CN116299735A - Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network - Google Patents
Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network Download PDFInfo
- Publication number
- CN116299735A CN116299735A CN202211096456.3A CN202211096456A CN116299735A CN 116299735 A CN116299735 A CN 116299735A CN 202211096456 A CN202211096456 A CN 202211096456A CN 116299735 A CN116299735 A CN 116299735A
- Authority
- CN
- China
- Prior art keywords
- neural network
- geomagnetic vector
- measurement system
- model
- magnetic field
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005291 magnetic effect Effects 0.000 title claims abstract description 117
- 238000005259 measurement Methods 0.000 title claims abstract description 87
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 63
- 238000003062 neural network model Methods 0.000 claims abstract description 56
- 238000012360 testing method Methods 0.000 claims abstract description 30
- 210000002569 neuron Anatomy 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 11
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 230000002068 genetic effect Effects 0.000 claims description 6
- 230000036962 time dependent Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims 1
- 230000005358 geomagnetic field Effects 0.000 description 12
- 238000013507 mapping Methods 0.000 description 9
- 238000012937 correction Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 3
- 230000036961 partial effect Effects 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000005294 ferromagnetic effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000005669 field effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000000696 magnetic material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V3/00—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation
- G01V3/40—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation specially adapted for measuring magnetic field characteristics of the earth
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V3/00—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation
- G01V3/38—Processing data, e.g. for analysis, for interpretation, for correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Remote Sensing (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- Geophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Measuring Magnetic Variables (AREA)
Abstract
The invention discloses an interference magnetic field compensation method of a geomagnetic vector measurement system based on a BP neural network, which comprises the following steps: s01, placing a geomagnetic vector measurement system at the center of a three-dimensional Helmholtz coil, and generating different magnetic vector fields in a central uniform area of the three-dimensional Helmholtz coil by controlling current flowing through the coil; s02, collecting representative magnetic vector data generated by the three-dimensional Helmholtz coil, and constructing a training data set and a testing data set; s03, training the BP neural network by using a training sample data set to obtain a trained BP neural network model, and testing the model by using a test data set until a model meeting the requirements is obtained; s04, obtaining a measured value of a geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using a BP neural network model. The method has the advantages of simple implementation method, strong operability, complete compensation model, high compensation precision and the like.
Description
Technical Field
The invention relates to the technical field of geomagnetic vector measurement, in particular to an interference magnetic field compensation method of a geomagnetic vector measurement system based on a BP neural network.
Background
Geomagnetic vector (north, east and vertical) measurement is widely used in the fields of surveying and searching for rescue, mineral exploration, vehicle detection, unmanned equipment positioning, geomagnetic navigation, etc. The geomagnetic field vector field can be measured by combining the triaxial magnetic sensor with an Inertial Navigation System (INS), the inertial navigation system is used for providing attitude information for the triaxial magnetic sensor, the vector is converted into a geographic coordinate system by utilizing the attitude information, and the geomagnetic field vector measuring system is the geomagnetic field vector measuring system. However, because of magnetic interference generated by the ferromagnetic component itself and other electrical devices such as an inertial element and a power circuit module, a large error exists between the measured value and the actual value, which becomes a main factor affecting the geomagnetic field vector measurement accuracy. Therefore, the magnetic field interference field compensation technology is a key to improve the measurement accuracy.
The main factors affecting the magnetic interference compensation performance of the geomagnetic vector measurement system are as follows: (i) A magnetic interference compensation model, namely, a mathematical formula is used for representing a magnetic interference source; (ii) a compensation strategy, i.e. how to build an error model equation; (iii) The compensation parameter estimation algorithm is how to accurately estimate the error parameters. Aiming at the correction compensation problem of the geomagnetic vector measurement system, the following correction compensation modes are generally adopted in the prior art:
(1) Based on the vector compensation model and taking the eddy current field into account to achieve correction compensation, however, the magnetic disturbance is very complex, and some other types of magnetic disturbance, such as power equipment and current magnetic fields, cannot be characterized at all.
(2) The compensation strategy based on the rotating platform comprises the construction of equations using three different attitude rotation strategies (symmetric rotation, orthogonal rotation and random rotation), wherein the compensation effect of the symmetric rotation strategy is optimal since the selected measurement positions are representative and cover the whole attitude space. However, this type of method needs a rotating geomagnetic vector measurement system, and has the problem of sensitivity to geomagnetic field gradients and environmental geomagnetic interference.
(3) Error parameters in the geomagnetic field vector measurement component compensation model are estimated by using a Lagrangian multiplier method to realize compensation, but a vortex field is not considered in the component compensation model.
In the prior art, the correction compensation for the geomagnetic vector measurement system is usually incomplete, and the following problems exist:
1. the eddy current field or other disturbance sources are not considered, but in moving geomagnetic vector measurement, the eddy current field is not ignored.
2. Only linear errors are considered and non-linear errors are not considered, for example, it is assumed that the soft magnetic disturbance is linearly related to the external geomagnetic field, the soft magnetic disturbance coefficient is a matrix of 3x3, but due to the complexity of the soft magnetic material characteristics, there is a certain error in characterizing it with a linear model, which also has a great influence on the measurement accuracy, and it cannot be completely characterized with a linear error model, where the cross field effect and hysteresis are two important factors that lead to non-linearity. Conventional compensation models typically treat the error parameter as a constant parameter, while in fact some sources of magnetic interference are nonlinear due to crossed fields and hysteresis effects, etc.
In summary, in the correction and compensation scheme for the geomagnetic vector measurement system in the prior art, the compensation model is usually incomplete, the vortex field or other interference sources are not considered, and the nonlinear error is not considered, so that the nonlinear network mapping relationship between the measured value (to be compensated) and the true value cannot be accurately described, the compensation precision is low, the compensation process is complex, and the operability is not strong in practice.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems existing in the prior art, the invention provides the geomagnetic vector measurement system interference magnetic field compensation method based on the BP neural network, which has the advantages of simple implementation method, strong operability, low cost, complete compensation model and high compensation precision.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a method for compensating an interference magnetic field of a geomagnetic vector measurement system based on a BP neural network comprises the following steps:
s01, placing a geomagnetic vector measurement system at the center of a three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling current flowing through the coil;
s02, collecting representative magnetic vector data generated by a three-dimensional Helmholtz coil, and constructing a training data set and a testing data set for forming a model;
s03, training the BP neural network by using the training sample data set, training to obtain a BP neural network model, and testing the BP neural network model obtained by training by using the test data set until the BP neural network model meeting the requirements is obtained;
s04, obtaining a measured value of a geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the BP neural network model obtained in the step S03 so as to realize interference magnetic field compensation of the geomagnetic vector measurement system.
Further, in the step S03, a genetic algorithm (genetic algorithm, GA) is used to optimize the BP neural network.
Further, the input layer of the BP neural network comprises 3 neurons for inputting the geomagnetic vector measurement system output value H m ＝[H mx ,H my ,H mz ] T ，H mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m The output layer includes 3 neurons for outputting a true geomagnetic vector value h= [ H ] generated by the three-dimensional helmholtz coil x ,H y ,H z ] T ，H x ,H y ,H z Three components of H, respectively.
Further, the hidden layer of the BP neural network is set to be 2 layers or 3 layers, and the number of neurons in the hidden layer is calculated according to the following formula:
wherein m and n respectively represent the number of neurons in an input layer and an output layer in the BP neural network, and beta is a constant between 1 and 10.
Further, the step S03 further includes a model evaluation and optimization step, when the model training process is finished, evaluating whether the BP neural network model obtained by training meets a preset requirement, if yes, obtaining a final BP neural network model, otherwise, adjusting the structure and parameters of the BP neural network model until the BP neural network model meets the preset requirement.
Further, in the model evaluation and optimization step, a root mean square error RMSE between a geomagnetic vector true value and a geomagnetic vector value predicted by a BP neural network model is used as a performance index to evaluate compensation and generalization performance of the BP neural network model, and a calculation expression of the root mean square error RMSE is as follows:
wherein H is actual Is the true value of geomagnetic vector, H predicted Geomagnetic vector values predicted by the BP neural network model are obtained, and N is the sampling number.
Further, the training sample set includes different H' s m And H, a time-dependent magnetic vector field, H m The output value of the geomagnetic vector measurement system is H, which is the real geomagnetic vector value generated by the three-dimensional Helmholtz coil, H m Time-dependent changes
From the measurement data of the triaxial magnetometer, H is varied over time>
Obtained from three-dimensional Helmholtz coils and current, wherein H m Variation over time->
And H>
The method comprises the following steps of:
wherein H is mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m Is H x ,H y ,H z Three components of the true geomagnetic vector value H generated by the three-dimensional Helmholtz coil are respectively, and delta represents the variation.
Further, in the step S02, when the representative magnetic vector data generated by the three-dimensional helmholtz coil is collected, the current sequences of the three orthogonal coils are controlled to generate different directions and magnitudes in the three-dimensional spherical involute.
Further, the three orthogonal coil current sequences are obtained specifically according to the following spherical involute equation:
wherein R represents the radial direction of the involute, θ represents the involute expansion angle, and α represents the involute pressure angle.
An interference magnetic field compensation system of a geomagnetic vector measurement system based on a BP neural network, comprising:
the test control system is used for placing the geomagnetic vector measurement system at the center of the three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling the current flowing through the coil;
the data collection module is used for collecting representative magnetic vector data generated by the three-dimensional Helmholtz coil and constructing a training data set and a testing data set for forming a model;
the model training module is used for training the BP neural network by using the training sample data set, obtaining a BP neural network model by training, and testing the BP neural network model obtained by training by using the test data set until obtaining the BP neural network model meeting the requirements;
the magnetic field compensation module is used for obtaining the measured value of the geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the obtained BP neural network model so as to realize the interference magnetic field compensation of the geomagnetic vector measurement system.
Compared with the prior art, the invention has the advantages that:
1. according to the invention, by utilizing the characteristics of the BP neural network, the BP neural network is used as a compensation model to establish a nonlinear network mapping relation between a value to be compensated and a true value, so that the interference magnetic field compensation is realized, as many interference sources as possible can be considered in the compensation model, and a compensation model with integrity is constructed, so that the problem that the traditional compensation model is incomplete is solved, the accuracy of the interference magnetic field compensation is improved, and the compensation process is simple and the efficiency is high.
2. The three-dimensional Helmholtz coil with the high-precision current source is combined to generate the magnetic field with any direction and amplitude, so that a sufficient neural network model data set can be conveniently obtained for model training, the data generated by the three-dimensional Helmholtz coil can be ensured to be representative in a three-dimensional space, the model precision is further ensured, meanwhile, personalized flexible compensation can be realized, an equation set for establishing an error model by a rotation measurement system is not needed, and the problem of sensitivity to geomagnetic gradients and environmental geomagnetic interference can be avoided.
3. The invention forms a compensation model based on the BP neural network, has simple operation and low cost, and can realize the compensation correction of the disturbing magnetic field of the geomagnetic vector measurement system in various environments.
Drawings
Fig. 1 is a schematic flow chart of an implementation of an interference magnetic field compensation method of a geomagnetic vector measurement system based on a BP neural network in the embodiment.
Fig. 2 is a schematic diagram of the structural principle of the present embodiment for obtaining model training and test samples using three-dimensional helmholtz coils.
Fig. 3 is a schematic diagram of the three-direction current confinement principle of the helmholtz coil obtained in a specific application example.
Fig. 4 is a schematic diagram of a three-directional partial current sequence of a helmholtz coil obtained in a specific application example.
Fig. 5 is a schematic diagram of the BP neural network employed in the present embodiment.
Fig. 6 is a detailed flow chart of the present invention for implementing the disturbing magnetic field compensation of geomagnetic vector measurement system in a specific application embodiment.
Detailed Description
The invention is further described below in connection with the drawings and the specific preferred embodiments, but the scope of protection of the invention is not limited thereby.
The neural network has strong nonlinear model learning and mapping capability, and a three-layer BP neural network can fit any continuous function, so that the BP neural network can be suitable for establishing an error compensation model. According to the invention, the characteristics of the BP neural network are utilized, the BP neural network is used as a compensation model to establish a nonlinear network mapping relation between a measured value (to be compensated) and a true value, so that the interference magnetic field compensation is realized, the problem of magnetic interference in a geomagnetic field vector measurement system is solved, as the BP neural network has strong nonlinear mapping capability, as many interference sources as possible can be considered in the compensation model, the problem of incomplete traditional compensation model is solved, meanwhile, a three-dimensional Helmholtz coil with a high-precision current source is combined to generate a magnetic field with any direction and amplitude, so that a sufficient neural network model dataset can be conveniently obtained for model training, the data generated by the three-dimensional Helmholtz coil can be ensured to be represented in a three-dimensional space, thereby ensuring model precision, and meanwhile, personalized flexible compensation can be realized, an equation set of an error model is not required to be established by means of a rotation measurement system, and the problem of sensitivity to geomagnetic field gradient and environmental geomagnetic interference can be avoided.
As shown in fig. 1, the steps of the interference magnetic field compensation method of the geomagnetic vector measurement system based on the BP neural network in this embodiment include:
s01, placing a geomagnetic vector measurement system at the center of a three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling current flowing through the coil;
s02, collecting representative magnetic vector data generated by a three-dimensional Helmholtz coil, and constructing a training data set and a testing data set for forming a model;
s03, training the BP neural network by using a training sample data set, training to obtain a BP neural network model, and testing the BP neural network model obtained by training by using a test data set until the BP neural network model meeting the requirements is obtained;
s04, obtaining a measured value of a geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the BP neural network model obtained in the step S03 so as to realize interference magnetic field compensation of the geomagnetic vector measurement system.
As shown in fig. 2, in this embodiment, the geomagnetic vector measurement system is located in a central area of a three-dimensional helmholtz coil, the three-dimensional helmholtz coil is composed of three completely orthogonal coils, the three-dimensional helmholtz coil is driven by currents, each coil is provided with a corresponding high-precision current controller, by controlling the currents flowing through the coils, arbitrary magnetic field components can be generated in a central uniform area of the three-dimensional coil, and specifically, magnetic vector fields with different amplitudes, different directions and different change rates can be generated in the central uniform area of the three-dimensional helmholtz coil according to actual requirements. In a specific application embodiment, three sensitive axis directions of the triaxial sensor in the geomagnetic vector measurement system should be aligned with three directions of the three-dimensional helmholtz coil, and a uniform magnetic field area of the three-dimensional helmholtz coil should be as large as possible and can generate a high-precision stable magnetic field.
In this embodiment, the output value H of the geomagnetic vector measurement system is obtained by generating magnetic vector fields with different amplitudes, different directions, and different rates of change in a uniform region of the three-dimensional helmholtz coil m ＝[H mx ,H my ,H mz ] T And is used as the input of the BP neural network, so that the input layer of the BP neural network selects 3 neurons; real geomagnetic vector value H= [ H ] generated by three-dimensional Helmholtz coil x ,H y ,H z ] T As an output of the BP neural network, the output layer of the BP neural network selects 3 neurons.
Considering that the vortex field and some low-frequency magnetic fields are complex, the direction, the magnitude and the time variation of the earth magnetic field are determined, and particularly the vortex field cannot be ignored in the mobile geomagnetic vector measurement, the training sample set in the embodiment also comprises different H m Time-dependent changes
And H>
To enable based on the magnetic vector field ofThe BP neural network simultaneously considers the vortex field, the low-frequency magnetic field and the like to form a more complete compensation model. The data set is obtained and then divided into a training set and a testing set so as to respectively perform model training and model testing. H m Variation over time->
And H>
Can be represented by formula (1):
wherein H is mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m Is H x ,H y ,H z Three components of the true geomagnetic vector value H generated by the three-dimensional Helmholtz coil are respectively, and delta represents the variation.
Above H m Time-dependent changes
The method can be precisely determined according to the measurement data of a triaxial magnetometer, and H is changed along with time +.>
The three-dimensional Helmholtz coil and the three-dimensional current can be precisely determined.
The structure of the BP neural network is a main factor affecting the performance, as shown in FIG. 5, the input layer of the BP neural network in this embodiment specifically includes 3 neurons for inputting the output value H of the geomagnetic vector measurement system m ＝[H mx ,H my ,H mz ] T ，H mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m The output layer includes 3 neurons for outputting the true geomagnetism produced by the three-dimensional helmholtz coilVector value h= [ H ] x ,H y ,H z ] T ，H x ,H y ,H z Three components of H, respectively. Meanwhile, too many hidden layers and nodes in the neural network may cause excessive fitting, while too few hidden layers and nodes in the neural network may affect mapping capability, under the condition of meeting the requirement of compensation precision, the hidden layers and the number of nodes of the neural network should be as small as possible, the hidden layers of the BP neural network in this embodiment are set to be 2 layers or 3 layers, and the number of neurons in the hidden layers is calculated specifically according to the following formula:
wherein m and n respectively represent the number of neurons in an input layer and an output layer in the BP neural network, and beta is a constant between 1 and 10.
The BP neural network formed by adopting the layer number and node number structure can form an optimal network structure for obtaining optimal compensation performance.
The generation of magnetic vector fields with different amplitudes and directions in the uniform region of the coil should be as rich and representative as possible, and the rate of change over time of the generation of magnetic vector fields with different amplitudes and directions in the uniform region of the coil should be as rich and representative as possible. In step S02 of this embodiment, the geomagnetic vector measurement system may specifically collect representative magnetic vector data generated by the three-dimensional helmholtz coil, and control current sequences of the three orthogonal coils to generate different directions and magnitudes in the three-dimensional spherical involute when collecting the data. The three orthogonal coil current sequences are obtained specifically according to the following spherical involute equation:
wherein R represents the radial direction of the involute, θ represents the involute expansion angle, and α represents the involute pressure angle.
By adopting the mode of controlling the three orthogonal coils to generate the current sequence, the embodiment can ensure that magnetic vector fields with different amplitudes and directions are as abundant and representative as possible, and ensure that the magnetic vector fields with different amplitudes and directions are generated in a uniform area of the coils, so that the change rate of the magnetic vector fields with different amplitudes and directions along with time is also as abundant and representative as possible, thereby effectively obtaining enough representative data for model training and further improving the performance of a compensation model. In a specific application embodiment, three-direction current constraints of the helmholtz coil are shown in fig. 3, and a partial current curve in current according to three orthogonal coils generated in a three-dimensional spherical involute is shown in fig. 4.
In step S03 of this embodiment, a genetic algorithm (genetic algorithm, GA) is specifically adopted to optimize the BP neural network. The training process needs to take a great deal of time to select proper parameters to build a network model, and is easy to fall into local optimum.
In this embodiment, step S03 further includes a model evaluation and optimization step, when the model training process is finished, evaluating whether the BP neural network model obtained by training meets a preset requirement, if yes, obtaining a final BP neural network model, otherwise, adjusting the structure and parameters of the BP neural network model until the BP neural network model meets the preset requirement, so that the BP neural network compensation model meeting the required compensation precision can be finally obtained.
In this embodiment, in the model evaluation and optimization step, a root mean square error RMSE between a geomagnetic vector true value and a geomagnetic vector value predicted by the BP neural network model is used as a performance index to evaluate compensation and generalization performance of the BP neural network model. The final goal of training the neural network is that the training model has good generalization ability for non-training samples, i.e. effectively approximates the intrinsic law of the samples. According to the embodiment, the root mean square error between the geomagnetic vector true value and the geomagnetic vector value predicted by the BP neural network model is used as a performance index, so that the generalization capability of the model can be well improved, and the compensation precision is further improved. The calculation expression of the root mean square error RMSE is specifically:
wherein H is actual Is the true value of geomagnetic vector, H predicted Geomagnetic vector values predicted by the BP neural network model are obtained, and N is the sampling number.
After the BP neural network is trained in the mode and model evaluation and optimization are carried out, a required BP neural network is finally formed, the BP neural network can be used for predicting a real geomagnetic vector value according to a measured value, the trained neural network can be regarded as a filter to filter an interference magnetic field, namely, a magnetic interference source generated by a system is eliminated through the BP neural network, and the interference magnetic field compensation of the geomagnetic vector measurement system is realized.
Errors in triaxial magnetic sensors, such as scale factors, non-orthogonality, and fixed bias, can also affect the accuracy of geomagnetic vector measurement systems. In the embodiment, the compensation module is constructed by using the BP neural network model, so that the error of the triaxial magnetic sensor can be integrated into the BP neural network model, and a nonlinear network mapping relation is established between the measured value and the real geomagnetic vector value.
According to the embodiment, the BP neural network is used as a compensation model to realize the interference magnetic field compensation of the geomagnetic vector measurement system, so that various different types of magnetic field interference and nonlinear errors can be simultaneously considered, a complete compensation model is constructed and formed, the problems of incompleteness, neglecting nonlinear errors and the like of a traditional compensation model are solved, and the elimination of various magnetic field interference can be rapidly and accurately realized; meanwhile, the three-dimensional Helmholtz coil is combined, the geomagnetic vector measurement system is arranged at the center of a uniform area of the three-dimensional Helmholtz coil, magnetic fields with different amplitudes, different directions and different change rates can be generated by the coil to create a reasonable enough data set for the BP neural network, good compensation precision can be ensured, and the method is simple to operate and low in cost, and can be used for personalized compensation according to different application scenes.
To verify the effectiveness of the above method of the present invention, a compensation test was performed using the method of the present invention in a specific application example, and an experimental apparatus is shown in fig. 2, comprising: 1) The geomagnetic field vector measurement system consists of a high-precision laser gyro inertial navigation (INS, providing attitude information) and a DM050 triaxial fluxgate magnetic sensor (measuring magnetic field component); 2) Three-dimensional helmholtz coil (creating arbitrary magnetic field components) construction; 3) Data processor, data acquisition software and data processing software. The sampling rate of the magnetic sensor is 20Hz. Wherein the triaxial magnetic sensor is calibrated before the experiment, and the output error is reduced to below 1nT after the calibration. Inertial navigation systems have also been calibrated in the laboratory using a turret with three degrees of freedom.
According to the DM050 magnetic sensor manual, the main performance indexes are as follows: magnetic field range for each sensor axis: 1mT;
quadrature error: < + -0.01 DEG; offset amount: < 5NT; resolution ratio: 2pT. According to the INS manual, the main performance specifications are as follows:
zero drift of laser gyroscope: <0.003 °; angular measurement resolution: 1e-4 °; accelerometer accuracy: 1e-6g. According to the three-dimensional helmholtz coil manual, the main performance specifications are as follows: coil size: 1 meter; uniformity: 0.1% in the central region of 260cm 3; quadrature error: less than plus or minus 0.01 degrees;
as shown in fig. 6, the detailed flow for implementing the disturbing magnetic field compensation of the geomagnetic vector measurement system in this embodiment specifically includes:
(1) and (5) experimental installation. The geomagnetic vector measurement system is placed in the center of a three-dimensional helmholtz coil, as shown in fig. 2.
(2) Data set preparation. The three-dimensional Helmholtz coil generates a magnetic field under the excitation of a predefined coil current, and the output value H of the triaxial magnetic field sensor starts to be recorded m And true value H generated by a three-dimensional Helmholtz coil 0 And H 0 And H m The values of the changes over time, and the obtained partial data are shown in table 1;
table 1: measurement data of triaxial magnetic field sensor and true value data generated by coil
(3) And (5) model training. And training the BP neural network by using the training sample data set, and improving the local searching and global searching capabilities by using a genetic algorithm so as to improve the training efficiency.
(4) Model evaluation and optimization. And when the training process of the model is finished, evaluating and optimizing the structure and parameters of the model, and if the model precision meets the requirement, judging that the model is well trained and can be used for magnetic interference compensation. Otherwise, returning to the step (3) until the requirement is met. The results of the comparison of the different network structures are shown in table 2. When the network performance is similar, the neural network with fewer hidden layers and nodes has better effect. Therefore, in this embodiment, 3×8×10×3 is specifically selected as the neural network structure.
Table 2: interference magnetic field compensation effect (nT)
| BP network structure | X component | Y | Z component | |
| 3*4*3 | 75.31 | 75.25 | 58.57 | |
| 3*10*3 | 70.45 | 74.75 | 41.05 | |
| 3*8*4*3 | 25.45 | 20.49 | 21.74 | |
| 3*8*10*3 | 16.72 | 13.58 | 11.42 | |
| 3*8*4*10*3 | 16.70 | 13.61 | 11.41 | |
| 3*8*8*10*3 | 15.24 | 12.57 | 11.43 | |
| 3*8*8*8*10*3 | 15.32 | 12.42 | 11.16 |
(5) And (5) compensating test. And predicting a real geomagnetic vector value according to the measured value by using the trained BP neural network model.
As shown in Table 3, after compensation using the BP neural network model, the root mean square error for the north, vertical, east and total intensities was reduced from 3688.9nT,3196.2nT, and 3796.2nT to 16.72nT,13.58nT,11.42nT, respectively.
Table 3: interference magnetic field compensation effect (nT)
According to the experimental result, the interference magnetic field compensation method of the geomagnetic vector measurement system based on the BP neural network can effectively eliminate the interference magnetic field around the triaxial magnetic sensor and effectively improve the accuracy and reliability of geomagnetic vector measurement. And the invention has the following advantages: firstly, a BP neural network is adopted as a compensation training model, so that the compensation training model has strong nonlinear mapping capability, linear error models such as a permanent magnetic field, an induced magnetic field and an eddy magnetic field are considered, nonlinear errors are considered, and as many interference sources as possible can be explained, so that a complete compensation model is constructed; secondly, different vector magnetic fields are generated by skillfully adopting three-dimensional Helmholtz coils to construct training and testing sample sets, a compensation equation is not required to be constructed through a rotary platform, and enough representative data can be quickly generated to establish a mapping relation of the BP neural network; thirdly, a personalized simulation scheme can be formulated without depending on the rotation of the measurement system in the geomagnetic field, and the compensation process is simple and quick.
The interference magnetic field compensation system of the geomagnetic vector measurement system based on the BP neural network of the embodiment comprises:
the test control system is used for placing the geomagnetic vector measurement system at the center of the three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling the current flowing through the coil;
the data collection module is used for collecting representative magnetic vector data generated by the three-dimensional Helmholtz coil and constructing a training data set and a testing data set for forming a model;
the model training module is used for training the BP neural network by using the training sample data set, obtaining a BP neural network model by training, and testing the BP neural network model obtained by training by using the test data set until obtaining the BP neural network model meeting the requirements;
the magnetic field compensation module is used for obtaining the measured value of the geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the obtained BP neural network model so as to realize interference magnetic field filtering.
The interference magnetic field compensation system of the geomagnetic vector measurement system based on the BP neural network in this embodiment corresponds to the interference magnetic field compensation method of the geomagnetic vector measurement system based on the BP neural network one by one, and will not be described in detail here.
The invention mainly considers the magnetic interference fields of the measurement system, the ferromagnetic component and the electrical equipment, and can be also suitable for the magnetic field interference from a small unmanned platform such as UUV or UAV when the geomagnetic field vector measurement system is carried on the small unmanned platform for mobile geomagnetic vector measurement. And as long as the uniform area of the three-dimensional Helmholtz coil is large enough, the interference source of the platform can be integrated in the BP neural network model to be considered together, and the uniform area of the three-dimensional Helmholtz coil can be made large enough in engineering according to actual needs, so that the interference source of the platform can be integrated in the BP neural network model to be considered together.
The foregoing is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. While the invention has been described with reference to preferred embodiments, it is not intended to be limiting. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present invention shall fall within the scope of the technical solution of the present invention.
Claims (10)
1. The interference magnetic field compensation method of the geomagnetic vector measurement system based on the BP neural network is characterized by comprising the following steps of:
s01, placing a geomagnetic vector measurement system at the center of a three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling current flowing through the coil;
s02, collecting representative magnetic vector data generated by a three-dimensional Helmholtz coil, and constructing a training data set and a testing data set for forming a model;
s03, training the BP neural network by using the training sample data set, training to obtain a BP neural network model, and testing the BP neural network model obtained by training by using the test data set until the BP neural network model meeting the requirements is obtained;
s04, obtaining a measured value of a geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the BP neural network model obtained in the step S03 so as to realize interference magnetic field compensation of the geomagnetic vector measurement system.
2. The method for compensating the interference magnetic field of the geomagnetic vector measurement system based on the BP neural network according to claim 1, wherein the BP neural network is optimized by adopting a genetic algorithm in the step S03.
3. The method for compensating for an interfering magnetic field of a geomagnetic vector measurement system based on a BP neural network according to claim 1, wherein an input layer of the BP neural network includes 3 neurons for inputting an output value H of the geomagnetic vector measurement system m ＝[H mx ,H my ,H mz ] T ，H mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m The output layer includes 3 neurons for outputting a true geomagnetic vector value h= [ H ] generated by the three-dimensional helmholtz coil x ,H y ,H z ] T ，H x ,H y ,H z Three components of H, respectively.
4. The method for compensating the disturbing magnetic field of the geomagnetic vector measurement system based on the BP neural network according to claim 3, wherein an hidden layer of the BP neural network is set to be 2 layers or 3 layers, and the number of neurons in the hidden layer is calculated according to the following formula:
wherein m and n respectively represent the number of neurons in an input layer and an output layer in the BP neural network, and beta is a constant between 1 and 10.
5. The method for compensating the interference magnetic field of the geomagnetic vector measurement system based on the BP neural network according to claim 1, wherein the step S03 further comprises a model evaluation and optimization step, when the model training process is finished, whether the BP neural network model obtained through training meets the preset requirement is evaluated, if yes, a final BP neural network model is obtained, and otherwise, the structure and the parameters of the BP neural network model are adjusted until the BP neural network model meets the preset requirement.
6. The method for compensating for an interference magnetic field of a BP neural network-based geomagnetic vector measurement system according to claim 5, wherein in the model evaluation and optimization step, a root mean square error RMSE between a geomagnetic vector true value and a geomagnetic vector value predicted by a BP neural network model is used as a performance index to evaluate compensation and generalization performance of the BP neural network model, and a calculation expression of the root mean square error RMSE is:
wherein H is actual Is the true value of geomagnetic vector, H predicted Geomagnetic vector values predicted by the BP neural network model are obtained, and N is the sampling number.
7. The interference magnetic field of the BP neural network-based geomagnetic vector measurement system of any one of claims 1 to 6The compensation method is characterized in that the training sample sets comprise different H m And H, a time-dependent magnetic vector field, H m The output value of the geomagnetic vector measurement system is H, which is the real geomagnetic vector value generated by the three-dimensional Helmholtz coil, H m Time-dependent changes
From the measurement data of the triaxial magnetometer, H is varied over time>
Obtained from three-dimensional Helmholtz coils and current, wherein H m Variation over time->
And H>
The method comprises the following steps of:
wherein H is mx ,H my ,H mz Output values H of geomagnetic vector measuring system respectively m Is H x ,H y ,H z Three components of the true geomagnetic vector value H generated by the three-dimensional Helmholtz coil are respectively, and delta represents the variation.
8. The method for compensating for the disturbing magnetic field of the geomagnetic vector measurement system based on the BP neural network according to any one of claims 1 to 5, wherein in the step S02, when collecting representative magnetic vector data generated by three-dimensional helmholtz coils, different directions and magnitudes are generated in three-dimensional spherical involute by controlling current sequences of three orthogonal coils.
9. The method for compensating for an interference magnetic field of a geomagnetic vector measurement system based on a BP neural network according to claim 8, wherein the three orthogonal coil current sequences are obtained specifically according to the following spherical involute equation:
wherein R represents the radial direction of the involute, θ represents the involute expansion angle, and α represents the involute pressure angle.
10. An interference magnetic field compensation system of a geomagnetic vector measurement system based on a BP neural network, which is characterized by comprising:
the test control system is used for placing the geomagnetic vector measurement system at the center of the three-dimensional Helmholtz coil, and generating magnetic vector fields with different amplitudes, different directions and different change rates in a central uniform area of the three-dimensional Helmholtz coil by controlling the current flowing through the coil;
the data collection module is used for collecting representative magnetic vector data generated by the three-dimensional Helmholtz coil and constructing a training data set and a testing data set for forming a model;
the model training module is used for training the BP neural network by using the training sample data set, obtaining a BP neural network model by training, and testing the BP neural network model obtained by training by using the test data set until obtaining the BP neural network model meeting the requirements;
the magnetic field compensation module is used for obtaining the measured value of the geomagnetic vector value, and predicting the geomagnetic vector value of the geomagnetic vector measurement system by using the obtained BP neural network model so as to realize the interference magnetic field compensation of the geomagnetic vector measurement system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211096456.3A CN116299735A (en) | 2022-09-08 | 2022-09-08 | Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211096456.3A CN116299735A (en) | 2022-09-08 | 2022-09-08 | Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116299735A true CN116299735A (en) | 2023-06-23 |
Family
ID=86832911
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211096456.3A Pending CN116299735A (en) | 2022-09-08 | 2022-09-08 | Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116299735A (en) |
-
2022
- 2022-09-08 CN CN202211096456.3A patent/CN116299735A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105891755B (en) | The bearing calibration of aircraft hanging fluxgate magnetic gradient tensor instrument | |
| JP4093861B2 (en) | Compensation of large magnetic errors for electronic compass and all orientation operations | |
| CN102252689A (en) | Electronic compass calibration method based on magnetic sensor | |
| CN109556631B (en) | INS/GNSS/polarization/geomagnetic combined navigation system alignment method based on least squares | |
| Liu et al. | Calibration and compensation of geomagnetic vector measurement system and improvement of magnetic anomaly detection | |
| Pang et al. | Calibration of three-axis magnetometers with differential evolution algorithm | |
| CN110146839A (en) | A kind of mobile platform magnetic gradient tensor system compensation method | |
| CN107894235B (en) | Model error compensation method for autonomous navigation system of ultra-high-speed aircraft | |
| CN111433634A (en) | Magnetic compensation method based on aeromagnetic compensation error model | |
| US20160356601A1 (en) | Hybrid inertial/magnetic system for determining the position and orientation of a mobile body | |
| CN113156355B (en) | Magnetic interference compensation method of superconducting full tensor magnetic gradient measuring device | |
| CN114674301B (en) | Electronic compass active compensation method and system under strong magnetic interference environment | |
| Liu et al. | Distortion magnetic field compensation of geomagnetic vector measurement system using a 3-D Helmholtz coil | |
| CN111220932B (en) | Unmanned aerial vehicle magnetic interference calibration method and distributed magnetic anomaly detection system | |
| CN109725361A (en) | A kind of locating magnetic objects method based on magnetic gradient tensor invariant | |
| Gnadt | Machine learning-enhanced magnetic calibration for airborne magnetic anomaly navigation | |
| Liu et al. | Compensation of geomagnetic vector measurement system with differential magnetic field method | |
| CN115542225A (en) | Correction method for improving magnetic measurement precision of suspension wire torsion balance device | |
| Jiang et al. | Scalar calibration of aeromagnetic data using BPANN and LS algorithms based on fixed-wing UAV platform | |
| Liu et al. | Simplified ellipsoid fitting-based magnetometer calibration for pedestrian dead reckoning | |
| Wan et al. | Improved component compensation for geomagnetic field vector measurement using Lagrange multiplier method | |
| Chen et al. | A novel calibration method for tri-axial magnetometers based on an expanded error model and a two-step total least square algorithm | |
| CN115524762A (en) | Geomagnetic vector measurement system compensation method based on three-dimensional Helmholtz coil | |
| CN110954081A (en) | Quick calibration device and method for magnetic compass | |
| CN116299735A (en) | Interference magnetic field compensation method of geomagnetic vector measurement system based on BP neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |