A data-driven hybrid eccentricity fault diagnosis method for outer-rotor hub permanent magnet motors

Abstract

A data-driven method for diagnosing hybrid eccentricity faults of an outer rotor hub permanent magnet motor relates to a method for diagnosing motor faults, and includes the following steps: 1) establishing an analytical model of open-circuit back-electromotive force of a hybrid eccentricity of an outer rotor hub permanent magnet motor; 2) constructing a large-sample hybrid eccentricity fault signal database based on the analytical model; 3) extracting the features of the hybrid eccentricity fault signal based on fast Fourier transform; 4) establishing a hybrid eccentricity fault diagnosis model based on a BP neural network; 5) eliminating interference faults; 6) diagnosing the hybrid eccentricity fault parameters of the outer rotor hub permanent magnet motor to be tested. The method proposed in the present invention is based on data-driven, and uses a parameterized analytical model to efficiently construct a large-sample hybrid eccentricity fault signal database, and can diagnose the hybrid eccentricity fault parameters of the outer rotor hub permanent magnet motor with high precision and non-invasively.

Description

Technical Field

[0001] The present invention relates to a method for diagnosing motor faults, and more specifically to a method for diagnosing hybrid eccentricity faults of an outer-rotor hub permanent magnet motor based on data driving. Background Art

[0002] Due to characteristics such as high integration and flexible control, the form of hub permanent magnet motor drive is considered an ideal choice for future transportation vehicles. The outer-rotor hub permanent magnet motor is widely used in the field of electric bicycles and will gradually be applied to distributed drive electric vehicles.

[0003] The outer-rotor hub permanent magnet motor has a high risk of hybrid eccentricity faults. On the one hand, the outer rotor and inner stator of the hub motor are constantly subjected to impacts from road unevenness and alternating loads of the vehicle body. On the other hand, manufacturing and assembly errors will also cause a certain degree of eccentricity faults in the motor. Severe hybrid eccentricity faults may cause the motor to wear out and be scrapped, endangering the driving safety of the vehicle. Therefore, there is an urgent need for a method for diagnosing hybrid eccentricity faults of outer-rotor hub permanent magnet motors.

[0004] Three fault parameters are required to completely describe the hybrid eccentricity fault: the static eccentricity circumferential angle γ0, the static eccentricity ratio ε s and the dynamic eccentricity ratio ε d , that is, (γ0, ε s , ε d ). The value range of γ0 is 0° ≤ γ0 ≤ 360°, and the value range of ε s is 0% ≤ ε s ≤ 100%, and the value range of ε d is 0% ≤ ε d ≤ 100%. Since the potential combination number of these three parameters reaches at least tens of thousands, the characteristics of hybrid eccentricity faults are extremely complex. The data-driven method is suitable for diagnosing such hybrid eccentricity faults with complex fault parameters. Through comprehensive research on the existing technology, the existing methods for diagnosing motor hybrid eccentricity faults mainly have the following problems:

[0005] (1) It is difficult for the existing technology to simultaneously diagnose the hybrid eccentricity fault parameters (γ0, ε s , ε d ). Most of the existing methods for diagnosing hybrid eccentricity faults only focus on fault classification, such as CN113659908A, CN112924090A, CN107091986A, and CN106602797A, or only diagnose the static eccentricity ratio and the dynamic eccentricity ratio, such as CN113659908A, CN112924090A, CN107091986A, CN102262215A, and CN106602797A, ignoring the influence of the static eccentricity circumferential angle γ0 in the hybrid eccentricity fault parameters on the static eccentricity ratio εs and the dynamic eccentricity ε d The coupling effect makes it impossible to detect the static eccentricity direction.

[0006] (2) It is difficult for the existing technologies to non-invasively diagnose the hybrid eccentricity fault parameters (γ0, ε s , ε d ). Some technologies have achieved the simultaneous diagnosis of three fault parameters, namely the static eccentricity circumferential angle, the static eccentricity, and the dynamic eccentricity, such as CN109541461A and CN114295979A. However, their fault signals are all obtained through invasive sensors. This method requires additional sensors to be installed inside the motor in advance, which will increase the manufacturing and maintenance costs of the motor.

[0007] (3) The existing data-driven hybrid eccentricity fault diagnosis methods cannot efficiently establish a large-sample fault signal database. To improve the diagnostic accuracy of the data-driven hybrid eccentricity fault diagnosis methods, a database with a large sample is generally urgently needed. The existing methods mainly rely on the finite element method or the experimental method to obtain the database, such as CN114295979A and CN107091986A. However, building a database using the finite element method is very time-consuming; building a database using the experimental method is not only time-consuming but also very costly. This results in the diagnostic methods based on the finite element method and the experimental method generally having the disadvantages of a small sample size, low sample diversity, low database construction efficiency, and weak generalization ability of the diagnostic model, and thus cannot accurately diagnose the static eccentricity circumferential angle, the static eccentricity, and the dynamic eccentricity. Summary of the Invention

[0008] The purpose of the present invention is to solve the problems that it is difficult for the existing technologies to simultaneously and non-invasively diagnose three hybrid eccentricity fault parameters (γ0, ε s , ε d ) and that it is impossible to efficiently establish a large-sample fault signal database, and to provide a data-driven hybrid eccentricity fault diagnosis method for an outer-rotor hub permanent magnet motor.

[0009] The technical solution adopted by the present invention to solve the above-mentioned deficiencies of the existing technologies is as follows:

[0010] A data-driven hybrid eccentricity fault diagnosis method for an outer-rotor hub permanent magnet motor, characterized by comprising the following steps:

[0011] S1: Establish an analytical model of the open-circuit back electromotive force of the hybrid eccentricity of the outer-rotor hub permanent magnet motor;

[0012] S2: Construct a large-sample hybrid eccentricity fault signal database based on the analytical model;

[0013] S3: Extract the features of the hybrid eccentricity fault signal based on the fast Fourier transform;

[0014] S4: Establish a hybrid eccentricity fault diagnosis model based on BP neural network;

[0015] S5: Eliminate interference faults;

[0016] S6: Diagnose the mixed eccentricity fault parameters of the outer rotor hub permanent magnet motor under test.

[0017] In the above S1, the specific method for establishing the analytical model of the hybrid eccentric open-circuit back-EMF of the outer rotor hub permanent magnet motor is:

[0018] S1.1 Calculate the radial magnetic field and tangential magnetic field of the permanent magnet of the outer rotor hub permanent magnet motor without considering the slots, multiply them by the complex relative magnetic permeance considering the stator slots of the outer rotor hub permanent magnet motor, and then linearly superpose them to obtain the air gap radial magnetic flux density of the outer rotor hub permanent magnet motor, which is expressed as:

[0019]

[0020] In the formula, B mr is the radial magnetic flux density of the air gap of the outer rotor hub permanent magnet motor, B Mr_slotless and B Mt_slotless They are respectively the radial magnetic field and tangential magnetic field of the permanent magnet of the outer rotor hub permanent magnet motor without considering the slotting effect; a and λ b are the real and imaginary parts of the complex relative permeance considering the stator slot effect of the outer rotor hub permanent magnet motor; B PMrη and B PMtη are the ηth harmonic amplitudes of the radial magnetic field and tangential magnetic field of the permanent magnet of the outer rotor hub permanent magnet motor without considering the slotting effect, η is an odd number, indicating the spatial order of the permanent magnet magnetic field of the outer rotor hub permanent magnet motor; p is the number of permanent magnet pole pairs, ω is the mechanical angular velocity, θ is the spatial angle, θ0 is the initial phase angle; λ aμ and λ bμ are the harmonic amplitudes of the real and imaginary parts of the complex relative permeability, μ is the harmonic order, N λ is the maximum harmonic order; Q is the number of stator slots, r is the average radius of the air gap, and t is time.

[0021] S1.2 Calculation of the mixed eccentric permeability correction coefficient ε using the permeability correction coefficient method M , the expression is:

[0022]

[0023] in:

[0024]

[0025]

[0026] Wherein, e s and e d are the static eccentricity and the dynamic eccentricity respectively, and ε s and ε d represent the static eccentricity ratio and the dynamic eccentricity ratio respectively; δ is the air-gap length of the non-eccentric outer-rotor hub permanent magnet motor; γ0 is the static eccentricity circumferential angle.

[0027] S1.3 Calculate the hybrid eccentricity air-gap radial magnetic density B Mmr of the outer-rotor hub permanent magnet motor, and the expression is as follows:

[0028]

[0029] S1.4 Establish an analytical model of the hybrid eccentricity three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor, and the expression is as follows:

[0030]

[0031]

[0032]

[0033] Wherein, e A , e B , e C are the A-phase, B-phase, and C-phase open-circuit back electromotive forces of the outer-rotor hub permanent magnet motor respectively; A, B, and C represent the slot vector matrices of the A-phase, B-phase, and C-phase of the outer-rotor hub permanent magnet motor respectively; N Z is the number of turns of the conductor of the outer-rotor hub permanent magnet motor, L is the axial length, and R s is the outer diameter of the stator. When only a single conductor of the A-phase or B-phase or C-phase is contained in the i-th slot, k i =1; when two conductors of the A-phase or B-phase or C-phase are contained in the i-th slot, k i =2. The slots mentioned in the present invention all refer to the slots opened on the stator of the outer-rotor hub permanent magnet motor.

[0034] In the above-mentioned S2, it includes the definition of the hybrid eccentricity ratio ε m :

[0035] ε m = ε s+ ε d

[0036] The specific method of S2 is as follows:

[0037] S2.1 Set the parameter change step size, and list all combinations of hybrid eccentricity fault parameters (γ0, ε s , ε d ): The value range of γ0 is 0°≤γ0≤360°, and ε sThe value range of ε is 0% ≤ ε s ≤ 100%, where ε d The value range of ε is 0% ≤ ε d ≤ 100%; Take the step size |Δγ0|, and let γ0 = {|Δγ0|, 2|Δγ0|, 3|Δγ0|,..., a|Δγ0|,...}, where 0° ≤ a|Δγ0| ≤ 360° and a ∈ N * ; Take the step size |Δε s |, and let ε s = {|Δε s |, 2|Δε s |, 3|Δε s |,..., s|Δε s |,...}, where 0% ≤ s|Δε s | ≤ 100% and s ∈ N * ; Take the step size |Δε d |, and let ε d = {|Δε d |, 2|△ε d |, 3|Δε d |,..., d|Δε d |,...}, where 0% ≤ d|Δε d | ≤ 100% and d ∈ N * ;

[0038] S2.2 Eliminate the combined hybrid eccentricity fault parameters (γ0, ε m ≥ 100%); Because when ε s , ε d ) where ε m is greater than 100%, the stator and rotor invade each other.

[0039] S2.3 Eliminate the combined hybrid eccentricity fault parameters (γ0, ε s = 0 and γ0 ≠ 0); Because when the static eccentricity is 0, there is no static eccentricity fault, so there is no static eccentricity circumferential angle. s , ε d )

[0040] S2.4 Based on the analytical model of the hybrid eccentricity three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor obtained in S1, traverse the combined hybrid eccentricity fault parameters (γ0, ε s , ε d ) to analytically calculate the time-domain waveforms (e A , e B , e C ) of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor. Thus, obtain a large-sample hybrid eccentricity fault signal database.

[0041] The specific method of S3 is as follows:

[0042] S3.1 Through fast Fourier transform (FFT), perform amplitude-frequency characteristic analysis on the time-domain waveforms (e A , e B , e C ) of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor for each combination of hybrid eccentricity fault parameters, and extract the fundamental wave amplitudes e A_c , e B_c and e C_c of phases A, B, and C, and the average values e A_s , e B_s and e C_s of the left and right sideband harmonic amplitudes of phases A, B, and C, to obtain (e A_c , e B_c , e C_c , e A_s , e B_s , e C_s );

[0043] S3.2 Based on the relationship that the open-circuit back electromotive force is proportional to the rotational speed, eliminate the influence of the rotational speed n, as shown in the following two equations, to obtain the hybrid eccentricity fault index (e A_cn , e B_cn , e C_cn , e A_sn , e B_sn , e C_sn ):

[0044] e i_cn = e i_c / n

[0045] e i_sn = e i_s / e i_c

[0046] In the equations, e i_c represents the fundamental wave amplitude of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor for phases A, B, and C, and e i_cn represents the fundamental wave amplitude of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor after eliminating the influence of the rotational speed for phases A, B, and C, where i = A, B, C. e i_s represents the average value of the left and right sideband harmonic amplitudes of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor for phases A, B, and C, and e i_cn represents the average value of the left and right sideband harmonic amplitudes of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor after eliminating the influence of the rotational speed for phases A, B, and C, where i = A, B, C.

[0047] In S3.1 and S3.2 described above, the fundamental wave amplitude refers to the open-circuit back electromotive force at the fundamental wave frequency f cThe amplitude at [specific location], the left and right sideband harmonic amplitudes respectively refer to the amplitudes of the open - circuit back - electromotive force at the frequencies of the left and right sideband harmonics f l 、f y The fundamental frequency f c 、mechanical frequency f r and the left - sideband harmonic frequency f l 、right - sideband harmonic frequency f y are defined as follows:

[0048] When the rotational speed is n, after performing a fast Fourier transform on the time - domain signal of the open - circuit back - electromotive force within one mechanical period, the frequency - domain distribution of the open - circuit back - electromotive force is obtained. Among them, the fundamental frequency f c = np / 60, where p is the number of pole pairs of the outer - rotor hub permanent - magnet motor; the mechanical frequency f r = n / 60; the left - sideband harmonic frequency f l = f c - f r and the right - sideband harmonic frequency f y = f c + f r .

[0049] S3.3 Construct a hybrid eccentricity fault index database I and a hybrid eccentricity fault parameter database O that contain all combinations of hybrid eccentricity fault parameters. The database dimension N c = d max ×[(s max - 1)×a max / 2 + 1], where a max , s max and d mac represent the number of steps of the hybrid eccentricity fault parameters γ0, ε s and ε d respectively. The database is represented in matrix form as follows:

[0050]

[0051] The corresponding hybrid eccentricity fault parameter database O is represented as follows:

[0052]

[0053] The specific method of S4 is as follows:

[0054] S4.1 Construct a BP neural network model: Using the hybrid eccentricity fault index database I as the input and the hybrid eccentricity fault parameter library O as the output, establish a model with n i inputs, n o outputs, N h hidden layers, and the number of nodes in each hidden layer is N h1 、N h2 、……、N(Nh) ) of the BP neural network model, its overall structure is: i -N y1 -N y2 -……-N (Ny) -n o The transfer function of each hidden layer is tansig, the transfer function of the output layer is purelin, and the training function during back propagation is trainlm.

[0055] S4.2 Data normalization: Randomly select Nt corresponding groups of data in databases I and O as training sets and normalize them.

[0056] S4.3 Train the BP neural network model. Set the number of training times to n t , the learning rate is n l , the minimum error of the training target is n e , the maximum number of failures is n f , train the neural network. When the error is less than n e Stop training when .

[0057] S4.4 Verify the BP neural network model. c -N t The group data is used as a validation set to verify the diagnostic accuracy of the trained BP neural network. If the diagnostic error is within e t The following meets the usage requirements.

[0058] The specific method of S5 is:

[0059] S5.1 tests and determines whether the resistances of the three-phase windings of the outer rotor hub permanent magnet motor to be tested are equal. If they are not equal, the motor to be tested has a winding imbalance fault and the diagnosis process ends; if they are equal, the motor to be tested has no winding fault and the next step is executed.

[0060] S5.2 collects the open-circuit back electromotive force of the outer rotor hub permanent magnet motor unit to be tested;

[0061] The specific test method of S5.2 is as follows:

[0062] S5.2.1 Install the outer rotor hub permanent magnet motor to be tested on the test platform, and connect the three-phase voltage terminal wires of its unit motor to the oscilloscope.

[0063] S5.2.2 Start the reverse-drag motor and reverse-drag the outer rotor hub permanent magnet motor to be tested to a speed n.

[0064] S5.2.3 Use an oscilloscope to collect the three-phase open-circuit back-EMF signal of the inner and outer rotor hub permanent magnet motor in one mechanical cycle.

[0065] S5.3 Using the Fourier transform method of S3, extract the spectral information of the three-phase open-circuit back electromotive force of the outer rotor hub permanent magnet motor to be measured in step S5.2.3. First, diagnose whether there is an uneven magnetization harmonic component ε c ±f r with a frequency not equal to f s ±kf r , where k = 1, 2, 3... Then, to tolerate the interference caused by test noise, when the amplitude of the harmonic with a frequency of f c ±kf r is greater than or equal to half the amplitude of the mixed eccentricity harmonic f s ±f r , it is judged as an uneven magnetization harmonic. If the uneven magnetization harmonic component exists, the measured motor has an uneven magnetization fault, and the diagnostic process ends; if not, the measured motor has uniform magnetization, and the next step is executed.

[0066] S5.4 Use the method of S3 to extract the fault index of the outer rotor hub permanent magnet motor to be measured

[0067] I T =[e AT_cn , e BT_cn , e CT_cn , e AT_sn , e BT_sn , e CT_sn ;

[0068] In the formula, e iT_cn represents the fundamental wave amplitude of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor obtained by testing, excluding the influence of speed; e iT_sn represents the average value of the left and right sideband harmonic amplitudes of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor obtained by testing, excluding the influence of speed, where i = A, B, C.

[0069] The specific implementation method of S6 is as follows:

[0070] Substitute I T into the BP neural network model established in S4 to diagnose and obtain the mixed eccentricity parameters ME'(γ0, ε s , ε dd ).

[0071] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0072] (1) Using the open-circuit back electromotive force as the fault signal, three fault parameters of the mixed eccentricity fault can be diagnosed simultaneously and non-invasively: the static eccentricity circumferential angle γ0, the static eccentricity ratio ε s and the dynamic eccentricity ratio ε d .

[0073] (2) Based on the analytical model of the open-circuit back electromotive force of the hybrid eccentric outer-rotor hub permanent magnet motor, a hybrid eccentric fault signal database can be efficiently established, so it has a large sample size, high sample diversity, high database construction efficiency, and strong generalization ability of the diagnostic model. Description of the Drawings

[0074] Figure 1 It is the winding distribution diagram of the outer-rotor hub permanent magnet motor.

[0075] Figure 2 It is the flow chart of the present invention.

[0076] Figure 3 It is the schematic diagram of the dimension of the hybrid eccentric fault signal database of the outer-rotor hub permanent magnet motor in the present invention.

[0077] Figure 4 It is the schematic diagram of the hybrid eccentric fault diagnosis model of the outer-rotor hub permanent magnet motor based on the BP neural network in the present invention.

[0078] Figure 5 It is the verification diagram of the hybrid eccentric fault diagnosis model of the outer-rotor hub permanent magnet motor based on the BP neural network in the present invention.

[0079] Figure 6 It is the schematic diagram of the structure of the hybrid eccentric outer-rotor hub permanent magnet motor in the present invention.

[0080] Figure 7 It is the schematic diagram of the test platform of the outer-rotor hub permanent magnet motor in the present invention.

[0081] Figure 8 It is the time-domain diagram of the test result of the open-circuit back electromotive force of the hybrid eccentric outer-rotor hub permanent magnet motor in the present invention.

[0082] Figure 9 It is the frequency-domain diagram of the test result of the open-circuit back electromotive force of the hybrid eccentric outer-rotor hub permanent magnet motor in the present invention. Detailed Embodiment

[0083] The present invention will be described in detail below with reference to the drawings and specific embodiments. The embodiments are implemented on the premise of the technical method of the present invention, and the detailed implementation manners and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.

[0084] Taking an outer-rotor hub permanent magnet motor with 46 poles and 51 slots and a rated speed of 440 rpm as an example, as Figure 1 shown, Figure 1 in which 1 is the rotor, 2 is the permanent magnet, 3 is the stator winding, and 4 is the stator. Diagnose the hybrid eccentric fault according to the flow of the proposed method, and the flow is as Figure 2 shown.

[0085] The specific diagnosis steps are as follows:

[0086] S1: Establish an analytical model of the open-circuit back electromotive force of a hybrid eccentric outer-rotor hub permanent magnet motor.

[0087] S1.1 Calculate the radial magnetic field and tangential magnetic field of the permanent magnets of the outer-rotor hub permanent magnet motor without considering slotting respectively, multiply them by the complex relative permeance considering the stator slotting of the outer-rotor hub permanent magnet motor, and then linearly superpose them to obtain the air-gap radial magnetic density of the outer-rotor hub permanent magnet motor, N λ Take the value of 2000, n takes the value of 440 rpm, ω = 2πn, p = 23, Q = 51, r = 0.1236 mm, and the expression is:

[0088]

[0089] S1.2 Calculate the hybrid eccentric permeance correction coefficient ε using the permeance correction coefficient method M , and the expression is:

[0090]

[0091] Where:

[0092]

[0093]

[0094] S1.3 Calculate the hybrid eccentric air-gap radial magnetic density B of the outer-rotor hub permanent magnet motor Mmr , and the expression is as follows:

[0095]

[0096] S1.4 Establish an analytical model of the hybrid eccentric three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor, and the expression is as follows:

[0097]

[0098]

[0099]

[0100] According to Figure 1 , it can be known that the slot vector matrices A, B, and C of phase A, phase B, and phase C of the example outer-rotor hub permanent magnet motor are respectively:

[0101]

[0102]

[0103]

[0104] k i The value of k can also be obtained according to Figure 1 the number of conductors in each slot. When the i-th slot contains only a single conductor of phase A, phase B, or phase C, k i = 1; when the i-th slot contains two conductors of phase A, phase B, or phase C, k i = 2.

[0105] S2: Construct a large-sample hybrid eccentricity fault signal database based on the analytical model.

[0106] S2.1 Set the parameter change step size and list all combinations of hybrid eccentricity fault parameters (γ0, ε s , ε d ); the value range of γ0 is 0° ≤ γ0 ≤ 360°, the value range of ε s s is 0% ≤ εs ≤ 100%, and the value range of ε d is 0% ≤ ε d ≤ 100%; take the step size |Δγ0| and let γ0 = {|Δγ0|, 2|Δγ0|, 3|Δγ0|,..., a|Δγ0|,...}, where 0° ≤ a|Δγ0| ≤ 360° and a ∈ N * ; take the step size |Δε s | and let ε s s = {|Δε s |, 2|Δε s |, 3|Δε s |,..., s|Δε s |,...}, where 0% ≤ s|Δε s | ≤ 100% and s ∈ N * ; take the step size |Δε d | and let ε d = {|Δε d |, 2|Δε d |, 3|Δε d |,..., d|Δε d |,...}, where 0% ≤ d|Δε d | ≤ 100% and d ∈ N * ;

[0107] S2.2 Eliminate the combinations of hybrid eccentricity fault parameters (γ0, ε m , ε s ) with a hybrid eccentricity ε d ≥ 100%; because when ε m is greater than 100%, the stator and rotor penetrate each other.

[0108] S2.3 Eliminate the combinations of hybrid eccentricity fault parameters (γ0, ε s s, ε s ) with a static eccentricity εd )Combination; since there is no static eccentricity fault when the static eccentricity is 0, there is no static eccentricity circumferential angle.

[0109] S2.4 Based on the analytical model of the hybrid eccentricity three-phase open-circuit back electromotive force of the outer rotor hub permanent magnet motor obtained in S1, traverse the hybrid eccentricity fault parameters (γ0, ε s , ε d ), and analytically calculate the time-domain waveforms (e A , e B , e C ) of the three-phase open-circuit back electromotive force of the outer rotor hub permanent magnet motor. Thus, a large sample hybrid eccentricity fault signal database is obtained.

[0110] In this embodiment, take the step size of the static eccentricity circumferential angle |Δγ0| = 2°, take the step size of the static eccentricity ratio |△ε s | = 5%, and take the step size of the dynamic eccentricity ratio |Δε d | = 5%. Take a max = 180, s max = 20, d max = 20. The number of combinations of hybrid eccentricity fault parameters (γ0, ε s , ε d ) that meet the requirements is N c = d max ×[(s max - 1)×a max / 2 + 1] = 34220. That is, the hybrid eccentricity fault signal database contains N c = 34220 samples, as shown in Figure 3 .

[0111] S3: Extract the characteristics of the fault signal based on the fast Fourier transform.

[0112] S3.1 Through the fast Fourier transform (FFT), perform amplitude-frequency characteristic analysis on the time-domain waveforms (e A , e B , e C ) of the three-phase open-circuit back electromotive force of the outer rotor hub permanent magnet motor for each combination of hybrid eccentricity fault parameters, extract the fundamental wave amplitudes e A_c , e B_c and e C_c of phases A, B, and C, extract the average values e A_s , e B_s and e C_s of the left and right sideband harmonic amplitudes of phases A, B, and C, and obtain (e A_c , e B_c , e C_c , e A_s , e B_s , eC_s );

[0113] S3.2 Based on the relationship that the open - circuit back - electromotive force is proportional to the rotational speed, eliminate the influence of the rotational speed n. As shown in the following two equations, obtain the hybrid eccentricity fault indicators (e A_cn , e B_cn , e C_cn , e A_sn , e B_sn , e C_sn ):

[0114] e i_cn = e i_c / n

[0115] e i_sn = e i_s / e i_c

[0116] In this embodiment, n = 440 rpm.

[0117] S3.3 Construct a hybrid eccentricity fault indicator database I and a hybrid eccentricity fault parameter database O that contain all combinations of hybrid eccentricity fault parameters. The dimension N of the database c = 34220. Represent the database in matrix form as follows:

[0118]

[0119] The corresponding hybrid eccentricity fault parameter database O is represented as follows:

[0120]

[0121] The methods for constructing the fault signal database include the experimental method, the finite - element method, and the analytical method. In this embodiment, the hybrid eccentricity fault signal database contains N c= 34,220 samples. If the experimental method is used, collecting the fault signals under these working conditions is very time-consuming and expensive. To compare the computational efficiency of building a database using the electromagnetic finite element method and the parametric analytical method proposed in the present invention, in this embodiment, the same computing resources, Precision 7920 Tower (Intel(R) Xeon(R) Gold 6240 CPU @ 2.60 GHz, 512 GB RAM), were used to perform comparative calculations on the two methods, and the data shown in Table 1 was obtained. As can be seen from Table 1, under the same computing conditions, the total time for building a database using the electromagnetic finite element model and the parametric analytical model is 7,984 hours and 16.19 hours, respectively. Here, the total time of the finite element calculation is obtained by multiplying the calculation time of a single sample by the number of samples, and the operation time of modeling is not considered. Based on the calculation speed of the electromagnetic finite element, the calculation rate of building a database based on the analytical model is 493 times that of it. In addition, the electromagnetic finite element calculation occupies 120,575 GB of memory, while the analytical calculation only occupies 0.18 GB of memory. Thus, under the current computing conditions, the calculation speed of the electromagnetic finite element is difficult to meet the actual needs and will occupy a large amount of storage resources. The method proposed in this embodiment has the advantages of high efficiency and low memory occupancy.

[0122] Table 1. Comparison of the efficiency and resource occupancy of the fault signal database building methods

[0123]

[0124] S4: Establish a hybrid eccentricity fault diagnosis model based on the BP neural network.

[0125] S4.1 Construct a BP neural network model as Figure 4 shown. Taking the hybrid eccentricity fault index database I as the input and the hybrid eccentricity fault parameter library O as the output, a BP neural network model with 6 inputs, 3 outputs, 3 hidden layers, and the number of nodes in each hidden layer being 36, 27, and 18 respectively is established, and its overall structure is: 6-36-27-18-3. The transfer function of each hidden layer is tansig, the transfer function of the output layer is purelin, and the training function during backpropagation is trainlm.

[0126] S4.2 Data normalization. Randomly select 70% of the corresponding data in databases I and O as the training set and standardize it.

[0127] S4.3 Train the BP neural network model. Set the number of training times to 1,000,000, the learning rate to 0.01, the minimum error of the training target to 0.0000000000001, and the maximum number of failures to 50, and train the neural network. When the error is less than 0.001, stop training.

[0128] S4.4 Verify the BP neural network model. Use the remaining 30% of the data sets in databases I and O as the verification set to verify the diagnostic accuracy of the trained BP neural network. The diagnostic error is below 0.21%, as Figure 5 shown, meeting the usage requirements.

[0129] S5: Eliminate interference faults.

[0130] The proposed hybrid eccentricity fault diagnosis method is applicable to outer rotor hub permanent magnet motors with any hybrid eccentricity fault parameters. In this embodiment, to verify the accuracy of the proposed method, an outer rotor hub permanent magnet motor is designed as a hybrid eccentricity fault prototype with hybrid eccentricity fault parameters ME(270°, 25%, 25%). Figure 6 For its structural schematic diagram, Figure 6 in it, 1 is the rotor, 2 is the outer eccentric sleeve, 3 is a non-standard bearing (inner diameter 25mm, outer diameter 32mm, thickness 7mm), 4 is the inner eccentric sleeve, and 5 is the stator shaft. Among them, O r , O s , O w are the rotor geometric center, the stator geometric center, and the rotor rotation center respectively. e s , e d and e m represent the static eccentricity, the dynamic eccentricity, and the hybrid eccentricity respectively.

[0131] The method for setting the hybrid eccentricity fault used in this embodiment is as follows:

[0132] (1) Make an inner eccentric sleeve with the center of the inner circle deviated from the center of the outer circle by 0.2mm, and align its eccentric direction (the direction in which the center of the inner circle deviates from the center of the outer circle) with the 90° direction of the stator (the opposite direction of the static eccentricity circumferential angle), and then install it on the stator shaft. In this way, a static eccentricity fault with a static eccentricity circumferential angle of 270° and a static eccentricity rate of 25% can be simulated (for example, the air gap length of the outer rotor hub permanent magnet motor is 0.8mm, and the static eccentricity length corresponding to 25% of the static eccentricity rate is 0.2mm).

[0133] (2) Make an outer eccentric sleeve with the center of the inner circle deviated from the center of the outer circle by 0.2mm, and install it in the original bearing seat of the rotor to simulate a 25% dynamic eccentricity fault (for example, the air gap length of the outer rotor hub permanent magnet motor is 0.8mm, and the dynamic eccentricity length corresponding to 25% of the dynamic eccentricity rate is 0.2mm).

[0134] (3) Install the stator with the inner eccentric sleeve and the rotor with the outer eccentric sleeve together through non-standard bearings, as Figure 6 shown.

[0135] S5.1 Test to obtain the resistance R of the three-phase windings of the outer rotor hub permanent magnet motor to be measuredA and R B and R C are both 0.036 Ω. If there is no winding fault in the motor under test, proceed to the next step.

[0136] S5.2 Collect the time-domain waveform of the open-circuit back electromotive force of the motor of the outer-rotor hub permanent magnet motor unit to be tested.

[0137] S5.2.1 Install the outer-rotor hub permanent magnet motor to be tested on the test platform and connect its voltage terminal wires to the oscilloscope. Figure 7 is a schematic diagram of the test platform. Figure 7 It includes a dynamometer mounting fixture 1, a drag motor 2, a coupling 3, a torque and speed sensor 4, an outer-rotor hub permanent magnet motor 6 to be tested, an oscilloscope probe wire 7, an oscilloscope 8, and a mounting platform 9.

[0138] S5.2.2 Start the drag motor and drag the outer-rotor hub permanent magnet motor to be tested to a speed of 440 rpm.

[0139] S5.2.3 Use the oscilloscope to collect the open-circuit back electromotive force signal of the outer-rotor hub permanent magnet motor to be tested within one mechanical cycle, as Figure 8 shown.

[0140] S5.3 Use the Fourier transform method of S3 to extract the spectrum information of the three-phase open-circuit back electromotive forces of the outer-rotor hub permanent magnet motor to be tested in step S5.2.3, as Figure 9 shown. Among them, there is no frequency that is not equal to f c ±f r , and the amplitude is greater than or equal to half of the amplitude of the hybrid eccentricity harmonic f s ±f r of the non-uniform magnetization harmonic component f s ±kf r , k = 1, 2, 3... The motor under test has uniform magnetization. Proceed to the next step.

[0141] S5.4 Use the method of S3 to extract the fault index I T = [0.006618, 0.006550, 0.006615, 0.008620, 0.003539, 0.008073];

[0142] S6: Diagnose the hybrid eccentricity fault parameters of the outer-rotor hub permanent magnet motor to be tested.

[0143] Let I TBringing in the BP neural network model established in S4, the hybrid eccentricity parameter ME’(273.1°, 27.4%, 28.2%) is diagnosed. Compared with the experimental design value ME(270°, 25%, 25%), the absolute diagnostic errors are 3.1°, 2.4%, and 3.2% respectively. The reason for the error may be due to the uncertainty during the installation of the experimental device. The present invention can effectively diagnose the hybrid eccentricity fault parameters.

Claims

1. A data-driven hybrid eccentricity fault diagnosis method for outer rotor hub permanent magnet motors, characterized in that The steps include: S1: Establish an analytical model of open-circuit back-EMF of a hybrid eccentric outer rotor hub permanent magnet motor; S2: Building a large sample hybrid eccentricity fault signal database based on analytical model; S3: Extracting the features of mixed eccentricity fault signal based on fast Fourier transform; S4: Establish a hybrid eccentricity fault diagnosis model based on BP neural network; S5: Eliminate interference faults; S6: diagnose the mixed eccentricity fault parameters of the outer rotor hub permanent magnet motor to be tested; The specific method for establishing the analytical model of the hybrid eccentric open-circuit back-EMF of the outer rotor hub permanent magnet motor described in S1 is: S1.1 Calculate the radial magnetic field and tangential magnetic field of the permanent magnet of the outer rotor hub permanent magnet motor without considering the slots, multiply them by the complex relative magnetic permeance considering the stator slots of the outer rotor hub permanent magnet motor, and then linearly superpose them to obtain the air gap radial magnetic flux density of the outer rotor hub permanent magnet motor, which is expressed as: where B mr is the radial air-gap magnetic density of the outer-rotor hub permanent magnet motor, B Mr_slotless and B Mt_slotless are the radial magnetic field and tangential magnetic field of the permanent magnet of the outer-rotor hub permanent magnet motor without considering the slotting effect, respectively; λ a and λ b are the real part and imaginary part of the complex relative permeance considering the stator slotting effect of the outer-rotor hub permanent magnet motor, respectively; B PMrη and B PMtη are the η-th harmonic amplitudes of the radial magnetic field and tangential magnetic field of the permanent magnet of the outer-rotor hub permanent magnet motor without considering the slotting effect, respectively. η is an odd number, representing the spatial order of the permanent magnet magnetic field of the outer-rotor hub permanent magnet motor; p is the number of permanent magnet pole pairs, ω is the mechanical angular velocity, θ is the spatial angle, and θ0 is the initial phase angle; λ aμ and λ bμ are the harmonic amplitudes of the real and imaginary parts of the complex relative permeance respectively, μ is the harmonic order, N λ is the maximum harmonic order; Q is the number of stator slots, r is the average air-gap radius, t is time; S1.2 Calculate the correction coefficient ε of the mixed eccentricity magnetic conductance using the magnetic conductance correction coefficient method M , and the expression is as follows: in: where, e s and e d are the static eccentricity and the dynamic eccentricity respectively, ε s and ε d represent the static eccentricity ratio and the dynamic eccentricity ratio respectively; δ is the air-gap length of the non-eccentric outer-rotor hub permanent magnet motor; γ0 is the static eccentric circumferential angle; S1.3 Calculate the hybrid eccentric air-gap radial magnetic density B of the outer-rotor hub permanent magnet motor Mmr , and the expression is as follows: S1.4 Establish an analytical model for the hybrid eccentric three-phase open-circuit back electromotive force of the outer rotor hub permanent magnet motor, and the expression is as follows: where e A , e B , e C are the open - circuit back electromotive forces of the A, B, and C phases of the outer - rotor hub permanent - magnet motor respectively; A, B, and C respectively represent the slot vector matrices of the A phase, B phase, and C phase of the outer - rotor hub permanent - magnet motor; N Z is the number of turns of the conductor of the outer - rotor hub permanent - magnet motor, L is the axial length, and R s is the outer diameter of the stator; when the i - th slot contains only a single conductor of the A phase or B phase or C phase, k i = 1; when the i - th slot contains two conductors of the A phase or B phase or C phase, k i = 2; the slots mentioned all refer to the slots opened on the stator of the outer - rotor hub permanent - magnet motor; The specific method described in S2 for constructing a large sample hybrid eccentricity fault signal database based on an analytical model is: S2.1 Set the step size of parameter change and list all the combined parameters of hybrid eccentricity faults (γ0, ε s , ε d ): The value range of γ0 is 0° ≤ γ0 ≤ 360°, and the value range of ε s is 0% ≤ ε s ≤ 100%, and the value range of ε d is 0% ≤ ε d ≤ 100%; Take the step size |Δγ0| and let γ0 = {|Δγ0|, 2|Δγ0|, 3|Δγ0|, …, a|Δγ0|, …}, where 0° ≤ a|Δγ0| ≤ 360° and a ∈ N * ; Take the step size |Δε s | and let ε s = {|Δε s |, 2|Δε s |, 3|Δε s |, …, s|Δε s |, …}, where 0% ≤ s|Δε s | ≤ 100% and s ∈ N * ; Take the step size |Δε d | and let ε d = {|Δε d |, 2|Δε d |, 3|Δε d |, …, d|Δε d |, …}, where 0% ≤ d|Δε ε | ≤ 100% and d ∈ N * ; S2.2 Reject the mixed eccentricity ε m Combination of mixed eccentricity fault parameters (γ0, ε s , ε d ) where ε m is greater than 100%, the stator and rotor penetrate each other; S2.3 Eliminate the static eccentricity ε s = 0 and the hybrid eccentricity fault parameters (γ0, ε s , ε d ) combination; because when the static eccentricity is 0, there is no static eccentricity fault, so there is no static eccentricity circumferential angle; S2.4 Based on the hybrid eccentricity three-phase open-circuit back electromotive force analytical model of the outer-rotor hub permanent magnet motor obtained in S1, traverse the combination of hybrid eccentricity fault parameters (γ0, ε s , ε d ), and analytically calculate the time-domain waveforms of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor (e A , e B , e C ); thereby, obtain a large-sample hybrid eccentricity fault signal database; Containing the mixed eccentricity ε m Definition: ε m = ε s+ ε d ; The specific method for extracting the characteristics of the mixed eccentricity fault signal based on fast Fourier transform described in S3 is: S3.1 By performing a Fast Fourier Transform (FFT) on the time-domain waveforms (e A , e B , e C ) of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor for each combination of hybrid eccentricity fault parameters, the fundamental wave amplitudes e A_c , e B_c , and e C_c of phases A, B, and C are extracted, as well as the average values e A_s , e B_s , and e C_s of the left and right sideband harmonic amplitudes of phases A, B, and C, obtaining (e A_c , e B_c , e C_c , e A_s , e B_s , e C_s ); S3.2 Based on the relationship that the open - circuit back - electromotive force is proportional to the rotational speed, the influence of the rotational speed \(n\) is excluded. As shown in the following two equations, the hybrid eccentricity fault index \((e A_cn ,e B_cn ,e C_cn ,e A_sn ,e B_sn ,e C_sn ) is obtained: e i_cn = e i_c / n e i_sn = e i_s / e i_c where, e i_c represents the fundamental wave amplitude of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor, and e i_cn represents the fundamental wave amplitude of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor after excluding the influence of speed, i = A, B, C; e i_s represents the average value of the harmonic amplitudes of the left and right sidebands of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor, and e i_cn represents the average value of the harmonic amplitudes of the left and right sidebands of the open-circuit back electromotive force of the three phases A, B, and C of the outer rotor hub permanent magnet motor after excluding the influence of speed, i = A, B, C; In S3.1 and S3.2 described above, the fundamental wave amplitude refers to the amplitude of the open-circuit back electromotive force at the fundamental wave frequency f c The left and right sideband harmonic amplitudes respectively refer to the amplitudes of the open-circuit back electromotive force at the left and right sideband harmonic frequencies f l , f y ; The definitions of the fundamental wave frequency f c , the mechanical frequency f r and the left sideband harmonic frequency f l , the right sideband harmonic frequency f y are as follows: When the rotational speed is n, after performing a fast Fourier transform on the time-domain signal of the open-circuit back electromotive force within one mechanical period, the frequency-domain distribution of the open-circuit back electromotive force is obtained; where the fundamental wave frequency f c = np / 60, p is the number of pole pairs of the outer-rotor hub permanent magnet motor; the mechanical frequency f r = n / 60; the left-side harmonic frequency f l = f c - f r , and the right-side harmonic frequency f y = f c + f r ; S3.3 Construct a hybrid eccentricity fault index database I and a hybrid eccentricity fault parameter database O that contain all combinations of hybrid eccentricity fault parameters; the database dimension N c = d max ×[(s max – 1) × a max / 2 + 1], where a max , s max and d max represent the number of steps of the hybrid eccentricity fault parameters γ0, ε s and ε d respectively; the database is represented in matrix form as follows: The corresponding hybrid eccentricity fault parameter database O is expressed as follows: The specific method for establishing a hybrid eccentricity fault diagnosis model based on BP neural network described in S4 is: S4.1 Construct a BP neural network model: Taking the hybrid eccentricity fault index database I as the input and the hybrid eccentricity fault parameter library O as the output, establish a BP neural network model with n i inputs, n o outputs, N h hidden layers, and the number of nodes in each hidden layer is N h1 , N h2 , ……, N (Nh) ). The overall structure of the BP neural network model is: n i -N y1 -N y2 -……-N (Ny) –n o ; The transfer function of each hidden layer is tansig, the transfer function of the output layer is purelin, and the training function during backpropagation is trainlm; S4.2 Data normalization: Randomly select N t sets of corresponding data from databases I and O as the training set, and perform standard normalization on it; S4.3 Train the BP neural network model: Set the number of training times to n t and the learning rate to n l and the minimum error of the training target to n e and the maximum number of failures to n f , and train the neural network; When the error is less than n e , stop training; S4.4 Verify the BP neural network model; use the remaining N c -N t sets of data in databases I and O as the validation set to verify the diagnostic accuracy of the trained BP neural network. If the diagnostic error is within e t or less, it meets the usage requirements; The specific method for eliminating interference faults described in S5 is: S5.1 tests and determines whether the resistances of the three-phase windings of the outer rotor hub permanent magnet motor to be tested are equal. If they are not equal, the motor to be tested has a winding imbalance fault and the diagnosis process ends; if they are equal, the motor to be tested has no winding fault and the next step is executed; S5.2 collects the open-circuit back electromotive force of the outer rotor hub permanent magnet motor unit to be tested; S5.3 Using the Fourier transform method of S3, extract the spectral information of the three-phase open-circuit back electromotive force of the outer-rotor hub permanent magnet motor to be measured collected in step S5.2; First, diagnose whether there is an uneven magnetization harmonic component f c ±f r whose frequency is not equal to f c ±kf r , where k = 1, 2, 3...; Then, in order to tolerate the interference caused by test noise, when the amplitude of the harmonic with frequency f c ±kf r is greater than or equal to half of the amplitude of the mixed eccentricity harmonic f c ±f r , it is judged as an uneven magnetization harmonic; If the uneven magnetization harmonic component exists, the measured motor has an uneven magnetization fault, and the diagnostic process ends; If not, the measured motor has uniform magnetization, and the next step is executed; S5.4 Extract the fault index of the outer rotor hub permanent magnet motor to be tested using the method of S3 I T = [e AT_cn , e BT_cn , e CT_cn , e AT_sn , e BT_sn , e CT_sn ; Where, e iT_cn represents the fundamental wave amplitude of the open-circuit back electromotive force of phases A, B, and C of the outer-rotor hub permanent magnet motor obtained through testing, excluding the influence of rotational speed; e iT_sn represents the average value of the harmonic amplitudes on the left and right sides of the open-circuit back electromotive force of phases A, B, and C of the outer-rotor hub permanent magnet motor obtained through testing, excluding the influence of rotational speed, where i = A, B, C; The specific method for diagnosing the hybrid eccentricity fault parameters of the outer rotor hub permanent magnet motor to be tested described in S6 is as follows: Bring I T into the BP neural network model established by S4, and the hybrid eccentricity parameter ME’(γ0, ε s , ε d ) is obtained by diagnosis.

2. The method for diagnosing the hybrid eccentricity fault of the outer rotor hub permanent magnet motor based on data driving according to claim 1, wherein The method for collecting the open-circuit back electromotive force of the outer rotor hub permanent magnet motor unit motor to be tested as described in S5.2 is as follows: S5.2.1 Install the outer rotor hub permanent magnet motor to be tested on the test platform, and connect the three-phase voltage terminal wires of its unit motor to the oscilloscope; S5.2.2 Start the reverse drag motor to reversely drag the outer rotor hub permanent magnet motor to be tested to a speed n; S5.2.3 Use an oscilloscope to collect the three-phase open-circuit back-EMF signal of the inner and outer rotor hub permanent magnet motor in one mechanical cycle.

Images