Method for measuring space pattern distribution of magnetic field of electromagnet

Abstract

The invention discloses a method for measuring space pattern distribution of an electromagnet magnetic field, relates to the technical field of electromagnetic measurement, and aims to solve the problem that pulse electromagnet magnetic field measurement is easily influenced by transient characteristics of a power supply, adopts a phase-locked amplification measurement circuit to perform synchronous detection, extracts a fundamental component through Fourier transform, and obtains the space pattern distribution of the electromagnet magnetic field. A laser interference type magnetic field reference calibration system is combined for calibration, so that magnetic field measurement errors caused by unstable pulse current waveforms are corrected, the phase consistency of measurement signals is improved, and the accuracy of magnetic field measurement is improved; in order to solve the problem of magnetic field reconstruction errors caused by the fact that measured data are influenced by the mutual inductance effect of electromagnets, a Transform magnetic field correction neural network based on a self-attention mechanism is adopted, global feature extraction and nonlinear regression are carried out on magnetic field data, magnetic field compensation is carried out in combination with a multi-layer perceptron (MLP), the measured data are made to be more stable and adaptive, and the accuracy of the magnetic field reconstruction is improved. And the magnetic field calculation deviation caused by the mutual inductance effect is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic measurement, and in particular to a method for measuring the spatial configuration distribution of the magnetic field of an electromagnet. Background Art

[0002] In fields such as high-power microwaves, plasma control, and magnetic confinement fusion, magnetic field measurement is crucial. In the environment of pulsed electromagnets, obtaining the spatial configuration distribution of the magnetic field plays a key role in optimizing the design and improving the system performance. Traditional measurement methods generally use Hall sensors or fluxgate sensors to perform point-by-point measurements at specific points, and reconstruct the magnetic field distribution through interpolation or fitting algorithms.

[0003] For steady-state electromagnets, such a scheme can achieve high-precision measurement through long-term data accumulation. However, when the measurement object becomes a pulsed electromagnet, the magnetic field of the pulsed electromagnet will be established and disappear rapidly within an extremely short time; the measurement accuracy of the magnetic field distribution is affected by the stability of the pulsed current, and the transient characteristics of the pulsed power supply system cause slight fluctuations in the current peak value during each discharge. Such fluctuations result in inconsistent magnetic field values at the same measurement point in different pulse periods, thereby affecting the reconstruction accuracy of the magnetic field spatial distribution; in addition, the heating effect of the electromagnet will cause resistance changes, making the current waveform drift after multiple discharges, further exacerbating the measurement instability; to reduce errors, some traditional methods use the method of taking the average value of multiple discharges for compensation, but due to the discharge variability of the pulsed electromagnet itself, even for multiple discharges under the same experimental conditions, it is difficult for the pulse peaks to be exactly the same.

[0004] In the existing patent technology such as the published number CN117647762A, for the measurement of the spatial configuration of the magnetic field of a pulsed electromagnet, interpolation or curve fitting is used for reconstruction. However, the fluctuation of the pulsed current peak value leads to measurement errors, resulting in deviations in the spatial distribution reconstruction; the magnetic field calculation uses a simple linear superposition model without introducing the mutual inductance effect between multiple electromagnets, reducing the accuracy of the magnetic field measurement; the existing scheme is difficult to perform efficient and accurate measurements in the face of a complex magnetic field environment. Therefore, there is an urgent need for a method that introduces a more stable measurement of the spatial configuration distribution of the magnetic field of an electromagnet to improve the accuracy of the measurement of the spatial configuration distribution of the magnetic field of a pulsed electromagnet. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a method for measuring the spatial configuration distribution of the magnetic field of an electromagnet to solve the problems that the measurement of the magnetic field of a pulsed electromagnet by traditional methods is affected by current fluctuations, has large reconstruction errors, ignores the mutual inductance effect, and has limited measurement accuracy.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for measuring the spatial configuration distribution of the magnetic field of an electromagnet, which includes:

[0009] Step S1: Pass a pulsed current through the pulsed electromagnet to generate a magnetic field to be measured, synchronously detect the pulsed current using a lock-in amplifier measurement circuit to obtain reference phase information, and calibrate it through a laser interferometric magnetic field reference calibration system;

[0010] Step S2: Based on the reference phase obtained in Step S1, measure the magnetic induction intensity at multiple points within a specific time window;

[0011] Step S3: Based on the measured magnetic induction intensity, calculate the ratio of the magnetic induction intensity at each measurement point to the peak value of the pulsed current, and process the data using a deep learning magnetic field correction model;

[0012] Step S4: Based on the ratio data of the magnetic induction intensity to the peak value of the pulsed current processed in Step S3, use a compressive sensing algorithm for interpolation and completion;

[0013] Step S5: Based on the magnetic field spatial distribution data obtained in Step S4, combined with the measured magnetic field data during the measurement process, use an adaptive filtering algorithm for noise suppression and trend modeling to calculate the magnetic field spatial configuration distribution data.

[0014] As a preferred solution of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: the peak value and duration of the pulsed current are affected by the transient characteristics of the power supply.

[0015] As a preferred solution of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: the laser interferometric magnetic field reference calibration system includes a laser interferometer, a magnetic field standard source, and a feedback control module;

[0016] The laser interferometer is used to measure the relative position change of the magnetic field standard source in the external magnetic field, and the magnetic field standard source is a reference magnet or a superconducting magnet with a known magnetic induction intensity;

[0017] The feedback control module is used to receive the displacement data measured by the laser interferometer, and combined with the characteristics of the magnetic field standard source, calculate the reference value of the ambient magnetic field to calibrate the magnetic field deviation generated by the pulsed electromagnet during the measurement process.

[0018] As a preferred solution of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: in Step S1, the specific method of synchronous detection of the lock-in amplifier measurement circuit and laser interferometric magnetic field reference calibration is:

[0019] Pass a pulsed current through the pulsed electromagnet. The waveform of this pulsed current is affected by the transient characteristics of the power supply and may exhibit short-time peak value changes;

[0020] A phase-locked amplifier measurement circuit is used to synchronously detect the pulsed current. The phase-locked amplifier extracts a reference signal, compares the phase with the pulsed current, calculates the phase deviation and outputs the reference phase information;

[0021] The pulsed current signal is analyzed by Fourier transform to decompose the fundamental wave and higher harmonic components, and the phase of the main frequency signal is extracted;

[0022] The phase difference between the fundamental wave component and the external reference signal is calculated to obtain the reference phase information of the pulsed current;

[0023] Calibration is carried out using a laser interferometric magnetic field reference calibration system. The laser interferometer measures the relative displacement of the magnetic field standard source in the magnetic field;

[0024] Combined with the known magnetic induction intensity of the magnetic field standard source and the measured displacement, the ambient magnetic field reference value is calculated, and the magnetic field deviation generated by the pulsed electromagnet is corrected.

[0025] As a preferred scheme of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet described in the present invention, wherein: the measurement is synchronously acquired using a superconducting quantum interference device SQUID or a fiber optic magnetic sensor.

[0026] As a preferred scheme of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet described in the present invention, wherein: the deep learning magnetic field correction model uses a Transformer magnetic field correction neural network;

[0027] The Transformer magnetic field correction neural network is a deep learning model based on the self-attention mechanism. Through global feature extraction and nonlinear regression, it learns the spatial relationship of the magnetic field distribution, compensates for the influence of the mutual inductance effect of the electromagnet on the magnetic field distribution, and dynamically adjusts the magnetic field calculation result.

[0028] As a preferred scheme of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet described in the present invention, wherein: the step of calculating the ratio of the magnetic induction intensity at each measurement point to the peak value of the pulsed current and processing the data using the deep learning magnetic field correction model is,

[0029] The magnetic induction intensities at multiple measurement points are measured. Let the magnetic induction intensity at a certain measurement point be B i :

[0030] B i = f(I p ),

[0031] Calculate the ratio of the magnetic induction intensity at this measurement point to the peak value of the pulsed current. The ratio formula is:

[0032] R i = Bi / I p ,

[0033] Among them, B i represents the magnetic induction intensity at the i-th measurement point, and I p represents the peak value of the pulsed current, and R i represents the ratio of the magnetic induction intensity at the i-th measurement point to the peak value of the pulsed current, and f(·) is the mapping function of the magnetic induction intensity to the peak value of the pulsed current;

[0034] Train the Transformer magnetic field correction neural network, collect the ratio data {R i} at multiple measurement points to form a magnetic field data set where x i is the spatial coordinate of the measurement point,

[0035] Input encoding is adopted using the position encoding function:

[0036] PE(x i ) = sin(ω k x i ) + cos(ω k x i ),

[0037] where PE(x i ) is the position encoding value, ω k is the encoding frequency parameter, and x i is the spatial coordinate of the measurement point,

[0038] The self-attention mechanism is used to calculate the magnetic field distribution relationship, and the calculation formula is:

[0039] Q = W q R, K = W k R, V = W v R,

[0040] Calculate the attention weights:

[0041]

[0042] Calculate the weighted output:

[0043] Z = AV,

[0044] where Q, K, and V are the query, key, and value matrices respectively, R represents the magnetic induction intensity ratio vector, W q , W k , W v are the weight matrices, d k is the dimension of the key vector, A is the attention weight, and Z is the output calculated by the Transformer;

[0045] Nonlinear regression is performed using a multi-layer perceptron (MLP) to correct the magnetic field distribution. The correction formula is as follows:

[0046]

[0047] Where is the corrected magnetic induction intensity ratio, and MLP(·) is a multi-layer perceptron model for nonlinear regression;

[0048] The mean squared error (MSE) is used as the loss function:

[0049]

[0050] Where L is the loss value and N is the total number of samples in the dataset.

[0051] The Adam optimizer is used to update the network parameters to converge the loss function.

[0052] As a preferred embodiment of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: the step of performing interpolation and completion using the compressive sensing algorithm is as follows.

[0053] Let the sparse representation of the magnetic field distribution data B under the basis Ψ be:

[0054] B = Ψs

[0055] Where B is the magnetic field spatial distribution data, Ψ is the sparse transformation basis matrix, and s is the sparse coefficient vector.

[0056] A measurement matrix is constructed and measured using the observation matrix Φ:

[0057] y = ΦB

[0058] Where y is the observed data and Φ is the measurement matrix.

[0059] B is recovered by solving the l 1 constrained optimization problem:

[0060] subject to y = ΦΨs

[0061] The interpolated and completed magnetic field distribution is calculated by solving the sparse vector s:

[0062] Where represents the magnetic field spatial distribution data after compressive sensing interpolation and completion.

[0063] As a preferred embodiment of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: in step S5, the magnetic field spatial configuration is calculated based on adaptive filtering. Specifically:

[0064] The magnetic field state is predicted using Kalman filtering. Let the magnetic field state vector be x j , and its state transition equation is:

[0065] x j+1 = F j x j + w j ,

[0066] where x j is the magnetic field state vector at the j-th time step, x j+1 is the state at the next time step, F j is the state transition matrix, w j is the process noise vector, which follows a zero-mean Gaussian distribution, Q j is the process noise covariance matrix,

[0067] The magnetic field state is updated by combining the measurement data. The update process is:

[0068] y j = H j x j + v j ,

[0069] where y j is the measurement vector, H j is the observation matrix, v j is the measurement noise vector, which follows a zero-mean Gaussian distribution, R j is the measurement noise covariance matrix.

[0070] As a preferred embodiment of the method for measuring the spatial configuration distribution of the magnetic field of an electromagnet according to the present invention, wherein: the step of calculating the spatial configuration of the magnetic field based on adaptive filtering further includes:

[0071] Performing Kalman filter iterative calculations:

[0072] Prediction step:

[0073]

[0074] where, is the predicted state at the (j + 1)-th time step, P j+1|j is the predicted error covariance matrix,

[0075] Update step:

[0076]

[0077] P j|j = (I - K j H j)P j|j-1 ,

[0078] where K j is the Kalman gain, is the updated magnetic field state, P j|j is the updated error covariance matrix, and I is the identity matrix;

[0079] The magnetic field distribution trend is fitted by a multi-order polynomial, and the formula is:

[0080]

[0081] where B trend (x) is the trend term of the magnetic field distribution, M is the highest order of fitting, c m is the fitting coefficient, and x is the spatial coordinate;

[0082] Combined with the filtered data, the final magnetic field distribution is calculated:

[0083] B final (x) = B trend (x) + KF(y),

[0084] where B final (x) is the final magnetic field spatial configuration distribution, and KF(y) is the measurement data correction term after Kalman filtering.

[0085] The beneficial effects of the present invention are as follows: For the problem that the magnetic field measurement of pulsed electromagnets is easily affected by the transient characteristics of the power supply, a lock-in amplifier measurement circuit is used for synchronous detection, the fundamental wave component is extracted by Fourier transform, and calibration is carried out in combination with a laser interferometric magnetic field reference calibration system, so as to correct the magnetic field measurement error caused by the unstable pulsed current waveform, improve the phase consistency of the measurement signal, and improve the accuracy of magnetic field measurement.

[0086] For the problem that the measurement data is affected by the mutual inductance effect of the electromagnet, resulting in magnetic field reconstruction error, the present invention adopts a Transformer magnetic field correction neural network based on the self-attention mechanism to extract global features and perform non-linear regression on the magnetic field data, and combines a multi-layer perceptron MLP for magnetic field compensation, making the measurement data more stable and self-adaptive, and effectively reducing the magnetic field calculation deviation caused by the mutual inductance effect.

[0087] In view of the problem that the limited number of magnetic field measurement points leads to limited spatial resolution, this invention uses the compressed sensing algorithm to interpolate and complete magnetic field data. By constructing a sparse transformation basis matrix and adopting the l1 regularization optimization method, it restores the complete magnetic field distribution data, thereby reducing the requirement for measurement points, improving the magnetic field spatial resolution, and at the same time reducing the experimental cost. In view of the problem of measurement instability caused by magnetic field measurement noise and transient changes, this invention combines Kalman filtering for dynamic magnetic field state estimation and uses the polynomial fitting method to extract the magnetic field distribution trend, making the magnetic field data more robust and capable of adapting to the measurement requirements in different environments, improving the stability and accuracy of the measurement results.

[0088] In summary, compared with the traditional method that relies on point-by-point scanning and simple interpolation reconstruction, this invention realizes efficient, accurate, and low-error calculation of magnetic field measurement by integrating signal processing, machine learning, and optimization calculation technologies, and can provide a more practically valuable technical solution in the scenario of pulsed electromagnet magnetic field measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] In order to more clearly illustrate the technical solutions of the embodiments of this invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of this invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0090] Figure 1 It is a schematic flowchart of the method for measuring the spatial configuration distribution of the magnetic field of the electromagnetic body of this invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0091] To make the above objects, features, and advantages of this invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of this invention with reference to the drawings in the specification.

[0092] Many specific details are set forth in the following description to facilitate a thorough understanding of this invention. However, this invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of this invention. Therefore, this invention is not limited by the specific embodiments disclosed below.

[0093] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation of this invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0094] Example 1, refer to Figure 1, this embodiment provides a method for measuring the spatial configuration distribution of the magnetic field of an electromagnet, including the following steps:

[0095] Step S1, apply a pulsed current to the pulsed electromagnet to generate a magnetic field to be measured, use a lock-in amplifier measurement circuit to synchronously detect the pulsed current, obtain reference phase information, and calibrate it through a laser interferometric magnetic field reference calibration system;

[0096] The peak value and duration of the pulsed current are affected by the transient characteristics of the power supply;

[0097] The laser interferometric magnetic field reference calibration system includes a laser interferometer, a magnetic field standard source, and a feedback control module;

[0098] The laser interferometer is used to measure the relative position change of the magnetic field standard source in the external magnetic field, and the magnetic field standard source is a reference magnet or superconducting magnet with a known magnetic induction intensity;

[0099] The feedback control module is used to receive the displacement data measured by the laser interferometer, and combine the characteristics of the magnetic field standard source to calculate the reference value of the ambient magnetic field and calibrate the magnetic field deviation generated by the pulsed electromagnet during the measurement process;

[0100] In step S1, the specific methods of synchronous detection of the lock-in amplifier measurement circuit and laser interferometric magnetic field reference calibration are as follows:

[0101] Apply a pulsed current to the pulsed electromagnet. The waveform of this pulsed current is affected by the transient characteristics of the power supply and may exhibit short-term peak value changes;

[0102] Use a lock-in amplifier measurement circuit to synchronously detect the pulsed current. The lock-in amplifier extracts a reference signal, compares the phase with the pulsed current, calculates the phase deviation, and outputs the reference phase information;

[0103] Analyze the pulsed current signal through Fourier transform, decompose the fundamental wave and higher harmonic components, and extract the phase of the main frequency signal;

[0104] Calculate the phase difference between the fundamental wave component and the external reference signal to obtain the reference phase information of the pulsed current;

[0105] Calibrate using a laser interferometric magnetic field reference calibration system. The laser interferometer measures the relative displacement of the magnetic field standard source in the magnetic field;

[0106] Combine the known magnetic induction intensity of the magnetic field standard source with the measured displacement, calculate the reference value of the ambient magnetic field, and correct the magnetic field deviation generated by the pulsed electromagnet;

[0107] Specifically, in step S1, a pulsed current is applied to the pulsed electromagnet, and synchronous detection is performed using a lock-in amplifier measurement circuit to keep the phase of the pulsed signal consistent with the measurement signal, thereby obtaining accurate reference phase information. Here, to further improve the magnetic field measurement accuracy, a laser interferometric magnetic field reference calibration system is adopted. The relative position change of the magnetic field standard source in the external magnetic field is measured using a laser interferometer, and the reference value of the ambient magnetic field is calculated through the feedback control module: first, synchronous detection is achieved through a lock-in amplifier, and then the magnetic field is calibrated based on a reference using high-precision interferometric measurement technology, thereby effectively reducing the magnetic field fluctuations caused by the transient characteristics of the power supply and improving the measurement accuracy.

[0108] Step S2, based on the reference phase obtained in step S1, measure the magnetic induction intensity at multiple points within a specific time window.

[0109] The measurement is synchronously collected using a superconducting quantum interference device (SQUID) or a fiber optic magnetic sensor.

[0110] Step S3, based on the measured magnetic induction intensity, calculate the ratio of the magnetic induction intensity at each measurement point to the peak value of the pulsed current, and process the data using a deep learning magnetic field correction model.

[0111] The deep learning magnetic field correction model uses a Transformer magnetic field correction neural network.

[0112] The Transformer magnetic field correction neural network is a deep learning model based on the self-attention mechanism. Through global feature extraction and nonlinear regression, it learns the spatial relationship of the magnetic field distribution, compensates for the influence of the mutual inductance effect of the electromagnets on the magnetic field distribution, and dynamically adjusts the magnetic field calculation results.

[0113] The steps of calculating the ratio of the magnetic induction intensity at each measurement point to the peak value of the pulsed current and processing the data using a deep learning magnetic field correction model are as follows:

[0114] The magnetic induction intensities at multiple measurement points are measured. Let the magnetic induction intensity at a certain measurement point be B i :

[0115] B i = f(I p ),

[0116] Calculate the ratio of the magnetic induction intensity at this measurement point to the peak value of the pulsed current. The ratio formula is:

[0117] R i = B i / I p ,

[0118] where B i represents the magnetic induction intensity at the i-th measurement point, and I prepresents the peak value of the pulsed current, R i represents the ratio of the magnetic induction intensity at the i-th measurement point to the peak value of the pulsed current, and f(·)i is the mapping function of the magnetic induction intensity to the peak value of the pulsed current;

[0119] Train the Transformer magnetic field correction neural network, collect the ratio data {R i} at multiple measurement points to form a magnetic field dataset where x i is the spatial coordinate of the measurement point,

[0120] Input encoding is adopted using the position encoding function:

[0121] PE(x i ) = sin(ω k x i ) + cos(ω k x i ),

[0122] where PE(x i ) is the position encoding value, ω k is the encoding frequency parameter, x i is the spatial coordinate of the measurement point,

[0123] The self-attention mechanism is used to calculate the magnetic field distribution relationship, and the calculation formula is:

[0124] Q = W q R, K = W k R, V = W v R,

[0125] Calculate the attention weights:

[0126]

[0127] Calculate the weighted output:

[0128] Z = AV,

[0129] where Q, K, V are the query, key, and value matrices respectively, R represents the magnetic induction intensity ratio vector, W q , W k , W v are the weight matrices, d k is the dimension of the key vector, A is the attention weight, and Z is the output calculated by the Transformer;

[0130] The multi-layer perceptron MLP is used for non-linear regression to correct the magnetic field distribution, and the correction formula is:

[0131]

[0132] where, is the corrected magnetic induction intensity ratio, and MLP(·) is a multi-layer perceptron model for non-linear regression;

[0133] The mean square error MSE is used as the loss function:

[0134]

[0135] where L is the loss value and N is the total number of samples in the data set,

[0136] The Adam optimizer is used to update the network parameters to converge the loss function;

[0137] Specifically, based on the measured magnetic induction intensity, the ratio of the magnetic induction intensity at each measurement point to the peak value of the pulse current is calculated here to construct a magnetic field feature data set; and a Transformer magnetic field correction neural network is introduced to extract the global features of the magnetic field distribution by the self-attention mechanism and model the influence of the mutual inductance effect through non-linear regression; the Transformer model is used to learn the spatial distribution relationship of the magnetic field, and MLP is used for magnetic field compensation and correction to make the measurement data more accurate. Through loss function optimization and Adam training, the magnetic field calculation results are dynamically adjusted to improve the robustness and accuracy of the measurement data;

[0138] Step S4: Based on the ratio data of the magnetic induction intensity to the peak value of the pulse current processed in step S3, the compressive sensing algorithm is used for interpolation and completion;

[0139] The steps of using the compressive sensing algorithm for interpolation and completion are as follows.

[0140] Suppose the sparse representation of the magnetic field distribution data B under the basis Ψ is:

[0141] B = Ψs,

[0142] where B is the magnetic field spatial distribution data, Ψ is the sparse transformation basis matrix, and s is the sparse coefficient vector.

[0143] A measurement matrix is constructed and measured using the observation matrix Φ:

[0144] y = ΦB,

[0145] where y is the observed data and Φ is the measurement matrix.

[0146] Restore B through the l 1 constrained optimization problem:

[0147] subject to y = ΦΨs,

[0148] Calculate the interpolated and completed magnetic field distribution by solving the sparse vector s:

[0149] wherein represents the magnetic field spatial distribution data after compressive sensing interpolation completion;

[0150] Specifically, in step S4, for the problem of incomplete spatial distribution of measurement data, a compressive sensing algorithm is used for interpolation completion; a sparse representation of magnetic field data is constructed using a sparse transformation basis matrix, and finite observation data is obtained through a measurement matrix. Subsequently, the l1 regularization optimization method is used to recover the complete magnetic field distribution data, solving the data missing problem caused by sparse measurement points, effectively utilizing the sparse characteristics of magnetic field data, improving the spatial resolution, and reducing the measurement cost;

[0151] In step S5, based on the magnetic field spatial distribution data obtained in step S4, combined with the measured magnetic field data during the measurement process, an adaptive filtering algorithm is used for noise suppression and trend modeling to calculate the magnetic field spatial configuration distribution data;

[0152] In step S5, the magnetic field spatial configuration is calculated based on adaptive filtering. Specifically:

[0153] The Kalman filter is used to predict the magnetic field state. Let the magnetic field state vector be x j , and its state transition equation is:

[0154] x j+1 = F j x j + w j ,

[0155] wherein, x j is the magnetic field state vector at the jth time step, x j+1 is the state at the next time step, F j is the state transition matrix, w j is the process noise vector, which follows a zero-mean Gaussian distribution, Q j is the process noise covariance matrix,

[0156] The magnetic field state is updated by combining the measurement data. The update process is:

[0157] y j = H j x j + v j ,

[0158] wherein, y j is the measurement vector, H j is the observation matrix, v j is the measurement noise vector, which follows a zero-mean Gaussian distribution, R j is the measurement noise covariance matrix;

[0159] The steps for calculating the magnetic field spatial configuration based on adaptive filtering further include:

[0160] Performing Kalman filter iterative calculation:

[0161] Prediction step:

[0162]

[0163] Wherein, is the predicted state at the (j + 1)-th time step, and P j+1|j is the predicted error covariance matrix,

[0164] Update step:

[0165]

[0166] P j|j =(I - K j H j )P j|j-1 ,

[0167] Wherein, K j is the Kalman gain, is the updated magnetic field state, and P j|j is the updated error covariance matrix, and I is the identity matrix;

[0168] Using a multi-order polynomial to fit the magnetic field distribution trend, the formula is:

[0169]

[0170] Wherein, B trend (x) is the trend term of the magnetic field distribution, M is the highest order of fitting, c m is the fitting coefficient, and x is the spatial coordinate;

[0171] Combining the filtered data, calculating the final magnetic field distribution:

[0172] B final (x)=B trend (x)+KF(y),

[0173] Wherein, B final (x) is the final magnetic field spatial configuration distribution, and KF(y) is the measurement data correction term after Kalman filtering;

[0174] Specifically, based on the magnetic field spatial distribution data supplemented in step S4 herein, combining with the original magnetic field data obtained during the measurement process, an adaptive filtering algorithm is used for noise suppression and trend modeling to calculate the final magnetic field spatial configuration distribution; specifically:

[0175] First, the Kalman filtering method is used to predict the magnetic field state, and it is dynamically updated through the observed data to filter out the measurement noise. Subsequently, the polynomial fitting method is adopted to model the spatial trend of the magnetic field, so as to extract the dominant change characteristics of the magnetic field. Finally, the filtered data is combined with the trend fitting result to calculate the final spatial configuration distribution of the magnetic field. Compared with the traditional filtering method, the adaptive filtering can adjust the filtering parameters according to the dynamic changes of the magnetic field data, improve the denoising accuracy, and enhance the adaptability of the system to measurement errors, which can effectively improve the stability and accuracy of the magnetic field distribution measurement.

[0176] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet, characterized in that: include, Step S1, passing a pulse current into a pulse electromagnet to generate a magnetic field to be measured, using a phase-locked amplification measurement circuit to synchronously detect the pulse current, obtaining reference phase information, and calibrating through a laser interferometer magnetic field reference calibration system; Step S2, based on the reference phase obtained in step S1, measuring the magnetic induction intensity of multiple points within a specific time window; Step S3, based on the measured magnetic induction intensity, calculate the ratio of the magnetic induction intensity at each measuring point to the peak value of the pulse current, and use a deep learning magnetic field correction model to process the data; Step S4, based on the magnetic induction intensity and pulse current peak ratio data processed in step S3, a compressed sensing algorithm is used for interpolation completion; Step S5, based on the magnetic field spatial distribution data obtained in step S4 and combined with the measured magnetic field data in the measurement process, an adaptive filtering algorithm is used to perform noise suppression and trend modeling to calculate the magnetic field spatial configuration distribution data.

2. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 1, characterized in that: The peak value and duration of the pulse current are affected by the transient characteristics of the power supply.

3. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 2, characterized in that: The laser interferometer magnetic field reference calibration system includes a laser interferometer, a magnetic field standard source and a feedback control module; The laser interferometer is used to measure the relative position change of a magnetic field standard source in an external magnetic field, and the magnetic field standard source is a reference magnet or a superconducting magnet with known magnetic induction intensity; The feedback control module is used to receive the displacement data measured by the laser interferometer, and calculate the reference value of the environmental magnetic field in combination with the characteristics of the magnetic field standard source, and calibrate the magnetic field deviation generated by the pulse electromagnet during the measurement process.

4. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 3, characterized in that: In step S1, the specific method of synchronous detection of the phase-locked amplification measurement circuit and laser interferometric magnetic field reference calibration is: A pulse current is passed through the pulse electromagnet, and the waveform of the pulse current is affected by the transient characteristics of the power supply; The phase-locked amplifier measurement circuit is used to synchronously detect the pulse current. The phase-locked amplifier extracts the reference signal, compares the phase with the pulse current, calculates the phase deviation and outputs the reference phase information. Analyze the pulse current signal through Fourier transform, decompose the fundamental wave and high-order harmonic components, and extract the phase of the main frequency signal; Calculate the phase difference between the fundamental component and the external reference signal to obtain the reference phase information of the pulse current; The laser interferometer magnetic field reference calibration system is used for calibration, and the laser interferometer measures the relative displacement of the magnetic field standard source in the magnetic field; The known magnetic induction intensity of the magnetic field standard source is combined with the measured displacement to calculate the ambient magnetic field baseline value and correct the magnetic field deviation generated by the pulse electromagnet.

5. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 4, characterized in that: The measurement is performed by using a superconducting quantum interference device SQUID or an optical fiber magnetic sensor for synchronous acquisition.

6. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 5, characterized in that: The deep learning magnetic field correction model adopts a Transformer magnetic field correction neural network; The Transformer magnetic field correction neural network is a deep learning model based on the self-attention mechanism. It learns the spatial relationship of the magnetic field distribution through global feature extraction and nonlinear regression, compensates for the influence of the mutual inductance effect of the electromagnet on the magnetic field distribution, and dynamically adjusts the magnetic field calculation results.

7. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 6, characterized in that: The steps of calculating the ratio of the magnetic induction intensity to the peak value of the pulse current at each measuring point and processing the data using a deep learning magnetic field correction model are as follows: The magnetic induction intensity of multiple measuring points is measured. Suppose the magnetic induction intensity of a certain measuring point is B i : B i =f(I p ), Calculate the ratio of the magnetic induction intensity at the measuring point to the peak value of the pulse current. The ratio formula is: R i =B i / I p , Among them, B i represents the magnetic induction intensity at the i-th measuring point, I p Indicates the peak value of the pulse current, R i represents the ratio of the magnetic induction intensity to the peak value of the pulse current at the i-th measurement point, and f(·) is the mapping function of the magnetic induction intensity to the peak value of the pulse current; Train the Transformer magnetic field correction neural network and collect ratio data of multiple measurement points {R i }, forming a magnetic field data set where x i is the spatial coordinate of the measuring point, Use the position encoding function to input the encoding: PE(x i )=sin(ω k x i) +cos(ω k x i) , Among them, PE(x i ) is the position encoding value, ω k is the encoding frequency parameter, x i is the spatial coordinate of the measuring point, The self-attention mechanism is used to calculate the magnetic field distribution relationship, and the calculation formula is: Q=W q R,K=W k R,V=W v R, Calculate the attention weights: Calculate the weighted output: Z=AV, Where Q, K, V are query, key and value matrices respectively, R represents the magnetic induction intensity ratio vector, W q ,W k ,W v is the weight matrix, d k is the dimension of the key vector, A is the attention weight, and Z is the output calculated by Transformer; The multi-layer perceptron MLP is used for nonlinear regression to correct the magnetic field distribution. The correction formula is: in, is the corrected magnetic induction intensity ratio, MLP(·) is the multi-layer perceptron model used for nonlinear regression; The mean square error MSE is used as the loss function: Among them, L is the loss value, N is the total number of data set samples, The Adam optimizer is used to update the network parameters to make the loss function converge.

8. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 7, characterized in that: The steps of using the compressed sensing algorithm to perform interpolation and completion are: Assume that the sparse representation of the magnetic field distribution data B under the basis Ψ is: B=Ψs, Among them, B is the magnetic field spatial distribution data, Ψ is the sparse transformation basis matrix, s is the sparse coefficient vector, Construct the measurement matrix and use the observation matrix Φ for measurement: y=ΦB, Among them, y is the observed data, Φ is the measurement matrix, Recover B through the l1-constrained optimization problem: By solving the sparse vector s, the interpolated magnetic field distribution is calculated: in Represents the spatial distribution data of the magnetic field after compressed sensing interpolation.

9. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 8, characterized in that: In step S5, the spatial configuration of the magnetic field is calculated based on adaptive filtering, specifically: Kalman filtering is used to predict the magnetic field state. Let the magnetic field state vector be x j , its state transfer equation is: x j+1 =F j x j +w j , Among them, x j is the magnetic field state vector at the jth time step, x j+1 is the state of the next time step, F j is the state transfer matrix, w j is the process noise vector, which obeys the zero-mean Gaussian distribution, Q j is the process noise covariance matrix, Combined with the measured data, the magnetic field state is updated. The updating process is: y j =H j x j +v j , Among them, y j is the measurement vector, H j is the observation matrix, v j is the measurement noise vector, which obeys the zero-mean Gaussian distribution, R j is the measurement noise covariance matrix.

10. A method for measuring the spatial configuration distribution of the magnetic field of an electromagnet as claimed in claim 9, characterized in that: The step of calculating the magnetic field spatial configuration based on adaptive filtering also includes: Perform Kalman filter iterative calculation: Prediction steps: in, is the predicted state at the j+1th time step, P j+1|j is the forecast error covariance matrix, Update steps: P j|j =(I-K j H j )P j|j-1 , Among them, K j is the Kalman gain, is the updated magnetic field state, P j|j is the updated error covariance matrix, I is the identity matrix; A multi-order polynomial is used to fit the magnetic field distribution trend, and the formula is: Among them, B trend (x) is the trend term of magnetic field distribution, M is the highest order of fitting, c m is the fitting coefficient, x is the spatial coordinate; Combined with the filtered data, the final magnetic field distribution is calculated: B final (x)=B trend (x)+KF(y), Among them, B final (x) is the final magnetic field spatial configuration distribution, and KF(y) is the measurement data correction term after Kalman filtering.

Images