Load radial electromagnetic force harmonic wave weakening effect evaluation method of motor

Abstract

The invention provides a motor load radial electromagnetic force harmonic wave weakening effect evaluation method, which belongs to the technical field of electromagnetic force harmonic waves, and comprises the following steps: firstly, acquiring sensor data such as flux density, vibration, current and torque of a motor, constructing a measurement vector, and carrying out segmentation processing according to a sliding time window; then establishing a coupled electromagnetic field and mechanical structure model, discretizing by using a finite element method, and calibrating model parameters by using measured data; a harmonic coefficient matrix of radial electromagnetic force is calculated through flux density data, and a dynamic model of a motor rotor system is established. Respectively inputting the standard harmonic coefficient and the fluctuation harmonic coefficient into the dynamic model to obtain a corresponding vibration response matrix, and defining a harmonic weakening coefficient matrix; finally, a neural network model is trained, the weakening effect of standard harmonic waves and fluctuation harmonic waves can be rapidly predicted, quantitative analysis is conducted, and a comprehensive evaluation result is output. The problem that existing evaluation often passes a large number of experiments and is time-consuming and labor-consuming is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromagnetic harmonics, and in particular, relates to a method for evaluating the weakening effect of radial electromagnetic harmonics of a motor load. Background Art

[0002] As a widely used power conversion device, the performance indicators of motors are crucial for various industrial applications. During the operation of the motor, the electromagnetic interaction between the rotor and the stator will generate periodic radial electromagnetic forces, which have complex harmonic characteristics. These radial electromagnetic force harmonics will cause vibration and fatigue loss in the motor bearings, thereby affecting the reliability and service life of the motor. Therefore, how to effectively weaken the radial electromagnetic force harmonics of the motor has become one of the key issues in motor design and operation management.

[0003] The existing technologies for weakening harmonics of radial electromagnetic force of motors mainly include: electromagnetic structure optimization, current waveform control and vibration absorber. Electromagnetic structure optimization adjusts the distribution of motor magnetic field by optimizing the geometric parameters of motor stator and rotor, such as tooth slot structure and winding distribution, so as to achieve the purpose of weakening harmonics. Current waveform control modulates the stator current to suppress the harmonic components in the current, thereby reducing the higher harmonics in the radial electromagnetic force. Vibration absorber adds damping devices to the motor structure and weakens the vibration response by using the principle of dynamic vibration absorption. These technologies have achieved certain results, but they also have some limitations. For example, electromagnetic structure optimization requires complex modeling and simulation processes, and the optimization design space is limited; current waveform control has high requirements for the drive circuit and can only suppress harmonics of specific frequencies; although vibration absorber can effectively weaken the vibration response, it requires additional structural design and debugging. At present, the evaluation of the harmonic weakening effect of radial electromagnetic force of motors is often time-consuming and labor-intensive through a large number of experiments and requires engineer experience. Therefore, there is an urgent need for a more systematic and intelligent method to evaluate the harmonic weakening effect of the radial electromagnetic force of the motor, which can comprehensively analyze the vibration characteristics during the operation of the motor, quickly evaluate the effects of different weakening measures, and provide a basis for motor design and operation optimization. Summary of the invention

[0004] In view of this, the present invention provides a method for evaluating the harmonic weakening effect of the radial electromagnetic force of a motor, which can solve the technical problem that the current evaluation of the harmonic weakening effect of the radial electromagnetic force of a motor often requires a large number of experiments and engineer experience, which is time-consuming and labor-intensive.

[0005] The present invention is achieved in that:

[0006] The present invention provides a method for evaluating the harmonic weakening effect of a load radial electromagnetic force of a motor, which comprises the following steps:

[0007] S10, collecting sensor data including magnetic density, vibration, current and torque, and constructing a measurement vector; segmenting the collected data according to a preset sliding time window; at the same time, establishing a matrix equation of the motor electromagnetic field and mechanical structure coupling model, and using the data in the measurement vector to calibrate and update the model parameters;

[0008] S20, discretizing the coupling model using the finite element method to obtain the electromagnetic field distribution matrix, and verifying and correcting it using the magnetic field data measured by the Hall sensor; calculating the radial electromagnetic force distribution vector in the air gap of the motor based on the Maxwell stress tensor method and the magnetic field data measured by the Hall sensor;

[0009] S30, performing fast Fourier transform on the radial electromagnetic force vector to obtain a harmonic coefficient matrix, and decomposing it into a standard harmonic coefficient matrix and a fluctuating harmonic coefficient matrix; at the same time, establishing a dynamic model matrix equation of the motor rotor system, and using the measurement data of the acceleration sensor to calibrate and update the relevant matrix;

[0010] S40, inputting the standard harmonic coefficient matrix and the fluctuating harmonic coefficient matrix into the dynamic model as excitations, calculating the standard vibration response matrix and the fluctuating vibration response matrix; and verifying and correcting these response matrices using measured data of the acceleration sensor;

[0011] S50, defining a standard harmonic attenuation coefficient matrix and a fluctuating harmonic attenuation coefficient matrix based on the vibration response matrix and the vibration response matrix when no attenuation measures are taken; calculating a standard evaluation index vector and a fluctuating evaluation index vector, using a weight matrix to reflect the importance of different harmonic orders;

[0012] S60, establishing a training data set containing multiple groups of historical data, including a measurement vector of each sliding time window, a standard harmonic coefficient matrix, a fluctuating harmonic coefficient matrix, and an evaluation index vector; training a neural network model to obtain a load radial electromagnetic force harmonic weakening effect evaluation model;

[0013] S70, for a new motor load condition, obtaining a measurement vector; using a trained load radial electromagnetic force harmonic weakening effect evaluation model, predicting a standard evaluation index vector and a fluctuation evaluation index vector, to achieve fast and accurate harmonic weakening effect evaluation;

[0014] S80, quantitatively analyzing the weakening effect of radial electromagnetic force harmonics of the motor load according to each component of the standard evaluation index vector and the fluctuation evaluation index vector; evaluating the weakening degree of the standard harmonics and the fluctuation harmonics, and identifying the main influencing factors;

[0015] S90. Output comprehensive evaluation results, including standard harmonic attenuation effects and fluctuating harmonic attenuation effects.

[0016] On the basis of the above technical solution, a method for evaluating the harmonic weakening effect of a load radial electromagnetic force of a motor of the present invention can also be improved as follows:

[0017] Wherein, the measurement vector is obtained by a sensor system, and the sensor system comprises:

[0018] The Hall sensor array is evenly arranged on the stator core to measure the air gap magnetic flux distribution;

[0019] An acceleration sensor is installed on the rotor bearing seat to measure the rotor vibration;

[0020] Select key points in the stator winding to install current sensors to measure the stator current;

[0021] A torque sensor is installed on the rotating shaft to measure the load torque.

[0022] Furthermore, the sampling frequency of the sensor system is 1kHz-10kHz.

[0023] Furthermore, the sensor system preprocesses the collected data, including denoising and normalization.

[0024] Furthermore, the length of the preset sliding time window is 10 to 20 times of the sampling period.

[0025] Furthermore, the method for performing fast Fourier transform on the radial electromagnetic force vector is a radix-2 FFT algorithm.

[0026] Furthermore, the neural network model adopts a convolutional neural network structure.

[0027] Specifically: The specific steps of step S10 are to collect sensor data including magnetic flux, vibration, current and torque, and construct a measurement vector {M}. First, the Hall sensor array is evenly arranged on the stator core of the motor to measure the air gap magnetic flux distribution. An acceleration sensor is installed on the rotor bearing seat to measure the rotor vibration. Current sensors are installed at key points in the stator winding to measure the stator current. A torque sensor is installed on the shaft to measure the load torque. The sampling frequency is usually set to 1kHz-10kHz, and the data of each sensor is collected synchronously during the operation of the motor. The collected data is preprocessed, such as denoising and normalization, to construct a measurement vector {M}. The purpose of this step is to obtain key physical quantity data during the operation of the motor.

[0028] Next, the collected data is processed in segments according to the preset sliding time window. At the same time, the matrix equation [A]{X}={B} of the motor electromagnetic field and mechanical structure coupling model is established. The initial value of the model matrix [A] is obtained by finite element analysis and then updated by iterative optimization. The state vector {X} and the input vector {B} are obtained by experimental measurement. The model error ∈ is initially set to the zero vector. During the iterative optimization process, the data in the measurement vector {M} is used to calibrate and update the model parameters. The purpose of this step is to establish a mathematical model of the dynamic behavior of the motor and optimize the model parameters using measured data.

[0029] The specific steps of step S20 are to discretize the coupling model using the finite element method to obtain the electromagnetic field distribution matrix [E]. The specific steps are as follows: first, establish the motor geometry model; set the material properties and boundary conditions; perform meshing; and then solve the Maxwell equations. After obtaining the [E] matrix, verify and correct the magnetic flux density data measured by the Hall sensor [E]. Based on the Maxwell stress tensor method and the magnetic flux density data measured by the Hall sensor, calculate the radial electromagnetic force distribution vector {F r The purpose of this step is to obtain the radial electromagnetic force distribution in the air gap of the motor, laying the foundation for subsequent harmonic analysis.

[0030] The specific steps of step S30 are to calculate the radial electromagnetic force vector {F r} to obtain the harmonic coefficient matrix [H]. The specific steps are as follows: First, {F r} is padded with zeros so that the number of sampling points N is an integer power of 2; then [H] is calculated using the radix-2 FFT algorithm. [H] is decomposed into the standard harmonic coefficient matrix [H std ] and the wave harmonic coefficient matrix [H fluc ]. At the same time, the dynamic model matrix equation of the motor rotor system is established The measurement data of the acceleration sensor is used to calibrate and update the correlation matrices [M], [C], and [K]. The purpose of this step is to obtain the harmonic characteristics of the radial electromagnetic force of the motor and establish a dynamic model of the motor rotor system.

[0031] The specific steps of step S40 are to convert the standard harmonic coefficient matrix [H std ] and the wave harmonic coefficient matrix [H fluc ] are input as excitations into the dynamic model and the standard vibration response matrix [R std ] and the wave vibration response matrix [R fluc The specific calculation process is as follows: First, construct the frequency response function (FRF) matrix ([K]-ω 2 [M]+jω[C]) -1 ; Then [H std] and [H fluc ] as input, and calculate the output response [R std ] and [R fluc ]. The measured data of the acceleration sensor are used to verify and correct these response matrices. The purpose of this step is to obtain the vibration response characteristics of the motor under standard harmonic and fluctuating harmonic excitation.

[0032] The specific steps of step S50 are to define the standard harmonic attenuation coefficient matrix [S std ] and the wave harmonic attenuation coefficient matrix [S fluc ]. [S std The calculation formula of [S std ]=[I]-[R std ][R0] -1 , where [I] is the unit matrix and [R0] is the vibration response matrix when no weakening measures are taken. [S fluc The calculation formula of [S fluc ]=[I]-[R fluc ][R0] -1 . Then calculate the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc {E std The calculation formula is {E std}=[W std ][S std ]{1}, where [W std ] is the weight matrix. fluc The calculation formula is {E fluc}=[W fluc ][S fluc ]{1}, where [W fluc ] is the weight matrix. The purpose of this step is to quantify the weakening effect of standard harmonics and fluctuating harmonics.

[0033] The specific steps of step S60 are to establish a training data set containing multiple sets of historical data, including the measurement vector {M} of each sliding time window, the standard harmonic coefficient matrix [H std ], wave harmonic coefficient matrix [H fluc ], evaluation index vector {E std} and {E fluc}. Then a neural network is trained to obtain a model that can quickly evaluate the harmonic attenuation effect. The forward propagation formula of the neural network is Where y is the output, f is the activation function, and w i is the weight, x i is the input, and b is the bias vector. The loss function of the neural network is Where m is the number of samples, yi is the true value, is the predicted value, and λ is the regularization coefficient. The purpose of this step is to train an intelligent model that can quickly evaluate the harmonic attenuation effect.

[0034] The specific steps of step S70 are to obtain a measurement vector {M} for a new motor load condition; and then use the trained evaluation model to predict the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc}. The prediction formula for the evaluation model is in is the predicted evaluation index vector, [W] is the weight matrix, b is the bias vector, and f is the activation function. The purpose of this step is to use the trained model to quickly and accurately evaluate the harmonic reduction effect under new working conditions.

[0035] The specific steps of step S80 are based on {E std} and {E fluc}, and quantitatively analyze the harmonic reduction effect of the radial electromagnetic force of the motor load. The quantitative index Q of the standard harmonic reduction effect std The calculation formula is Where N is the dimension of the evaluation indicator, E std,i is the i-th component of the standard evaluation index vector. Quantitative index Q of the fluctuation harmonic weakening effect fluc The calculation formula is Comprehensive evaluation results total The calculation formula is Q total =w std Q std +w fluc Q fluc , where w std and w fluc are the weight coefficients of standard harmonics and fluctuating harmonics respectively. The purpose of this step is to quantitatively evaluate the weakening effect of standard harmonics and fluctuating harmonics and give a comprehensive harmonic weakening effect.

[0036] Furthermore, the expression of the measurement vector M is as follows:

[0037] M=[m1,m2,…,m n ] T ;

[0038] In the formula, m i represents the measurement value of the i-th sensor, and n is the total number of sensors. M is obtained experimentally, and the specific steps are as follows:

[0039] Step 1: Install high-precision sensors, including Hall sensors, acceleration sensors, current sensors, and torque sensors;

[0040] Step 2: Set the sampling frequency, usually 1kHz-10kHz;

[0041] Step 3: During the operation of the motor, synchronously collect data from each sensor;

[0042] Step 4: Preprocess the collected data, such as denoising and normalization.

[0043] The matrix equation of the motor electromagnetic field and mechanical structure coupling model is:

[0044] [A]X=B+∈;

[0045] Where [A] is the system matrix, X is the state vector, B is the input vector, and ∈ is the error vector. The initial value of [A] is obtained by finite element analysis and then updated by iterative optimization. X and B are obtained by experimental measurement. ∈ represents the model error and is initially set to the zero vector.

[0046] The update formula of the system matrix [A] is:

[0047]

[0048] In the formula, k represents the number of iterations and η is the learning rate. The value range of η is usually 0.01-0.1, and the optimal value is determined by cross-validation.

[0049] Discretization expression of electromagnetic field distribution matrix [E]:

[0050]

[0051] Among them, N e is the number of finite elements, [N] is the shape function matrix, [D] is the electromagnetic material property matrix, Ω e is the unit integration domain. [E] is obtained by calculation using finite element software (such as ANSYS). The calculation steps are as follows:

[0052] Step 1: Establish the motor geometry model;

[0053] Step 2: Set material properties and boundary conditions;

[0054] Step 3: Mesh division;

[0055] Step 4: Solve Maxwell's equations.

[0056] Radial electromagnetic force distribution vector F r The calculation formula is:

[0057]

[0058] Where μ0 is the vacuum magnetic permeability, which is 4π×10 -7H / m; B n and B t are the normal and tangential magnetic induction intensities, respectively, which are measured by Hall sensors.

[0059] Radial electromagnetic force vector F r The fast Fourier transform of :

[0060]

[0061] Where k is the frequency index, N is the number of sampling points, and j is the imaginary unit. H(k) is calculated using the FFT algorithm. The specific steps are as follows:

[0062] Step 1: F r Pad with zeros so that N is an integer power of 2;

[0063] Step 2: Calculate H(k) using the radix-2 FFT algorithm.

[0064] The matrix equation of the motor rotor system dynamics model is:

[0065]

[0066] Where [M] is the mass matrix, [C] is the damping matrix, [K] is the stiffness matrix, x is the displacement vector, F(t) is the external force vector, and ξ(t) is the random disturbance. [M], [C], and [K] are obtained through modal analysis and system identification. F(t) is calculated by the electromagnetic force. ξ(t) represents the unmodeled disturbance, which is assumed to be Gaussian white noise.

[0067] Standard vibration response matrix [R std ] is calculated by:

[0068] [R std ]=([K]-ω 2 [M]+jω[C]) -1 [H std ];

[0069] Where ω is the angular frequency. std ] is obtained through frequency domain analysis and calculation. The specific steps are as follows:

[0070] Step 1: Construct the frequency response function (FRF) matrix ([K]-ω 2 [M]+jω[C]) -1 ;

[0071] Step 2: [H std ] as input and calculate the output response [R std ].

[0072] Fluctuation vibration response matrix [R fluc] is calculated by:

[0073] [R fluc ]=([K]-ω 2 [M]+jω[C]) -1 [H fluc ];

[0074] [R fluc The calculation method of [R std ] is the same as [H fluc ].

[0075] Standard harmonic attenuation coefficient matrix [S std Definition of ]:

[0076] [S std ]=[I]-[R std ][R0] -1 ;

[0077] Where [I] is the identity matrix and [R0] is the vibration response matrix when no mitigation measures are taken. [R0] is measured by the benchmark test.

[0078] The harmonic attenuation coefficient matrix of fluctuations [S fluc Definition of ]:

[0079] [S fluc ]=[I]-[R fluc ][R0] -1 ;

[0080] [S fluc The calculation method of [S std ] is the same as [R std ] is replaced by [R fluc ].

[0081] Standard evaluation index vector E std Calculation:

[0082] E std =[W std ][S std ]1;

[0083] Among them, [W std ] is the weight matrix, and 1 is the full 1 column vector. [W std ]Determined by expert experience or optimization algorithm.

[0084] Volatility Assessment Index Vector E fluc Calculation:

[0085] F fluc =[W fluc ][S fluc ]1;

[0086] E fluc The calculation method of E std Same as [w std ] and [S Std ] are replaced by [W fluc ] and [S fluc ].

[0087] The forward propagation formula of the neural network is:

[0088]

[0089] Where y is the output, f is the activation function (such as ReLU or sigmoid), and w i is the weight, x i is the input, b is the bias vector, and n is the number of input features. The weights and biases are optimized through the back propagation algorithm.

[0090] The loss function of the neural network is:

[0091]

[0092] Among them, m is the number of samples, y i is the true value, is the predicted value, and λ is the regularization coefficient. λ is determined by cross-validation and usually ranges from 0.001 to 0.1.

[0093] The prediction formula for the evaluation model is:

[0094]

[0095] In the formula, is the predicted evaluation index vector, [W] is the weight matrix, M is the measurement vector, b is the bias vector, and f is the activation function. [W] and b are obtained through neural network training.

[0096] Quantitative indicators of standard harmonic reduction effect:

[0097]

[0098] Among them, N is the dimension of the evaluation index, E std,i is the i-th component of the standard evaluation index vector.

[0099] Quantitative indicators of the wave harmonic weakening effect:

[0100]

[0101] Q fluc The calculation method and Q std Same as E std,iReplace with F fluc,i .

[0102] The calculation formula for the comprehensive evaluation results is:

[0103] Q total =w std Q std +w fluc Q fluc ;

[0104] In the formula, Q total For the overall weakening effect, w std and w fluc are the weight coefficients of standard harmonics and fluctuating harmonics respectively. The weight coefficients are determined by multi-objective optimization methods, such as genetic algorithm or particle swarm optimization algorithm.

[0105] Compared with the prior art, the method for evaluating the harmonic reduction effect of the load radial electromagnetic force of a motor provided by the present invention collects key physical quantities such as magnetic flux density, vibration, current and torque during the operation of the motor, constructs an electromagnetic-mechanical coupling model, and uses frequency domain analysis and dynamic modeling to calculate the reduction effect of standard harmonics and fluctuating harmonics respectively. Finally, the method trains an intelligent evaluation model based on a neural network, which can quickly and accurately predict the harmonic reduction effect under the motor load condition.

[0106] Compared with the existing motor radial electromagnetic force harmonic weakening technology, the method of the present invention has the following significant advantages:

[0107] 1. Comprehensively analyze the motor operation characteristics. Through multi-sensor data fusion, rich physical quantity information during the motor operation process can be obtained, including magnetic flux density distribution, vibration response, current waveform and load torque, etc. This lays a solid foundation for the subsequent electromagnetic-mechanical coupling modeling and harmonic reduction effect evaluation.

[0108] 2. Accurately calculate the harmonic reduction effect. Based on frequency domain analysis and dynamic modeling methods, the present invention can calculate the reduction effect of standard harmonics and fluctuating harmonics respectively, and give quantitative indicators. This sophisticated evaluation method can better guide the optimization of harmonic reduction measures in engineering practice.

[0109] 3. Realize fast intelligent evaluation. The intelligent evaluation model built by neural network can quickly predict the harmonic reduction effect based on a small amount of input data. This data-driven evaluation method greatly improves the efficiency of diagnosis and optimization.

[0110] 4. Support the evaluation of multiple mitigation measures. The method of the present invention is suitable for evaluating the effects of various harmonic mitigation measures such as electromagnetic structure optimization, current waveform control and vibration absorption, providing comprehensive support for motor design and operation optimization.

[0111] In general, the inventive method realizes a rapid and accurate evaluation of the harmonic weakening effect of the radial electromagnetic force of the motor load through technical means such as multi-source sensor data fusion, electromagnetic-mechanical coupling modeling and intelligent evaluation model, solving the technical problem that the current evaluation of the harmonic weakening effect of the radial electromagnetic force of the motor often requires a large number of experiments and the experience of engineers, which is time-consuming and labor-intensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 A flow chart of the method provided by the present invention;

[0113] Figure 2 is the air gap magnetic density map;

[0114] Figure 3 is the rotor vibration acceleration diagram;

[0115] Figure 4 is the stator current diagram;

[0116] Figure 5 is the rotor torque diagram;

[0117] Figure 6 is the radial electromagnetic force distribution diagram in the motor air gap;

[0118] Figure 7 It is the harmonic amplitude distribution diagram under standard working conditions and fluctuating working conditions;

[0119] Figure 8 It is the frequency response function diagram under standard working conditions;

[0120] Fig. 9 It is the frequency response function diagram under fluctuating conditions. DETAILED DESCRIPTION

[0121] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0122] like Figure 1 FIG. 1 is a flow chart of a method for evaluating the harmonic weakening effect of a load radial electromagnetic force of a motor provided by the present invention. The method comprises the following steps:

[0123] S10, collecting sensor data including magnetic density, vibration, current and torque, and constructing a measurement vector; segmenting the collected data according to a preset sliding time window; at the same time, establishing a matrix equation of the motor electromagnetic field and mechanical structure coupling model, and using the data in the measurement vector to calibrate and update the model parameters;

[0124] S20, discretizing the coupling model using the finite element method to obtain the electromagnetic field distribution matrix, and verifying and correcting it using the magnetic field data measured by the Hall sensor; calculating the radial electromagnetic force distribution vector in the air gap of the motor based on the Maxwell stress tensor method and the magnetic field data measured by the Hall sensor;

[0125] S30, performing fast Fourier transform on the radial electromagnetic force vector to obtain a harmonic coefficient matrix, and decomposing it into a standard harmonic coefficient matrix and a fluctuating harmonic coefficient matrix; at the same time, establishing a dynamic model matrix equation of the motor rotor system, and using the measurement data of the acceleration sensor to calibrate and update the relevant matrix;

[0126] S40, inputting the standard harmonic coefficient matrix and the fluctuating harmonic coefficient matrix into the dynamic model as excitations, calculating the standard vibration response matrix and the fluctuating vibration response matrix; and verifying and correcting these response matrices using measured data of the acceleration sensor;

[0127] S50, defining a standard harmonic attenuation coefficient matrix and a fluctuating harmonic attenuation coefficient matrix based on the vibration response matrix and the vibration response matrix when no attenuation measures are taken; calculating a standard evaluation index vector and a fluctuating evaluation index vector, using a weight matrix to reflect the importance of different harmonic orders;

[0128] S60, establishing a training data set containing multiple groups of historical data, including a measurement vector of each sliding time window, a standard harmonic coefficient matrix, a fluctuating harmonic coefficient matrix, and an evaluation index vector; training a neural network model to obtain a load radial electromagnetic force harmonic weakening effect evaluation model;

[0129] S70, for a new motor load condition, obtaining a measurement vector; using a trained load radial electromagnetic force harmonic weakening effect evaluation model, predicting a standard evaluation index vector and a fluctuation evaluation index vector, to achieve fast and accurate harmonic weakening effect evaluation;

[0130] S80, quantitatively analyzing the weakening effect of radial electromagnetic force harmonics of the motor load according to each component of the standard evaluation index vector and the fluctuation evaluation index vector; evaluating the weakening degree of the standard harmonics and the fluctuation harmonics, and identifying the main influencing factors;

[0131] S90. Output comprehensive evaluation results, including standard harmonic attenuation effects and fluctuating harmonic attenuation effects.

[0132] The specific implementation methods of the above steps are described in detail below:

[0133] The specific implementation of step S10 is to collect sensor data including magnetic flux density, vibration, current and torque, and construct a measurement vector {M}. The expression of the measurement vector M is as follows:

[0134]

[0135] Among them, m i represents the measurement value of the i-th sensor, and n is the total number of sensors. The measurement vector M is obtained experimentally, and the specific steps are as follows:

[0136] Step 1: Install high-precision sensors, including Hall sensors, acceleration sensors, current sensors, and torque sensors.

[0137] Step 2: Set the sampling frequency, usually 1kHz-10kHz.

[0138] Step 3: While the motor is running, synchronously collect data from each sensor.

[0139] Step 4: Preprocess the collected data, such as denoising and normalization.

[0140] According to the preset sliding time window, the collected data is processed in segments. At the same time, the matrix equation [A]{X}={B} of the motor electromagnetic field and mechanical structure coupling model is established, where [A] is the system matrix, {X} is the state vector, and {B} is the input vector. The initial value of [A] is obtained by finite element analysis and then updated by iterative optimization. {X} and {B} are obtained by experimental measurement. The model error ∈ is initially set to the zero vector. The update formula of the system matrix [A] is:

[0141]

[0142] Where k is the number of iterations and η is the learning rate. The value range of η is usually 0.01-0.1, and the optimal value is determined by cross-validation.

[0143] The specific implementation of step S20 is to discretize the coupling model using the finite element method to obtain the electromagnetic field distribution matrix [E]. The discretization expression of the electromagnetic field distribution matrix [E] is:

[0144]

[0145] Among them, N e is the number of finite elements, [N] is the shape function matrix, [D] is the electromagnetic material property matrix, Ω e is the unit integration domain. [E] Obtained by finite element software (such as ANSYS).

[0146] Based on the Maxwell stress tensor method and the magnetic flux density data measured by the Hall sensor, the radial electromagnetic force distribution vector {F r}. Radial electromagnetic force distribution vector {F r The calculation formula of} is:

[0147]

[0148] Where μ0 is the vacuum magnetic permeability, which is 4π×10 -7 H / m; B n and B t are the normal and tangential magnetic induction intensities, respectively, which are measured by Hall sensors.

[0149] The specific implementation method of step S30 is to calculate the radial electromagnetic force vector {F r} Perform fast Fourier transform to obtain the harmonic coefficient matrix [H]. The radial electromagnetic force vector {F r The fast Fourier transform expression of} is:

[0150]

[0151] Where k is the frequency index, N is the number of sampling points, and j is the imaginary unit. H(k) is calculated using the FFT algorithm.

[0152] At the same time, the dynamic model matrix equation of the motor rotor system is established:

[0153]

[0154] Where [M] is the mass matrix, [C] is the damping matrix, [K] is the stiffness matrix, x is the displacement vector, F(t) is the external force vector, and ξ(t) is the random perturbation. [M], [C], and [K] are obtained through modal analysis and system identification. F(t) is calculated from the electromagnetic force. ξ(t) represents the unmodeled perturbation, which is assumed to be Gaussian white noise.

[0155] The specific implementation of step S40 is to convert the standard harmonic coefficient matrix [H std ] and the wave harmonic coefficient matrix [H fluc ] are input as excitations into the dynamic model and the standard vibration response matrix [R std ] and the wave vibration response matrix [R fluc ]. Standard vibration response matrix [R std The calculation formula of ] is:

[0156] [R std ]=([K]-ω 2 [M]+jω[C]) -1 [H std ];

[0157] Fluctuation vibration response matrix [R fluc The calculation formula of ] is:

[0158] [R fluc ]=([K]-ω 2[M]+jω[C]) -1 [H fluc ];

[0159] Where ω is the angular frequency.

[0160] The specific implementation of step S50 is to define a standard harmonic attenuation coefficient matrix [S std ] and the wave harmonic attenuation coefficient matrix [S fluc ]. Standard harmonic attenuation coefficient matrix [S std ] is defined as:

[0161] [S std ]=[I]-[R std ][R0] -1 ;

[0162] The harmonic attenuation coefficient matrix of fluctuations [S fluc ] is defined as:

[0163] [S fluc ]=[I]-[R fluc ][R0] -1 ;

[0164] Where [I] is the identity matrix and [R0] is the vibration response matrix when no weakening measures are taken.

[0165] Then calculate the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc}. Standard evaluation index vector {E std The calculation formula of} is:

[0166] {E std}=[W std ][S std ]{1};

[0167] The volatility assessment index vector {E fluc The calculation formula of} is:

[0168] {E fluc}=[W fluc ][S fluc ]{1};

[0169] Among them, [W std ] and [W fluc ] is the weight matrix.

[0170] The specific implementation of step S60 is to establish a training data set containing multiple sets of historical data and train a convolutional neural network. The forward propagation formula of the neural network is:

[0171]

[0172] Among them, y is the output, f is the activation function (such as ReLU or sigmoid), w i is the weight, x i is the input, b is the bias vector, and n is the number of input features. The loss function of the neural network is:

[0173]

[0174] Among them, m is the number of samples, y i is the true value, is the predicted value, and λ is the regularization coefficient. λ is determined by cross-validation and usually ranges from 0.001 to 0.1.

[0175] The specific implementation of step S70 is to use the trained evaluation model to predict the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc}. The prediction formula of the evaluation model is:

[0176]

[0177] in, is the predicted evaluation index vector, [W] is the weight matrix, M is the measurement vector, b is the bias vector, and f is the activation function.

[0178] The specific implementation of step S80 is based on {E std} and {E fluc}, and quantitatively analyze the harmonic reduction effect of the radial electromagnetic force of the motor load. The quantitative index Q of the standard harmonic reduction effect std The calculation formula is:

[0179]

[0180] Quantitative index Q of the wave harmonic weakening effect fluc The calculation formula is:

[0181]

[0182] Comprehensive evaluation results total The calculation formula is:

[0183] Q total =w std Q std +w fluc Q fluc ;

[0184] Among them, w std and w fluc are the weight coefficients of standard harmonics and fluctuating harmonics respectively.

[0185] In summary, this method collects motor operation data through multiple sensors, establishes an electromagnetic-mechanical coupling model, uses frequency domain analysis and dynamic modeling methods to calculate the attenuation effect of standard harmonics and fluctuating harmonics, and finally trains an intelligent evaluation model based on a neural network to achieve fast and accurate harmonic attenuation effect evaluation.

[0186] Specifically, the principle of the present invention is:

[0187] 1. Multi-sensor data fusion to build measurement vectors. Collect sensor data including key physical quantities such as magnetic flux density, vibration, current and torque to build measurement vectors {M}. These physical quantities can fully reflect the electromagnetic and mechanical characteristics of the motor during operation.

[0188] 2. Establish an electromagnetic-mechanical coupling model. According to the preset sliding time window, the collected data is processed in segments. At the same time, the matrix equation [A]{X}={B} for the coupling between the motor electromagnetic field and the mechanical structure is established. This coupling model can describe the dynamic response process of the motor system under electromagnetic force excitation.

[0189] 3. Calculate the harmonic characteristics of the radial electromagnetic force. The coupling model is discretized using the finite element method to obtain the electromagnetic field distribution matrix [E]. Combining the Maxwell stress tensor theory and the magnetic flux density data measured by the Hall sensor, the radial electromagnetic force distribution vector {F r}. Further r} Perform fast Fourier transform to obtain the standard harmonic coefficient matrix [H std ] and the wave harmonic coefficient matrix [H fluc ].

[0190] 4. Analyze the vibration response characteristics. std ] and [H fluc ] are respectively used as excitation inputs to the dynamic model of the motor rotor system, and the standard vibration response matrix [R std ] and the wave vibration response matrix [R fluc ]. This enables the prediction of the vibration response characteristics of the motor under standard harmonic and fluctuating harmonic excitations.

[0191] 5. Quantify the harmonic reduction effect. Based on [R std ]、[R fluc ] and the baseline response [R0] when no mitigation measures are taken, and the standard harmonic mitigation coefficient matrix [S std ] and the wave harmonic attenuation coefficient matrix [S fluc ]. We further define the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc}, which is used to quantitatively describe the effect of harmonic attenuation.

[0192] 6. Train the intelligent evaluation model. Create a training set containing historical data, including the measurement vector {M}, the harmonic coefficient matrix [H std ]、[H fluc ] and the evaluation index vector {E std} and {E fluc}. A neural network-based model is trained to quickly predict the harmonic reduction effect under new operating conditions.

[0193] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: A three-phase asynchronous motor produced by a certain enterprise is widely used in production devices in metallurgy, chemical industry and other industries. The rated power of the motor is 800kW and the rated voltage is 6kV. In order to evaluate the radial electromagnetic force harmonic weakening effect of the motor load, the enterprise adopts the method proposed in the present invention for testing and analysis.

[0194] First, 18 Hall sensors are evenly arranged on the stator core of the motor to measure the air gap magnetic flux distribution. Three acceleration sensors are installed on the rotor bearing seat to measure rotor vibration. Six key points in the stator winding are selected to install current sensors to measure stator current. A torque sensor is installed on the shaft to measure load torque. The sampling frequency is set to 5kHz, and the data of each sensor is synchronously collected under the rated load condition of the motor. The collected raw data is denoised and normalized to construct the measurement vector {M}. The typical waveforms of each physical quantity in the measurement vector {M} are as follows: Figure 2-5 As shown in the figure, it can be seen that there are obvious harmonic components in the waveforms of various physical quantities, which will have an adverse effect on the operating performance of the motor.

[0195] Based on the data in the measurement vector {M}, the matrix equation [A]{X}={B} for the coupling between the motor electromagnetic field and the mechanical structure is first established. The initial parameters of the coupling model are obtained through finite element analysis, and the model parameters are subsequently calibrated and updated using the measurement data through iterative optimization.

[0196] Next, the coupling model is discretized using the finite element method to obtain the electromagnetic field distribution matrix [E]. By solving the [E] matrix and combining it with the magnetic flux density data measured by the Hall sensor, the radial electromagnetic force distribution vector {F r}. Figure 6 The distribution of radial electromagnetic force in the air gap of the motor with the rotor position is given.

[0197] It can be seen that there are obvious periodic fluctuations in the radial electromagnetic force, which indicates the existence of complex harmonic components.

[0198] In order to further analyze the harmonic characteristics, {F r} Perform fast Fourier transform to obtain the standard harmonic coefficient matrix [H std ] and the wave harmonic coefficient matrix [H fluc ]. Figure 7 The amplitude distribution of each order harmonic under standard working condition and fluctuating working condition is given. It can be seen from the figure that under standard working condition, there are mainly 3rd, 5th and 7th order harmonic components, while under fluctuating working condition, there are more high-order harmonic components. These harmonic components will have an adverse effect on the vibration characteristics of the motor.

[0199] In order to quantitatively evaluate the harmonic reduction effect, the method of the present invention establishes the matrix equation of the dynamic model of the motor rotor system: The parameters [M], [C] and [K] of the dynamic model are obtained through modal analysis and system identification. std ] and [H fluc ] are used as excitation inputs, and the standard vibration response matrix [R std ] and the wave vibration response matrix [R fluc ]. Figure 8 and Fig. 9 The frequency response functions of the motor rotor system under standard working conditions and fluctuating working conditions are given. It can be seen from the figure that the main resonance frequency of the motor system under standard working conditions is concentrated around 300Hz, while there are more resonance peaks under fluctuating working conditions, and the frequency range is wider. This shows that the influence of fluctuating harmonics on the vibration response of the motor is more significant.

[0200] Next, the method of the present invention defines a standard harmonic attenuation coefficient matrix [S std ] and the wave harmonic attenuation coefficient matrix [S fluc ]. std ] and [S fluc ] and the preset weight matrix [W std ] and [W fluc ], we can get the standard evaluation index vector {E std} and the volatility assessment index vector {E fluc Table 1 shows the evaluation index results under different working conditions.

[0201] Table 1 Evaluation index results under different working conditions

[0202] Working conditions <![CDATA[E std,1 ]]> <![CDATA[E std,2 ]]> <![CDATA[E std,3 ]]> <![CDATA[E fluc,1 ]]> <![CDATA[E fluc,2 ]]> <![CDATA[E fluc,3 ;]]> standard 0.72 0.81 0.75 0.85 0.88 0.91 fluctuation 0.61 0.68 0.63 0.78 0.83 0.86

[0203] As can be seen from the table, the evaluation index under standard working conditions is generally high, indicating that the standard harmonics are weakened well. However, the evaluation index under fluctuating working conditions is relatively low, indicating that the weakening effect of fluctuating harmonics needs to be further improved. Overall, the radial electromagnetic force harmonic weakening effect of the motor load needs to be further optimized.

[0204] In order to quickly evaluate the harmonic reduction effect under new working conditions, the method of the present invention also trains an intelligent evaluation model based on a neural network. The input of the model is the measurement vector {M}, and the output is {E std} and {E fluc The model consists of three branches, which are used to predict the standard harmonic evaluation index, the fluctuating harmonic evaluation index and the comprehensive evaluation index. Each branch consists of three fully connected layers, and the activation function is ReLU. The Adam optimization algorithm is used for model training, and the loss function is the weighted sum of the mean square error and L2 regularization. The optimal regularization coefficient λ = 0.05 was determined by cross-validation.

[0205] After the intelligent evaluation model is trained, the prediction accuracy on the test set is shown in Table 2.

[0206] Table 2 Prediction accuracy of neural network model

[0207] index Predicted value Actual value Relative error <![CDATA[E std,1 ]]> 0.70 0.72 2.8％ <![CDATA[E std,2 ]]> 0.79 0.81 2.5％ <![CDATA[E std,3 ]]> 0.73 0.75 2.7％ <![CDATA[E fluc,1 ]]> 0.82 0.85 3.5％ <![CDATA[E fluc,2 ]]> 0.86 0.88 2.3％ <![CDATA[E fluc,3 ]]> 0.89 0.91 2.2％ <![CDATA[Q total ]]> 0.78 0.80 2.5％

[0208] It can be seen from the table that the intelligent evaluation model can accurately predict the weakening effect of standard harmonics and fluctuating harmonics, and the average relative error is controlled within 3%, which meets the requirements of engineering applications.

[0209] Based on the above analysis results, the following conclusions can be drawn:

[0210] 1. Under standard working conditions, the motor mainly has 3rd, 5th and 7th order harmonic components, while under fluctuating working conditions, there are more high-order harmonic components. These harmonics have a significant impact on the vibration characteristics of the motor.

[0211] 2. The main resonance frequency of the motor system under standard working conditions is concentrated around 300Hz, while there are more resonance peaks under fluctuating working conditions, and the frequency range is wider. This shows that the influence of fluctuating harmonics on the vibration response of the motor is more significant.

[0212] 3. Through the quantitative analysis of the evaluation index, the weakening effect of standard harmonics is relatively good, while the weakening effect of fluctuating harmonics needs to be further optimized. Overall, the weakening effect of radial electromagnetic harmonics of the motor load needs to be improved.

[0213] 4. The intelligent evaluation model based on neural network can quickly and accurately predict the harmonic reduction effect under new working conditions, providing effective support for subsequent motor design optimization.

[0214] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor load, characterized in that: The following steps are involved: S10, collecting sensor data including magnetic density, vibration, current and torque, and constructing a measurement vector; segmenting the collected data according to a preset sliding time window; at the same time, establishing a matrix equation of the motor electromagnetic field and mechanical structure coupling model, and using the data in the measurement vector to calibrate and update the model parameters; S20, discretizing the coupling model using the finite element method to obtain the electromagnetic field distribution matrix, and verifying and correcting it using the magnetic field data measured by the Hall sensor; calculating the radial electromagnetic force distribution vector in the air gap of the motor based on the Maxwell stress tensor method and the magnetic field data measured by the Hall sensor; S30, performing fast Fourier transform on the radial electromagnetic force vector to obtain a harmonic coefficient matrix, and decomposing it into a standard harmonic coefficient matrix and a fluctuating harmonic coefficient matrix; at the same time, establishing a dynamic model matrix equation of the motor rotor system, and using the measurement data of the acceleration sensor to calibrate and update the relevant matrix; S40, inputting the standard harmonic coefficient matrix and the fluctuating harmonic coefficient matrix into the dynamic model as excitations, calculating the standard vibration response matrix and the fluctuating vibration response matrix; and verifying and correcting these response matrices using measured data of the acceleration sensor; S50, defining a standard harmonic attenuation coefficient matrix and a fluctuating harmonic attenuation coefficient matrix based on the vibration response matrix and the vibration response matrix when no attenuation measures are taken; Calculate the standard evaluation index vector and the fluctuation evaluation index vector, using a weight matrix to reflect the importance of different harmonic orders; S60, establishing a training data set containing multiple groups of historical data, including a measurement vector of each sliding time window, a standard harmonic coefficient matrix, a fluctuating harmonic coefficient matrix, and an evaluation index vector; training a neural network model to obtain a load radial electromagnetic force harmonic weakening effect evaluation model; S70, for a new motor load condition, obtaining a measurement vector; using a trained load radial electromagnetic force harmonic weakening effect evaluation model, predicting a standard evaluation index vector and a fluctuation evaluation index vector, to achieve fast and accurate harmonic weakening effect evaluation; S80, quantitatively analyzing the weakening effect of radial electromagnetic force harmonics of the motor load according to each component of the standard evaluation index vector and the fluctuation evaluation index vector; evaluating the weakening degree of the standard harmonics and the fluctuation harmonics, and identifying the main influencing factors; S90. Output comprehensive evaluation results, including standard harmonic attenuation effects and fluctuating harmonic attenuation effects.

2. The method for evaluating the harmonic weakening effect of the load radial electromagnetic force of a motor according to claim 1, characterized in that: The measurement vector is obtained by a sensor system, and the sensor system comprises: The Hall sensor array is evenly arranged on the stator core to measure the air gap magnetic flux distribution; An acceleration sensor is installed on the rotor bearing seat to measure the rotor vibration; Select key points in the stator winding to install current sensors to measure the stator current; A torque sensor is installed on the rotating shaft to measure the load torque.

3. The method for evaluating the harmonic weakening effect of the load radial electromagnetic force of a motor according to claim 2, characterized in that: The sampling frequency of the sensor system is 1kHz-10kHz.

4. The method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor according to claim 3, characterized in that: The sensor system preprocesses the collected data, including denoising and normalization.

5. The method for evaluating the radial electromagnetic force harmonic weakening effect of a motor according to claim 4, characterized in that: The preset length of the sliding time window is 10 to 20 times of the sampling period.

6. The method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor according to claim 5, characterized in that: The method for performing fast Fourier transform on the radial electromagnetic force vector is the radix-2 FFT algorithm.

7. The method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor load according to claim 6, characterized in that: The neural network model adopts a convolutional neural network structure.

8. The method for evaluating the radial electromagnetic force harmonic weakening effect of a motor according to claim 7, characterized in that: The quantitative indicators of the standard harmonic reduction effect are: In the formula, N is the dimension of the evaluation index, E std,i is the i-th component of the standard evaluation index vector.

9. The method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor according to claim 8, characterized in that: Quantitative indicators of the effect of weakening the harmonics of fluctuations are: In the formula, E fluc,i is the ith component of the volatility evaluation index vector.

10. The method for evaluating the harmonic weakening effect of radial electromagnetic force of a motor load according to claim 9, characterized in that: The calculation formula for the comprehensive evaluation results is: Q total =w std Q std +w fluc Q fluc ; In the formula, Q total For the overall weakening effect, w std and w fluc are the weight coefficients of standard harmonics and fluctuating harmonics respectively.

Images