A magnetic compensation method for aircraft cabin taking into account magnetic interference of electronic equipment

Abstract

The present invention discloses a method for magnetic compensation in an aircraft cabin that takes into account the magnetic interference of electronic equipment, and belongs to the field of aeromagnetic compensation. The method models the onboard electronic (OBE) interference in the cabin as functions proportional to the current of different electronic equipment and their time changes. In addition, in the data processing process, fast Fourier transform and curvature calculation are used to determine the optimal filter cutoff frequency, and the measurement value of useful current or voltage is selected by correlation calculation. In addition, since the extended onboard electronic (OBE) interference in the model additionally aggravates the multicollinearity of the model, the method also uses principal component analysis (PCA) to reduce the impact of the multicollinearity of the model. The method takes into account more interference sources in the cabin environment, improves the compensation accuracy of the model, and reduces the modification cost of the flight platform.

Description

Technical Field

[0001] The invention belongs to the field of aeromagnetic compensation, and in particular relates to a magnetic compensation method in an aircraft cabin taking into account magnetic interference of electronic equipment. Background Art

[0002] Geomagnetic navigation has the advantage of being available at any time and in any weather, and does not rely on satellites or other external communications, which has proven to be a viable alternative to the Global Positioning System (GPS). The basic principle of geomagnetic navigation is to match the real-time magnetic field data collected by the magnetometer with the Earth's magnetic field map. However, the magnetic signals required for geomagnetic navigation are subject to interference from the aircraft platform. In addition, for the sake of flight safety and the convenience of carrying the magnetometer, the magnetometer needs to be placed in the cabin. Existing aeromagnetic compensation methods are mostly based on the classic Tolles–Lawson (TL) model, which takes into account platform interference including permanent magnetic field, induced magnetic field, and eddy current magnetic field, but ignores the interference caused by the onboard electronic (OBE) system. When the magnetometer is placed in the cabin, the impact of OBE interference will be more serious, which reduces the compensation effect of the TL model. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method for magnetic compensation in the aircraft cabin that takes into account the magnetic interference of electronic equipment, and models the OBE interference in the cabin as a function proportional to the current of different electronic equipment and its time change. In addition, in the data processing process, fast Fourier transform and curvature calculation are used to determine the optimal filter cutoff frequency, and useful measurement values ​​are selected by correlation calculation. In addition, since the extended OBE interference in the model additionally aggravates multicollinearity, the method also uses principal component analysis (PCA) to reduce the impact of multicollinearity of the model.

[0004] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] A method for magnetic compensation in an aircraft cabin taking into account magnetic interference of electronic equipment comprises the following steps:

[0006] Step 1, establishing a TL model; including, according to physical principles, dividing the magnetic field interference caused by the aircraft platform itself into inherent magnetic field interference, induced magnetic field interference and eddy current magnetic field interference; in the TL model, the inherent magnetic field interference is modeled as three error parameter items fixed on the aircraft coordinate axis, the induced magnetic field interference is modeled as six error parameter items that are proportional to the size of the external magnetic field and related to the flight attitude, and the eddy current magnetic field interference is modeled as nine error parameter items related to the magnetic field change rate and the flight attitude;

[0007] Step 2, modeling of airborne electronic equipment interference; including, according to the Biot-Savart law and the magnetic effect of the current, projecting the airborne electronic equipment interference component to the direction of the measured magnetic field, and modeling the induced magnetic field interference and eddy current magnetic field interference caused by the airborne electronic equipment interference as error terms related to the current magnitude and the current change rate, which are combined into 6 error parameter terms, and then combined with the 18 error parameter terms in step 1 to form a complete in-cabin magnetic compensation model;

[0008] Step 3, obtaining magnetic compensation calibration data;

[0009] Step 4, performing filtering frequency analysis, current and voltage feature selection and dimension reduction preprocessing on the magnetic compensation calibration data;

[0010] Step 5: Recursive solution of the in-cabin magnetic compensation model parameters and magnetic field interference compensation.

[0011] Furthermore, the step 3 includes calibrating the flight of the aircraft in four directions, completing three sets of actions of pitch, yaw and roll in each direction in turn, and each set of actions is repeated no less than 3 times, wherein the standard amplitude of the pitch action is ±5°, the standard amplitude of the yaw action is ±10°, and the standard amplitude of the roll is ±5°, and periodically engaging and separating the airborne electronic equipment in each direction in turn, and there is a preset time interval between the engaging and separating of the airborne electronic equipment, so as to obtain the simulated state signal of the aircraft and the magnetic field change data.

[0012] Furthermore, step 4 includes fitting the spectrum curve of the magnetic field interference signal with an inverse proportional function to obtain a spectrum fitting curve, and then calculating the maximum curvature point of the spectrum fitting curve, which is the optimal low-frequency filter cutoff frequency; performing correlation calculation on the current measurement data and scalar magnetic field measurement data processed by the bandpass filter using the optimal low-frequency filter cutoff frequency, and after uniform scaling, evaluating the contribution of each current measurement data and each scalar magnetic field measurement data by the softmax function, and finally selecting a useful in-cabin magnetic compensation model input according to the contribution ranking and proportion of different electronic devices; decentralizing the input data matrix, and calculating the eigenvalues ​​and eigenvectors of the covariance matrix, sorting the eigenvalues ​​of the covariance matrix by size, and selecting the eigenvectors corresponding to the eigenvalues ​​that meet the empirical threshold to form a dimensionality reduction matrix;

[0013] Furthermore, the step 5 includes recursively solving the in-cabin magnetic compensation model parameters using the in-cabin magnetic compensation calibration data obtained and preprocessed in steps 3-4 as input. The recursive solution process is based on a recursive least squares algorithm, and regularization terms of varying degrees are added based on the idea of ​​elastic weight solidification to achieve iterative update of the in-cabin magnetic compensation model parameters, and then compensation for magnetic field interference is achieved based on the solved in-cabin magnetic compensation model parameters through the in-cabin magnetic compensation model in steps 1-2.

[0014] The beneficial effects of the present invention are:

[0015] Firstly, based on the traditional aeromagnetic compensation model, this method takes into account the situation where the magnetometer is installed inside the cabin, avoiding the need to mount the magnetometer outside the cabin in the form of a rod in the traditional aeromagnetic compensation method. This not only reduces the modification cost of the flight platform, but also improves the safety of flight.

[0016] Secondly, based on the classic aeromagnetic compensation model, this method adds the modeling of interference from airborne electronic equipment and the induced magnetic field and eddy current magnetic field caused by it, and considers more interference sources in the cabin environment to improve the compensation accuracy of the model;

[0017] In addition, this method determines the filter cutoff frequency by curve fitting and curvature calculation, selects useful model inputs by correlation and contribution calculation, and uses principal component analysis to reduce the dimension of model input features to reduce the impact of model multicollinearity;

[0018] Finally, based on the recursive least squares method, different regularizations are added to different parameters to control the updating direction of the parameters and achieve adaptive optimization of the model parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a magnetic compensation method for an aircraft cabin that takes into account magnetic interference of electronic equipment according to the present invention;

[0020] Figure 2 Schematic diagram of the carrier coordinate system of an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0022] like Figure 1As shown in the figure, a flow chart of a magnetic compensation method for aircraft cabin that takes into account the magnetic interference of electronic equipment is given: first, by establishing a TL model and modeling the interference of airborne electronic equipment, and then obtaining magnetic compensation data, the parameters of the in-cabin magnetic compensation model are solved, including fast Fourier transform of the preliminary processing results of the TL model, and then curve fitting and curvature calculation of the obtained spectrum diagram to determine the cutoff frequency, that is, the optimal cutoff frequency of the bandpass filter. Then, the correlation and contribution calculation of the magnetic measurement data and current data processed by bandpass filtering are performed to obtain the best feature selection as the input for model improvement. Then, the current characteristics of the airborne electronic equipment are combined with the classical aeromagnetic compensation model, and the principal component analysis method is used to reduce the dimension of the input features of the model (i.e., PCA dimension reduction). Finally, the final compensation result is obtained through the adaptive optimization of the parameters of the in-cabin magnetic compensation model.

[0023] The specific principles of each step are introduced in detail below.

[0024] Step 1: Establish TL model;

[0025] The measurement of the geomagnetic field and the analysis of the carrier magnetic field interference are all based on the carrier coordinate system. The schematic diagram of the carrier coordinate system is shown in Figure 2 As shown, the carrier coordinate system is defined as: the Y axis along the aircraft's forward direction, the X axis along the left wing, and the Z axis perpendicular to the ground and downward, and the three axes are perpendicular to each other. To measure the magnetic field, α, β, and γ are the angles between the measured magnetic field vector and the three axes of the carrier coordinate system;

[0026] According to the classic TL model, the carrier magnetic field interference can be equivalent to the superposition of the magnetic fields generated by several magnetic dipoles on the magnetic sensor, that is, the measured magnetic field It can be regarded as the Earth's magnetic field Interfering magnetic field with carrier The vector sum of :

[0027] (1)

[0028] Among them, the carrier interferes with the magnetic field It is mainly composed of three types of magnetic field interference: inherent magnetic field interference , Inductive magnetic field interference , eddy current magnetic field interference :

[0029] (2)

[0030] The inherent magnetic field interference can be regarded as a constant in a short time, while the induction magnetic field interference and eddy current magnetic field interference depend on the size and direction of the external excitation magnetic field. When the magnetometer is installed outside the cabin, the excitation magnetic field can be regarded as the geomagnetic field because the amplitude of the interference of the onboard electronic equipment is much smaller. Based on this assumption, the measured magnetic field modulus It can be approximated as:

[0031] (3)

[0032] in, Represents the modulus of a vector. The vector interferes with the magnetic field To project the direction of the measured magnetic field, the direction cosines of the measured magnetic field in the carrier coordinate system are required:

[0033] (4)

[0034] in, is a column vector of direction cosines, is the three-axis unit vector in the carrier coordinate system, To measure the angle between the magnetic field vector and the three axes of the carrier coordinate system, the subscript T is the transpose of the matrix or vector. Then, the standard Tolles-Lawson model can be summarized as:

[0035] (5)

[0036] in, are the 18 unknown parameters to be solved in the model, is the derivative of the column vector composed of the direction cosine column, which is implemented in the form of difference in the actual processing. The simplified TL model can be expressed as:

[0037] (6)

[0038] Where A is the input data matrix consisting of the cross terms of geomagnetic measurements and direction cosines and their derivatives. is a column vector consisting of 18 unknown parameters to be solved. In order to ensure that the TL model parameters are solvable, both sides of the equation need to be bandpass filtered at the same time:

[0039] (7)

[0040] Where bpf is a bandpass filter. In practical applications, the frequency range of the bandpass filter is 0.1-0.9 Hz. The data used to calculate the model coefficients comes from a high-altitude calibration flight, which contains a specific set of roll, pitch and yaw aircraft maneuvers. The model compensation coefficients are then calculated using ridge regression:

[0041] (8)

[0042] Among them, the corner mark -1 is the matrix inversion operation, To measure the magnetic field modulus, The bandpass filtering result is , is the bandpass filtering result of the input data matrix A, that is . is the identity matrix, is a hyperparameter that can be determined by the ridge trace method in the ridge regression algorithm. Finally, the modulus of the geomagnetic field can be calculated as:

[0043] (9)

[0044] Step 2: Modeling of interference of airborne electronic equipment;

[0045] According to the Biot-Savart law, OBE interference can be regarded as the current flowing through the electronic device. Therefore, it is reasonable to model OBE interference using current observation.

[0046] (10)

[0047] In the formula, For OBE interference, is the three-axis unit vector in the carrier coordinate system, k is the label of the electronic device, is the current measurement value. To compensate for the unknown parameters of the model. Similarly, the interference component needs to be projected to the measured magnetic field direction for analysis:

[0048] (11)

[0049] As for the excitation source of interference, in the TL model, the magnitude of the induced magnetic field and the eddy current magnetic field mainly depends on the external geomagnetic field. When the magnetometer is installed in the cabin, the magnetic field caused by the electronic equipment in the cabin should also become part of the excitation source of the interference magnetic field. The induced magnetic field caused by OBE interference can be calculated as:

[0050] (12)

[0051] in, represents the induced magnetic field intensity coefficient caused by OBE interference, where m is used to distinguish the three-axis distribution of OBE interference and n is used to distinguish the three-axis distribution of the generated induced magnetic field. In addition, the eddy current magnetic field caused by the time variation of OBE interference can be calculated as:

[0052] (13)

[0053] in, is the time variation of current, is the eddy current interference compensation coefficient caused by OBE interference. Similarly, after projecting the above-mentioned induced magnetic field interference and eddy current magnetic field interference into the carrier coordinate system, we can get:

[0054] (14)

[0055] in, is the modulus of the magnetic field interference caused by the kth electronic device, is the unknown OBE interference model parameter. According to Ohm's law, the current in the wire is usually proportional to the voltage. In actual data processing, the voltage measurement value can also be used to represent the OBE interference characteristics, replacing the current measurement value in the above formula. The OBE interference caused by the kth electronic device in the cabin and its additional interference can be characterized as:

[0056] (15)

[0057] in, is the current characteristic matrix composed of the cross terms of the current and its derivative and the direction cosine, There are 6 unknown parameters The parameter vector composed of . At this time, the measurement value composition of the in-cabin magnetometer can be expressed as:

[0058] (16)

[0059] In order to facilitate parameter calculation, the attitude information matrix can be combined with the current interference matrix. Then the compensation model can be summarized as:

[0060] (17)

[0061] in, is the input data matrix A and the current characteristic matrix The result of the merger, is the parameter vector to be determined and After merging, the bandpass filtered model can be expressed as:

[0062] (18)

[0063] in, To measure the magnetic field modulus The result of bandpass filtering is , is the combined input feature matrix The result of bandpass filtering is .

[0064] Step 3: Obtain magnetic compensation calibration data;

[0065] After the UAV magnetic survey platform is assembled, the compensation coefficient in the model needs to be solved according to the interference compensation model to achieve real-time compensation of the interference field in the aeromagnetic measurement.

[0066] In the compensation calibration of attitude disturbance, in order to obtain effective attitude disturbance compensation data, the aircraft compensation calibration flight needs to be carried out according to certain requirements. The compensation flight needs to be carried out in four directions, and each direction must complete three groups of actions in turn: pitch, yaw and roll, and each group of actions must be repeated no less than 3 times. The standard amplitude of the pitch action is "±5°", the standard amplitude of the yaw is "±10°", and the standard amplitude of the roll is "±5°".

[0067] The magnetic compensation of airborne electronic equipment can be calibrated on the ground or in the air. The experiments are carried out in the north-south and east-west directions. The electronic equipment is periodically engaged and separated in each direction. There must be a certain time interval between the engagement and separation of the electronic equipment to obtain the analog status signal (usually voltage / current) and magnetic field change data of the equipment in the calibration experiment.

[0068] Step 4: Perform filtering frequency analysis, current and voltage feature selection and dimension reduction preprocessing on the magnetic compensation calibration data;

[0069] Since the current characteristics of multiple electronic devices are additionally introduced into the model, the complex collinearity problem of the model itself is aggravated. Therefore, principal component analysis (PCA) is used as a preprocessing step for coefficient estimation to decorrelate and reduce the dimension of the input features before parameter solution. The feature matrix after bandpass filtering is rewritten into vector form:

[0070] (19)

[0071] in, are the column vectors of the matrix, and N is the number of columns of the matrix. The matrix is ​​then centered:

[0072] (20)

[0073] Among them, s and t are used to distinguish different column vectors. Then, the eigenvalues ​​and eigenvectors of the covariance matrix are calculated and arranged in order from large to small. , where N is the number of columns in the matrix, that is, the number of original dimensions, is the characteristic matrix corresponding to each eigenvector, and then the corresponding eigenvector is obtained , taking this opportunity, the dimension N of the matrix can be reduced to , it must meet the following conditions:

[0074] (twenty one)

[0075] in, is the eigenvalue with the highest order, and d is the number of the eigenvalues ​​sorted from large to small. is an empirical threshold determined by experiments. Then, the order is The eigenvectors of the reduced dimension matrix are The dimensionality reduction process can be expressed as:

[0076] (twenty two)

[0077] in, is the input data matrix after dimensionality reduction, and then the ridge regression estimation can be used to calculate the compensation parameters:

[0078] (twenty three)

[0079] Since the interference signal needs to be analyzed in the frequency domain to determine the optimal filter cutoff frequency of the bandpass finite impulse response filter. The manual selection of the cutoff frequency from the spectrum diagram through fast Fourier transform is highly subjective, and the spectrum shape of the geomagnetic field signal is similar to an inverse function, which is doped with many burrs. Therefore, the curve fitting method of the spectrum curve is adopted here to determine the optimal cutoff frequency through curvature calculation. Assume that the selected fitting curve is:

[0080] (twenty four)

[0081] in, and is the unknown curve fitting parameter. The curvature of the discrete points is calculated as follows:

[0082] Where and are unknown curve fitting parameters. The curvature of discrete points is calculated as follows:

[0083] (25)

[0084] in, and are the first and second derivatives of the fitting function, respectively. is the calculated curvature value. The first and second derivatives of the fitting function are calculated as follows:

[0085] (26)

[0086] (27)

[0087] Where e is the subscript of the discrete data point. In addition, it is necessary to select useful signals from a series of current and voltage measurements. Useful features are selected by calculating the following correlation coefficient:

[0088] (28)

[0089] Where a and b represent the magnetic field measurement value and current measurement value respectively. And, , represents the average of the measurements. After uniformly scaling the correlation coefficient, the contribution of each measurement can be evaluated by the softmax function as follows:

[0090] (29)

[0091] Among them, r k is the correlation coefficient for different electronic devices, so that all measurements contribute The sum of is 1. We can choose useful metrics by ranking and contribution ratio.

[0092] Step 5: Recursive solution of the in-cabin magnetic compensation model parameters and magnetic field interference compensation.

[0093] In previous research work, the solution of TL model parameters is usually solved as a post-processing problem, that is, after all data are collected, offline batch processing is performed to solve the corresponding parameters. However, we prefer to use a sequential analysis method to process the acquired data online. Obviously, this method can avoid a large amount of data storage, saving time and effort. At the same time, it can update the parameters online according to the actual situation and has a certain degree of adaptive ability to environmental changes.

[0094] Recursive least squares (RLS) is a classic sequential parameter solution algorithm. This algorithm uses the sum of squared errors as the cost function and uses the orthogonality principle and sequential analysis data processing method to solve the model parameters. Each iteration only inputs a set of data to update the old parameters, and avoids complex matrix inversion calculations through the matrix inversion lemma, and converges quickly.

[0095] The recursive least squares algorithm is as follows:

[0096] Table 1 Recursive least squares method

[0097] ,

[0098] However, since the forgetting factor λ in recursive least squares has the same regularization constraint for all model parameters, it does not have the ability to distinguish each model parameter. Elastic Weight Consolidation (EWC) is one of the classic algorithms for continuous learning. The main idea is to estimate the importance of each parameter and add regularization terms with different weights to each parameter during the parameter update process, thereby controlling the update direction of the parameters and achieving the effect of suppressing "catastrophic forgetting". This method is mainly divided into two steps: (1) calculating the importance of each parameter to the old task; (2) applying regularization terms with different weights to each parameter to control the direction of parameter update.

[0099] For the model training process, given a data set D, the criterion for finding the optimal parameters is to maximize the posterior probability of the parameters, that is:

[0100] (30)

[0101] Where P is the probability, is the model parameter to be solved. In order to facilitate calculation, it is generally taken logarithm, as shown below:

[0102] (31)

[0103] For the sequential learning process of two tasks, let the old task be A, the new task be B, and their corresponding data sets be and , that is, divide the data set D into two, and and Independent of each other. At this time, the model parameters to be solved The posterior probability can be written as:

[0104] (32)

[0105] in, It can be deduced as:

[0106] (33)

[0107] Therefore, the posterior probability of the parameter is:

[0108] (34)

[0109] because and Independent, yes , , and then we get:

[0110] (35)

[0111] Taking the logarithm of both sides at the same time, we can get the final parameter posterior probability formula, which is also the core of the entire EWC method:

[0112] (36)

[0113] The first term on the right side of the equal sign can be understood as the loss function obtained for task B, denoted as , the third term and parameter It is irrelevant and can be ignored. The objective function of model optimization is:

[0114] (37)

[0115] Right now:

[0116] (38)

[0117] In optimizing the posterior probability Before, assume that the prior probability is a Gaussian distribution, that is ,in, For expectations, is the variance, and its probability distribution is:

[0118] (39)

[0119] Taking the logarithm, we get:

[0120] (40)

[0121] The logarithm of the prior probability is ,Right now ; The optimal solution of task A is , then .right exist Performing a second-order Taylor expansion at , we can obtain:

[0122] (41)

[0123] in, is the second-order derivative, is a high-order infinitesimal quantity expanded by Taylor series. In the above formula, the first term on the right side of the equal sign is a constant, and the second term is 0. It is not difficult to obtain:

[0124] (42)

[0125] Right now Assumptions In line with the uniform distribution, Is a constant, according to the Bayesian formula:

[0126] (43)

[0127] The posterior probability also approximately follows a Gaussian distribution: .

[0128] Therefore, the optimization objective function can be rewritten as:

[0129] (44)

[0130] If the M-dimensional parameter vector The components in are independent of each other and can be expanded as follows:

[0131] (45)

[0132] Among them, t is used to distinguish each component in the M-dimensional parameter vector, For the dataset D A The tth model parameter solved on It is just the diagonal element of the Fisher information matrix. In the actual calculation process, the Fisher information matrix item of each parameter is generally obtained by calculating the square of the parameter gradient, as shown below:

[0133] (46)

[0134] Among them, N data is the data length, and the part in brackets uses the gradient of the parameters to replace the approximate calculation.

[0135] In the problem of aeromagnetic error compensation, we use the TL model to sequentially process the data collected from two sorties. The calibration flight process completed in advance can be regarded as task A, and the subsequent level flight process can be regarded as task B. The processing is divided into two stages. By integrating the above EWC idea into the RLS algorithm, the following fused parameter solution method can be obtained:

[0136] Table 2 Recursive least squares method with EWC idea

[0137] ,

[0138] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A magnetic compensation method for aircraft cabin taking into account magnetic interference of electronic equipment, characterized in that: The steps include: Step 1: Establish a TL model; Including, according to the physical principles, the magnetic field interference caused by the aircraft platform itself is divided into inherent magnetic field interference, induced magnetic field interference and eddy current magnetic field interference; In the TL model, the inherent magnetic field interference is modeled as three error parameter items fixed on the aircraft coordinate axes, the induced magnetic field interference is modeled as six error parameter items that are proportional to the magnitude of the external magnetic field and related to the flight attitude, and the eddy current magnetic field interference is modeled as nine error parameter items related to the magnetic field change rate and the flight attitude. Step 2: Modeling of interference of airborne electronic equipment; Including, according to Biot-Savart law and the magnetic effect of current, projecting the interference component of airborne electronic equipment to the direction of the measured magnetic field, and modeling the induced magnetic field interference and eddy current magnetic field interference caused by the airborne electronic equipment interference as error terms related to the current magnitude and the current change rate, which are combined into 6 error parameter terms, and then combined with the 18 error parameter terms in step 1 to form a complete in-cabin magnetic compensation model; Step 3, obtaining magnetic compensation calibration data; Step 4, performing filtering frequency analysis, current and voltage feature selection and dimension reduction preprocessing on the magnetic compensation calibration data; Step 5, recursive solution of the in-cabin magnetic compensation model parameters and compensation for magnetic field interference, including: taking the in-cabin magnetic compensation calibration data obtained and preprocessed in steps 3-4 as input to recursively solve the in-cabin magnetic compensation model parameters, the recursive solution process is based on the recursive least squares algorithm, and based on the idea of ​​elastic weight solidification, regularization terms of varying degrees are added to achieve iterative update of the in-cabin magnetic compensation model parameters, and then based on the solved in-cabin magnetic compensation model parameters, compensation for magnetic field interference is achieved through the in-cabin magnetic compensation model in steps 1-2.

2. The method for magnetic compensation in an aircraft cabin taking into account magnetic interference of electronic equipment according to claim 1, characterized in that: The step 3 includes calibrating the flight of the aircraft in four directions, completing three sets of actions of pitch, yaw and roll in each direction in turn, and each set of actions is repeated no less than 3 times, wherein the standard amplitude of the pitch action is ±5°, the standard amplitude of the yaw action is ±10°, and the standard amplitude of the roll is ±5°, and periodically engaging and separating the airborne electronic equipment in each direction in turn, and there is a preset time interval between the engaging and separating of the airborne electronic equipment, so as to obtain the simulated state signal of the aircraft and the magnetic field change data.

3. The method for magnetic compensation in an aircraft cabin taking into account magnetic interference of electronic equipment according to claim 2, characterized in that: The step 4 includes fitting the spectrum curve of the magnetic field interference signal with an inverse proportional function to obtain a spectrum fitting curve, and then calculating the maximum curvature point of the spectrum fitting curve, which is the optimal low-frequency filter cutoff frequency; performing correlation calculation on the current measurement data and the scalar magnetic field measurement data processed by the bandpass filter with the optimal low-frequency filter cutoff frequency, and evaluating the contribution of each current measurement data and each scalar magnetic field measurement data through a softmax function after uniform scaling, and finally selecting a useful in-cabin magnetic compensation model input according to the contribution ranking and proportion of different electronic devices; decentralizing the input data matrix, and calculating the eigenvalues ​​and eigenvectors of the covariance matrix, sorting the eigenvalues ​​of the covariance matrix by size, and selecting the eigenvectors corresponding to the eigenvalues ​​that meet the empirical threshold to form a dimensionality reduction matrix.

Images