment set the population size N to 30, the learning factors c1 and c2 to 1.5 and 2 respectively, the inertia weight w ranges from 0.4 to 0.9, the number of iterations u is 30, and the speed V ranges from -20 to 20. Before starting to apply PSO to optimize the parameters L and T of MCKD, it is also necessary to set the value of M and the optimization range of L and T. The value of M is calculated to be 7, the optimization range of L is set to 100 to 1000, and the optimization range of T is calculated to be 20 to 110. After determining the values ​​and ranges of each parameter, start the optimization. Figure 2 As shown in the figure, it is the iteration curve of PSO optimizing MCKD. The algorithm converges at the 18th generation, and the optimal parameter combination is (725,93). The parameters are input into the MCKD algorithm for analysis. Figure 3 and Figure 4 The time domain diagrams before and after MCKD processing respectively show the effectiveness of this method.

[0104] 5. Perform envelope spectrum analysis on the time domain signal after MCKD processing. Figure 5 and Figure 6 The envelope spectrum of the original data before and after MCKD is analyzed, and the diagnosis is performed based on the calculated theoretical gear fault characteristic frequency. The envelope spectrum obtained from the time domain signal without MCKD processing cannot accurately extract the early gear fault characteristics and is submerged by noise. Figure 6 It can be seen that the characteristic frequency of gear fault is 130.9Hz. Figure 4It can also be seen that the obvious fault impact cycle is 0.008s, and the inverse of the cycle corresponds to the fault characteristic frequency. A variable frequency band with an interval of the gear fault characteristic frequency also appears clearly near the meshing frequency, which can be judged that the gear has an early crack fault.

[0105] Table 1 Gear fault characteristic frequency table

[0106]

[0107] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A gear early fault diagnosis method based on PSO-MCKD, characterized in that: The following steps are involved: A. Use a vibration acceleration sensor to collect vibration data of the reducer throughout its life cycle to obtain the original vibration acceleration data; B. Determine the optimization range of MCKD parameters L and T, initialize PSO parameters, and start optimizing parameters L and T, updating the speed and position of particles until the maximum number of iterations is reached or the optimal parameter combination is obtained; C. Assign the optimal MCKD parameter combination (L, T) suitable for gear early fault vibration signal processing obtained by PSO adaptive optimization to MCKD, and obtain the processed output signal; D. Perform envelope spectrum demodulation on the processed fault signal, extract the fault characteristics and compare them with the calculated theoretical fault frequency, and analyze the diagnosis results.

2. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: The vibration acceleration signal described is a radial vibration signal of the shaft; wherein the signal is obtained by acquiring Y channel data of a three-axis vibration acceleration sensor.

3. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized in that: When using MCKD for processing, the value of the deconvolution period T is not fixed, but a range centered on the theoretical value of T. When only one theoretical value of T is used to process the signal for MCKD processing, the processing result may be inaccurate. The range of the deconvolution period T can be calculated as follows: T＝ f g f pit In the formula, f g is the vibration signal sampling frequency; f pit is the fluctuation range of fault characteristic frequency. The optimization range of filter length L is 100~1000.

4. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: PSO is a global random adaptive optimization algorithm. When searching, it updates the optimal position of itself and the group. The group can search for the best position through the cooperation of each particle. If there is a particle group with N particles in the D-dimensional space, X can be used to i and V i Respectively represent the position vector and velocity vector of the i-th particle; use p i and g i They represent the local optimal position of the ith particle in the space and the global optimal position of the entire group respectively; the ith particle can update its speed and position through the following formula. Where i = 1, 2, ..., N; c1 and c2 are the individual factor and group factor in the optimization process respectively; r1 and r2 are random numbers in the range of 0 to 1; w is the inertia weight; u is the number of iterations; c1 means the acceleration weight of the particle to find the best position of the individual based on its own experience, and c2 means the acceleration weight of the particle to find the best position of the group based on the experience of other particles in the group; c1 and c2 are 0 to 2, c1 is 0 and it is easy to fall into the local optimum, and c2 is 0, which will cause the algorithm to converge slowly; the inertia weight w represents the ability of the particle to explore the space, and takes values ​​between 0.4 and 2; the number of iterations u is 10 to 30.

5. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: Also determine the particle group size N and the particle position and speed limit X max and V max , and the position vector X i and the velocity vector V i Cannot exceed X max With V max The particle swarm size N is generally between 20 and 200. The more complex the problem, the higher the value of N should be. However, if N is too large, it will waste computing resources, and if it is too small, it will fall into the local optimum. The maximum speed V max Setting it too large will lead to going beyond the optimal position, while setting it too small will also lead to falling into the local optimum. When initializing parameters, the balance between the calculation accuracy and calculation efficiency of actual engineering applications should be considered. Parameters including population size, particle optimization range, maximum number of iterations, optimization speed and error range should be set according to working conditions.

6. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: To determine whether the result of PSO's optimization of MCKD adaptive parameters is the optimal solution, it is necessary to determine a fitness function and use the output value of the function as the basis for judgment; from the vibration signal impact theory, it can be seen that the more the number of periodic characteristic impacts appearing in the vibration signal, the greater the impact peak of the characteristic frequency after the signal is analyzed by envelope spectrum; considering the periodicity and intensity of the signal, the dimensionless envelope spectrum peak factor E is used sp As a fitness function, this indicator can be understood as the ratio of the maximum value and the effective value of the envelope spectrum peak within the target frequency range. The calculation formula is as follows: Where f(n) (n=1,2,3,…,N) is the envelope spectrum peak value within the fault frequency range. If the parameters of L and T obtained after PSO adaptive optimization are accurate, the processed signal will contain obvious periodic pulse components, and the fitness function value E sp The larger the value, the larger the fitness function value E sp The maximum value of is used to determine whether the optimal result is obtained.

7. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: The output signal after MCKD processing with the optimal parameters is a time domain signal after noise reduction and feature enhancement.

8. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: The characteristic frequencies of early gear failure described specifically include gear meshing frequency, gear rotation frequency, gear failure frequency and related side frequencies.

9. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: The envelope spectrum transform of the experimental time domain signal can be defined as: In the formula, t is time and τ is a certain moment.

10. The gear early fault diagnosis method based on PSO-MCKD according to claim 1 is characterized by: After envelope spectrum analysis, what is obtained is the envelope spectrum diagram of the time domain signal. It can be determined whether the characteristic frequencies of early gear failure can be observed on the spectrum diagram, including the gear meshing frequency, gear rotation frequency, gear failure frequency and the characteristic amplitudes of related side frequencies and their respective integer multiples. If so, the gear fault information is determined according to the characteristic category. If not, normal information is output, where the diagnosis result includes fault prompt information and the normal prompt information.

Images