CN115130379A - Magnet modeling and magnetic field estimation method and device - Google Patents
Magnet modeling and magnetic field estimation method and device Download PDFInfo
- Publication number
- CN115130379A CN115130379A CN202210738001.0A CN202210738001A CN115130379A CN 115130379 A CN115130379 A CN 115130379A CN 202210738001 A CN202210738001 A CN 202210738001A CN 115130379 A CN115130379 A CN 115130379A
- Authority
- CN
- China
- Prior art keywords
- magnetic
- magnetic field
- virtual point
- training
- virtual
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000013528 artificial neural network Methods 0.000 claims abstract description 142
- 238000012549 training Methods 0.000 claims abstract description 122
- 238000009826 distribution Methods 0.000 claims description 16
- 230000010365 information processing Effects 0.000 claims description 2
- 238000005259 measurement Methods 0.000 description 38
- 238000010586 diagram Methods 0.000 description 16
- 238000004364 calculation method Methods 0.000 description 7
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 238000000691 measurement method Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000005389 magnetism Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
Magnet modeling and magnetic field estimation methods and apparatus are provided. A target magnet modeling method is provided, comprising: establishing one or more virtual point magnetic sources representing a target magnet; arranging a plurality of training measuring points in the space around the target magnet; measuring the magnetic field intensity of each training measuring point under the influence of a target magnet; generating training samples according to the magnetic field intensity of each training measuring point and the positions of the training measuring points to train one or more neural networks; the one or more virtual point magnetic sources are respectively provided with one or more corresponding neural networks, the neural networks corresponding to the virtual point magnetic sources are trained, and the output of each neural network represents the magnetic source parameters of the corresponding virtual point magnetic source; and using the position and the magnetic source parameter of each of the one or more virtual point magnetic sources as a magnetic model of the target magnet.
Description
Technical Field
The present invention relates to magnetism and neural network technology, and in particular, to a method and apparatus for modeling a magnet using a neural network and estimating the magnetic field strength at any point using the constructed magnet model.
Background
There is a need to measure and predict magnetic field strength. When the magnet model of the target magnet is not known, the magnetic field strength at a specified position around the target magnet can be obtained by measurement. The strength of the magnetic field at its location can be obtained by, for example, a three-axis magnetometer. However, for locations in space that are difficult to access or reach, it is not possible to measure with a magnetometer. For example, a ship is in motion, the hull of which is magnetized by the earth's magnetic field and forms a magnetic field in space. The magnetic field intensity formed by the ship at the relevant position in water is difficult to measure by a magnetometer.
The traditional ship magnetic field measurement method mainly comprises a seabed large-area array method, a seabed travelling measurement method and an artificial measurement method. Although the former two measurement accuracy is high, the construction cost and the maintenance cost are extremely high, and only the field intensity on a fixed plane can be measured, and the magnetic field distribution condition in the whole space cannot be measured. The manual measurement method has low efficiency, and the position and the posture of a measurement point are difficult to be accurately controlled, so that the accuracy of only a vertical component of a measured magnetic field is high, and the overall distribution condition of the magnetic field is difficult to obtain.
There are methods of calculating (or estimating) the magnetic field strength at a given location in space from a magnet model. For example, a formula (inverse cubic formula, dipole formula, etc.) for calculating the magnetic field strength from the magnetic moment is proposed in "calculating the magnetic field strength from the magnetic moment" (shin Xin bang et al, digital ocean and underwater attack and defense, 6.2019, vol 2, 2 nd paragraph).
Fig. 1A shows a prior art schematic for measuring the strength of an underwater magnetic field around a ship's hull.
The magnetic field intensity measuring equipment is arranged underwater at the berthing position of a ship body in facilities such as a wharf, a dock and the like, and the magnetic field intensity is measured by the measuring equipment when the ship is berthed. The drawback of this practice is evident, on the one hand, by the dependence on large installations able to accommodate the mooring of the ship, and by the need to arrange measuring equipment under water, which requires a considerable cost; on the other hand, the need for the ship to drive into the facility and berth to await completion of the measurement also introduces a significant time overhead and impacts on other tasks of the ship. And the measured magnetic field intensity is only the magnetic field intensity of the position of the measuring equipment, and the magnetic field intensity of other positions of the water area near the ship body still cannot be obtained.
Fig. 1B shows a schematic diagram of estimating magnetic field strength using a neural network.
A neural network is used to predict the magnetic field strength at a given location. In training of the neural network, the position P of a measuring device, such as a magnetometer, is provided as an input to the neural network, which outputs an estimated magnetic field strength at the position P. And (3) obtaining an error of the actual magnetic field intensity measured by the magnetometer at the position P according to the magnetic field intensity of the position P estimated by the neural network and the actual magnetic field intensity, and feeding back the error to the neural network. And training the neural network through the position data of a plurality of positions P and the real magnetic field intensity data of a plurality of measurements to obtain the neural network for estimating the magnetic field intensity of any position.
For a trained neural network, it is adapted to estimate the magnetic field strength at any location of a designated target magnet. By inputting location information thereto, the neural network outputs a predicted value of the magnetic field strength at that location.
However, such network structures are highly susceptible to overfitting. In the estimation of the magnetic field strength, when the distribution of positions to be estimated (also called measuring points) is consistent with the distribution of positions at which magnetic field strength data are acquired during training, the prediction result of the neural network is more accurate. However, if the measured points to be estimated are not consistent with the measured point distribution of the magnetic field intensity data acquired during training, for example, when the measured points are on different sides of the target magnet, the result predicted by the neural network is likely to have great errors, and even the direction of the result is opposite to the actual value.
Disclosure of Invention
It is desirable to obtain a magnet model of a target magnet of any shape to be measured, and to obtain a magnetic field distribution in the magnet surrounding space formed by the target magnet based on the magnet model, and then to obtain a magnetic field intensity at any position in the magnet surrounding space, thereby eliminating the burden of measuring the magnetic field in the field. For this purpose, there is also a need for a method of modeling a magnet model of an arbitrarily shaped target magnet, which can accurately obtain a magnet model of the target magnet.
According to the embodiment of the application, an arbitrary (complex-shaped) target magnet is modeled by magnetic field intensity data collected at partial observation points, and the magnetic field intensity of an arbitrary point is calculated according to the model. Furthermore, the magnetic model of the ship is modeled by using the field intensity data obtained by onshore measurement, so that the field intensity distribution of the water area around the ship is estimated, and the indirect measurement of the magnetic field intensity is realized.
According to a first aspect of the present application, there is provided a first target magnet modeling method according to the first aspect of the present application, comprising: establishing one or more virtual point magnetic sources representative of the target magnet, the point virtual point magnetic sources having a specified location within the space occupied by the target magnet; arranging a plurality of training measuring points in the space around the target magnet, wherein the training measuring points have designated positions; measuring the magnetic field intensity of each training measuring point under the influence of a target magnet; generating training samples according to the magnetic field intensity of each training measuring point and the positions of the training measuring points to train one or more neural networks; the one or more virtual point magnetic sources are respectively provided with one or more corresponding neural networks, the neural networks corresponding to the virtual point magnetic sources are trained, and the output of each neural network represents the magnetic source parameters of the corresponding virtual point magnetic source; according to the magnetic source parameters and positions of the one or more virtual point magnetic sources output by the neural network and the positions of the training measuring points in the training sample, calculating the magnetic field intensity generated by the one or more virtual point magnetic sources at the positions of the training measuring points in the training sample as estimated magnetic field intensity; taking the difference between the estimated magnetic field strength and the magnetic field strength of the training sample as an error for training a neural network corresponding to each of the one or more virtual point magnetic sources; using the output of the trained neural network as the magnetic source parameter of the virtual point magnetic source corresponding to the output of the trained neural network; and using the position and the magnetic source parameter of each of the one or more virtual point magnetic sources as a magnetic model of the target magnet.
According to a first target magnet modeling method of the first aspect of the present application, there is provided a second target magnet modeling method of the first aspect of the present application, wherein the one or more virtual point magnet sources are arranged at equal intervals within a space occupied by the target magnet.
According to the first or second target magnet modeling method of the first aspect of the present application, there is provided a third target magnet modeling method according to the first aspect of the present application, wherein the target magnet is represented by a single virtual point magnet source; or setting the one or more virtual point magnetic sources so that 6 times of the distance between any virtual point magnetic source is smaller than the distance between any training measuring point and any virtual point magnetic source.
According to one of the first to third target magnet modeling methods of the first aspect of the present application, there is provided the fourth target magnet modeling method according to the first aspect of the present application, further comprising: to train the one or more neural networks, virtual inputs are provided to each of the one or more neural networks, the virtual inputs provided to the one or more neural networks obeying a specified distribution.
According to a fourth target magnet modeling method of the first aspect of the present application, there is provided the fifth target magnet modeling method of the first aspect of the present application, wherein the one or more neural networks are trained such that an estimated magnetic field strength at each training measurement point calculated from a magnetic source parameter of a magnetic source at each virtual point obtained from an output of the one or more neural networks is as consistent as possible with a magnetic field strength of a corresponding training sample.
According to a fifth object magnet modeling method of the first aspect of the present application, there is provided the sixth object magnet modeling method of the first aspect of the present application, wherein the one or more neural networks are trained, and further, magnetic source parameters of the magnetic sources of the virtual points obtained from outputs of the one or more neural networks are made to be as consistent as possible under the condition of different virtual inputs.
According to a sixth target magnet modeling method of the first aspect of the present application, there is provided the seventh target magnet modeling method of the first aspect of the present application, wherein the one or more virtual point magnetic sources each have a corresponding plurality of neural networks; using the statistic value of the output of a plurality of neural networks corresponding to the one or more virtual point magnetic sources respectively as the magnetic source parameter of the virtual point magnetic source represented by the one or more virtual point magnetic sources; and taking the difference between the outputs of the plurality of neural networks respectively corresponding to the one or more virtual point magnetic sources as a further error of training the neural networks.
According to a seventh target magnet modeling method of the first aspect of the present application, there is provided the eighth target magnet modeling method of the first aspect of the present application, wherein in the training of the neural network, a superposition of magnetic field strengths generated by the magnetic source of all virtual points at the positions of the training points of the training sample respectively is calculated as the estimated magnetic field strength according to a magnetic field strength formula, based on the magnetic source parameters and the positions of all virtual point magnetic sources obtained from the neural network, and the positions of the training points in the training sample.
According to an eighth target magnet modeling method of the first aspect of the present application, there is provided the ninth target magnet modeling method of the first aspect of the present application, wherein a plurality of virtual inputs are provided, the distributions to which the plurality of virtual inputs respectively follow are identical to each other, and the number of the plurality of virtual inputs is identical to the number of the plurality of neural networks to which the one or more virtual point magnet sources respectively correspond; providing one of the plurality of virtual inputs to each of a plurality of neural networks to which the one or more virtual point magnetic sources respectively correspond; each of the plurality of virtual inputs is simultaneously provided to a respective one of a plurality of neural networks for each virtual point magnetic source.
According to a ninth target magnet modeling method of the first aspect of the present application, there is provided the tenth target magnet modeling method of the first aspect of the present application, wherein the background magnetic field strength of each training station is measured in the absence of the target magnet; measuring the real magnetic field intensity of each training measuring point under the condition that the target magnet exists; and obtaining the magnetic field intensity of each training measuring point under the influence of the target magnet according to the difference between the real magnetic field intensity and the background magnetic field intensity.
According to a second aspect of the application, a magnetic field strength estimation method of a target measuring point under the influence of a target magnet is provided, and comprises the following steps: establishing a magnetic model of a target magnet represented by one or more virtual point magnetic sources according to one of the target magnet modeling methods of the first aspect of the present application; and calculating the superposition of the magnetic field intensity generated by all the virtual point magnetic sources of the one or more virtual point magnetic sources at the position of the target measuring point according to the magnetic field intensity formula and the magnetic source parameters and the positions of the one or more virtual point magnetic sources to obtain the magnetic field intensity estimation of the target measuring point under the influence of the target magnet.
According to a third aspect of the present application, there is provided an information processing apparatus comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor implements one of the methods according to the first and second aspects of the present application when executing the program.
Drawings
The application, as well as a preferred mode of use, further objectives and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying drawings, wherein:
fig. 1A shows a prior art schematic for measuring the strength of an underwater magnetic field around a ship's hull.
Fig. 1B shows a schematic diagram of estimating magnetic field strength using a neural network.
Fig. 2 shows a schematic diagram of an array of point sources for a target magnet (bar magnet).
Fig. 3 shows a schematic diagram of magnet modeling and magnetic field strength prediction using a neural network according to an embodiment of the present application.
Fig. 4 illustrates a flow chart for modeling a target magnet and estimating magnetic field strength according to the present application.
FIG. 5 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to an embodiment of the application.
Fig. 6 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to yet another embodiment of the present application.
Fig. 7 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to yet another embodiment of the present application.
Fig. 8 illustrates a schematic diagram of magnet modeling and magnetic field strength prediction using a neural network according to yet another embodiment of the present application.
Fig. 9 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to still another embodiment of the present application.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application.
According to an embodiment of the application, a target magnet is fitted using an array of point magnetic sources formed from a plurality of point magnetic sources (referred to as virtual point magnetic sources). Magnetic source parameters of each virtual point magnetic source are predicted (instead of predicting the magnetic field intensity of the designated position) through machine learning by utilizing a neural network, the magnetic field intensity of the designated position under the influence of each virtual point magnetic source is calculated according to the magnetic source parameters of each virtual point magnetic source, and the measured value of the magnetic field intensity generated (calculated) by superposing each virtual point magnetic source at any position is close to the measured value of the corresponding position as much as possible.
The target magnet has a designated shape, for example, a ship is generally shuttle-shaped. And arranging the plurality of virtual point magnetic sources at the specified positions, so that the point magnetic source array formed by the plurality of virtual point magnetic sources is close to the target magnet in shape. It can be understood that by arranging more virtual point magnetic sources, the shape of the point magnetic source array is closer to the shape of the target magnet, which is beneficial to improving the accuracy of magnet modeling and magnetic field intensity prediction, but the complexity and the calculation amount are increased correspondingly. According to embodiments of the present application, accurate prediction of magnetic field strength has been achieved with a small number of virtual point magnetic sources arranged in a shape substantially similar to the target magnet, e.g., modeling of a ship magnet can be achieved with a magnetic source array formed of 3 virtual point magnetic sources arranged in a line. For another example, for a vehicle target magnet with a substantially rectangular shape, for example, 8 points are selected in the vehicle interior space to set virtual point magnetic sources, the rectangular hexahedron of the 8 points in the city enclosure has a shape similar to that of the vehicle, and the vehicle target magnet is fitted with a point magnetic source array formed by the 8 virtual point magnetic sources. For another example, when the target magnet is a relatively long distance from the survey point (e.g., the survey point is located more than 6 times the size of the target magnet relative to the target magnet), the target magnet can be better fit by a single virtual point magnet source.
In some embodiments of the present application, a virtual point magnetic source is modeled as a point having a spatial location and magnetic source parameters. In yet other examples, the virtual point magnetic source may also have a volume or shape. Whether the virtual point magnetic source represents a point or has a specified volume or shape, the magnetic field intensity formed by the point magnetic source at a specified position in space can be calculated according to the magnetic source parameters.
As an example, to create an array of point sources for fitting to a target magnet, the spatial locations of the virtual point sources are set within the contour of the target magnet and distributed as uniformly (equally spaced) as possible. Generally, the more complex the shape of the target magnet, the closer the region of magnetic field strength to be estimated is to the target magnet, and the greater the number of virtual point sources needed to achieve the same accuracy. The number and spacing of the point sources is related to the size of the target magnet to be modeled and the distance of the measured points from the target magnet, as well as the accuracy of the magnetic field strength estimation that needs to be achieved. Generally speaking, the more the number of point magnetic sources used for fitting, the higher the fitting accuracy, but the more training measured point data need to be collected when the neural network is trained.
When the distance of the target measurement point relative to the target magnet exceeds 6 times the size of the target magnet, the target magnet can be approximately regarded as one point (within 1% error, see "calculate magnetic field strength from magnetic moment"). At the moment, only a single virtual point magnetic source is needed to be arranged at the central position of the target magnet, and the target magnet can be well fitted by modeling the target magnet through the virtual point magnetic source.
The position of the target measurement point depends on the need to estimate the magnetic field strength. For example, to estimate the magnetic field strength around a ship, the distance from the target measurement point to the ship does not usually satisfy the condition that the distance of the measurement point from the target magnet exceeds 6 times the size of the target magnet due to the large size of the ship itself. In such a case, the point magnetic source array can also be made to fit the target magnet well by arranging a plurality of virtual point magnetic sources at equal intervals within the contour range of the target magnet, so that each virtual point magnetic source represents a part of the magnet to be measured, and so that the distance from the target measuring point to the nearest virtual point magnetic source exceeds, for example, 6 times the distance between adjacent virtual point magnetic sources.
Fig. 2 shows a schematic diagram of an array of point sources for a target magnet (bar magnet).
By way of example, the target magnet is a magnetic rod of 1 meter length and 1cm diameter. The target measuring point P is 0.6 meter away from the magnetic rod on the central side of the magnetic rod. Then 11 virtual point magnetic sources M may need to be arranged at equal intervals along the length direction of the magnetic rod, so that the distance between adjacent virtual point magnetic sources is within 0.1 meter, and the requirement that the distance from the target measuring point to the nearest virtual point magnetic source exceeds 6 times of the distance between the virtual point magnetic sources is met.
It can be understood that in some cases (for example, the required accuracy requirement is not high, and/or the spatial position of the target magnet and the measuring point is limited), the point magnetic source array is arranged so that the distance from the target measuring point to the nearest virtual point magnetic source is less than 6 times of the distance between adjacent virtual point magnetic sources, and although the accuracy of magnetic field intensity estimation is affected, the purpose of modeling the target magnet can also be achieved.
Fig. 3 shows a schematic diagram of magnet modeling and magnetic field strength prediction using a neural network according to an embodiment of the present application.
In the example of fig. 3, the target magnet is fitted with 2 virtual point magnetic sources (denoted as a and B, respectively, not shown). The virtual point magnetic source has a known spatial location.
According to an embodiment of the application, for each virtual point magnetic source used to fit the target magnet, a multi-layer neural network (310, 312) is provided, for example. The multilayer neural network corresponds to the virtual point magnetic sources, for example, one-to-one. The neural networks (310, 312) are trained to output magnetic source parameters for the virtual point magnetic sources they represent. The magnetic source parameters are used for describing that the virtual point magnetic source is a parameter which can represent the magnetic characteristics of the virtual point magnetic source. For example, to calculate the magnetic field strength using the inverse cubic formula of "calculate the magnetic field strength from the magnetic moment", the corresponding magnetic source parameter is the magnetic moment. If the magnetic field strength is calculated using a magnetic dipole model, the corresponding magnetic source parameters include the magnetic pole strength and the inter-pole distance. If the magnetic field strength is calculated using a current loop model, the corresponding magnetic source parameters include current loop area and current.
The operation of the neural network requires input. According to an embodiment of the application, the inputs provided to each neural network (10, 312) are, for example, specified constants. Optionally, virtual inputs are also provided to each neural network (310, 312). The virtual input is a random variable that obeys some distribution (e.g., a uniform distribution that obeys the interval 0, 1).
The magnetic field intensity calculating units (320, 322) calculate the magnetic field intensity generated by the virtual point magnetic source at the measuring point according to the magnetic source parameters and the position of the virtual point magnetic source provided by the neural networks (310, 312) and the position of the measuring point. In the example of fig. 3, the magnetic field strength calculation unit 320 calculates the magnetic field strength generated by the virtual point magnetic source a at each measuring point according to the magnetic source parameters and the position of the virtual point magnetic source a; the magnetic field intensity calculation unit 322 calculates the magnetic field intensity generated by the virtual point magnetic source B at each measurement point according to the magnetic source parameters and the position of the virtual point magnetic source B. There are also a plurality of measurement points (denoted as P). The number of stations P with different positions is large to provide enough training samples to train the neural network (310, 312).
The method of calculating the magnetic field intensity generated by the virtual point magnetic source at the designated position (the measured point) according to the magnetic source parameters and the position of the virtual point magnetic source belongs to the prior art. For example, the inverse cubic formula and the dipole formula are provided in "calculating magnetic field strength from magnetic moment".
The magnetic field strength at the survey point is, for example, the superposition (e.g., vector sum) of the magnetic field strengths generated at the survey point by all virtual point magnetic sources. The output of the magnetic field strength calculation unit (320, 322) is provided to a summing unit. The summing unit sums the outputs of the magnetic field strength calculation units (320, 322) to obtain an estimated value of the magnetic field strength at each measurement point P, which is collectively generated by, for example, 2 virtual point magnetic sources.
Also in the presence of the target magnet, at each station, a measurement value of the magnetic field strength generated by the target magnet at each station P is measured with, for example, a magnetometer.
The estimated value and the measured value of the magnetic field strength of each measured point P are compared, and the difference is fed back as an error to each neural network (310, 312) to supervise the learning of the neural network (310, 312). The learning goal (goal one) of the neural network (310, 312) is to make the estimated value of the magnetic field strength output by the summation unit and the measured value of the magnetic field strength of the corresponding measuring point as close as possible through the magnetic source parameters of each virtual magnetic source point output by the neural network.
In order to obtain a training sample of the neural network, under the condition that a target magnet exists, the measured value of the magnetic field intensity generated by the target magnet at each measuring point P is measured at each measuring point, and the training sample is marked as < measuring point coordinates and measuring point magnetic field intensity measured value >. According to each training sample (from the measuring point Pi), the neural networks (310 and 312) respectively generate output (the training samples are not directly supplied to the neural networks (310 and 312), the input of the neural networks (310 and 312) is from virtual input), the estimated value of the magnetic field strength of the measuring point Pi is calculated through the magnetic field strength calculating units (320 and 322) and the summing unit, and error feedback is generated according to the measured value of the magnetic field strength of the measuring point provided by the training samples and is fed back to the neural networks (310 and 312). Wherein the station position and the virtual point magnetic source position are each, for example, three-dimensional coordinates.
Further, the further learning goal (goal two) of the neural network (310, 312) is to make the magnetic source parameters of the virtual point magnetic source output by it constant no matter what value its virtual input takes.
And converging to meet a first training target and a second optional target through the training neural network, wherein the magnetic source parameters output by the neural networks (310 and 312) are the magnetic source parameters of the virtual point magnetic source A and the magnetic source parameters of the virtual point magnetic source B respectively, so that the virtual point magnetic source A and the virtual point magnetic source B are obtained. And the obtained virtual point magnetic source A and virtual point magnetic source B can effectively fit the target magnet. The virtual point magnetic source A and the virtual point magnetic source B can effectively fit the target magnet, which means that the magnetic field intensity generated by the target magnet at a specified point in space around the target magnet is consistent with the sum of the magnetic field intensities generated by the virtual point magnetic source A and the virtual point magnetic source B at the specified point.
After the magnetic source parameters of the virtual point magnetic source a and the virtual point magnetic source B are obtained, the target magnet model represented by the virtual point magnetic source a and the virtual point magnetic source B are also obtained, where the virtual point magnetic source a and the virtual point magnetic source B have known positions (in the space where the target magnet is located). The magnetic field strength (generated by the target magnet) at any point in space around the target magnet can also be estimated from the model of the target magnet represented by virtual point magnetic source a and virtual point magnetic source B. For example, for a specified point Pj (having a specified position) around the target magnet, the magnetic field strengths generated at the point Pj by the virtual point magnetic source a and the virtual point magnetic source B, respectively, are calculated from the positions of the virtual point magnetic source a and the virtual point magnetic source B and the magnetic source parameters, and the sum is an estimated value of the magnetic field strength generated at the point Pj by the target magnet.
Fig. 4 illustrates a flow chart for modeling a target magnet and estimating magnetic field strength according to the present application.
To model the target magnet, one or more virtual point magnetic sources representing the target magnet are established (410). For example, the number of virtual point magnetic sources used for modeling the target magnet is determined from the positions (distances) of the survey points with respect to the target magnet. When the distance between the measuring point and the target magnet is far, a small number of virtual point magnetic sources are used for modeling the target magnet; and when the measuring points are close to the target magnet, modeling the target magnet by using a large number of virtual point magnetic sources. By way of example, the position (coordinates) of each virtual point magnetic source is set by arranging the virtual point magnetic sources at uniform or equal intervals in the space in which the target magnet is located. It will be appreciated that the virtual point magnetic source is not the actual magnetic source present, but rather the object for modeling the magnetic field strength calculation set by the target magnet, with the specified location and unknown magnetic source parameters.
Stations (420) for training the neural network are positioned relative to the target magnet. The measured points are location points in the space around the target magnet. The number of stations is usually large, so that a training sample for training the neural network is obtained by using each station. The stations used to collect the training samples are referred to as training stations. The measuring points needing the estimated magnetic field intensity are called target measuring points.
According to embodiments of the application, the positions of the training stations need not be close to the target stations. Optionally, the training stations are distributed, for example, on different curved surfaces and different curves, in order to describe as much as possible the magnetic field distribution around the target magnet by the position distribution of the training stations. When the magnetometer is used to measure the magnetic field strength, optionally, in the field around the target magnet, the variation of the magnetic field strength in the space is obtained according to the magnetometer, and the interval/position of the measuring points is adjusted accordingly, so that a greater number of training measuring points (or the interval of the set training measuring points is reduced) is provided in the region (or direction) where the magnetic field strength variation is greater (faster), and a lesser number of training measuring points (or the interval of the set training measuring points is increased) is provided in the region (or direction) where the magnetic field strength variation is smaller (slower). By way of example, the number of training stations is no less than 3 times the number of virtual point sources used to model the target magnet, and preferably no less than 6 times the number of virtual point sources used to fit the target magnet.
In the presence of a target magnet, magnetic field strength and spatial coordinates are collected at each training station as training samples (430). Optionally, when the training sample is collected, the influence of the earth magnetic field or the environment magnetic field on the magnetic field strength of the training measuring point is also eliminated. For example, first, in the absence of a target magnet, a magnetometer is used at each training station to measure the magnetic field strength, which is referred to as the background field strength, and data is recorded. Next, the target magnet is made to appear at a specified location (which is the same location at which the target magnet modeled by the virtual point magnet source in step 410 is located), and the magnetic field strength (also referred to as apparent field strength) is measured at each of the training stations. When the training samples are collected, the coordinates of each training measuring point are recorded, and the 'field intensity value caused by a target magnet' obtained by deducting the background field intensity value from the measuring value of the magnetometer when the training samples exist is recorded in a manner of correlating with the coordinates and is used as the labeled magnetic field intensity of the training samples. Optionally, the external environment (e.g., weather conditions) at which the background field strength is measured is made to coincide with the external environment at which the apparent field strength of the training stations is measured.
Each neural network is trained (440) using the collected training samples. Referring also to FIG. 3, one or more neural networks (310, 312) are provided, each neural network corresponding to one of the virtual point magnetic sources created in step 410. The input of the neural network is a fixed value or a virtual input conforming to a specified distribution, and the output of the neural network is a magnetic source parameter of a magnetic source of a corresponding virtual point. And calculating the magnetic field intensity of all the virtual point magnetic sources generated at each training measuring point according to the point magnetic source parameters output by each neural network and the positions of the corresponding virtual point magnetic sources. And training the neural network to ensure that the magnetic field intensity generated by all virtual point magnetic sources at the calculated position at each training measuring point is consistent with the magnetic field intensity labeled in the training sample representing the training measuring point as much as possible. For example, the difference between the calculated magnetic field strength and the magnetic field strength in the training sample is made smaller than a specified threshold value.
And for each trained neural network, outputting the magnetic source parameters of the corresponding virtual point magnetic source. Therefore, the virtual point magnetic sources with the magnetic source parameters are used for obtaining a magnetic model of the target magnet, and modeling of the target magnet is completed. By utilizing the magnetic model of the target magnet, the magnetic field intensity of any position around the target magnet can be obtained according to the existing magnetic field intensity calculation method.
When the magnetic field intensity of the target measuring points appointed around the target magnet needs to be estimated, the positions (coordinates) of the target measuring points are obtained (450), and the sum of the magnetic field intensities generated by all virtual point magnetic sources representing the target magnet at the target measuring points is calculated to be used as the estimation of the magnetic field intensity of the target measuring points (460). Wherein, the magnetic source parameters of the magnetic sources of the virtual points are obtained from the outputs of the neural networks trained in step 440.
FIG. 5 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to an embodiment of the application.
In fig. 5, a rectangle surrounded by a dotted line represents a target magnet, a single virtual point magnet source (a rectangle surrounded by a dotted line) is provided inside a spatial position of the target magnet, and a surrounding space outside the target magnet includes a plurality of measurement points (a rectangle surrounded by a solid line). In the example of fig. 5, a single virtual point magnetic source has been able to effectively fit the magnetic model of the target magnet, since the survey point is farther from the target magnet.
In FIG. 5, the measurement points to the right and below the target magnet are training measurement points (marked by P1, P2 … … Pm … … Pn) and the measurement points to the left of the target magnet are target measurement points (marked by Pq). Thus, the training stations need not be adjacent to the target stations, thereby facilitating the collection of training samples (as the location of the target stations may be inconvenient or inaccessible or unavailable for placement of magnetometers).
In order to obtain a training sample, when a target magnet exists, the position coordinates of the measuring points and the magnetic field intensity of the measuring points are collected at the positions of the training measuring points. Optionally, at each measurement point, a background field strength (magnetic field strength at the measurement point when the target magnet is not present) and an apparent field strength (magnetic field strength at the measurement point when the target magnet is present) are collected, and the difference between the apparent field strength and the local field strength is used as the magnetic field strength of the measurement point in the training sample.
And training the neural network by using the training sample to obtain the magnetic source parameters of the virtual point magnetic source. And calculating the magnetic field intensity of the target measuring point (Pq) by using the position of the virtual point magnetic source and the magnetic source parameters as the estimation of the magnetic field intensity generated by the target magnet at the target measuring point.
Fig. 6 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to yet another embodiment of the present application.
In fig. 6, a rectangle surrounded by a chain line represents a rod-shaped target magnet (magnetic rod 600), a plurality of virtual point magnetic sources (rectangles surrounded by broken lines) (denoted as a1 and a2 … … An) are provided inside a space position where the magnetic rod 600 is located, and a plurality of measurement points (rectangles surrounded by solid lines) (denoted as P1 and P2 … … Pm) are provided in a surrounding space outside the target magnet. In the example of FIG. 6, because the survey points are closer to the magnetic bar 600, a plurality of virtual point magnetic sources are used to fit a magnetic model of the magnetic bar 600. Optionally, the virtual point magnetic sources are arranged such that 6 times the spacing between each virtual point magnetic source is still less than the distance between any measuring point to any virtual point magnetic source.
In FIG. 6, the stations include training stations (labeled P1, P2 … … Pm) and target stations (labeled Pq). In order to obtain training samples, when the magnetic bar 600 exists, the position coordinates of the measuring points and the magnetic field intensity of the measuring points are collected at the positions of the training measuring points. Optionally, at each measurement point, a background field strength (magnetic field strength at the measurement point when the magnetic bar 600 is not present) and an apparent field strength (magnetic field strength at the measurement point when the magnetic bar 600 is present) are collected, and the difference between the apparent field strength and the local field strength is used as the magnetic field strength of the measurement point in the training sample.
And training the neural network by using the training sample to obtain the magnetic source parameters of the virtual point magnetic source. And calculating the magnetic field intensity of the target measuring point (Pq) by using the position of the virtual point magnetic source and the magnetic source parameter as the estimation of the magnetic field intensity generated by the magnetic bar 600 at the target measuring point (Pq).
Fig. 7 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to yet another embodiment of the present application.
In fig. 7, an ellipse surrounded by a dotted line represents a target magnet (ship). To facilitate the acquisition of training samples, the vessel is docked in a port or dock. The harbor or the dock is in a groove shape, and the two sides of the ship are provided with lands which are convenient for arranging training measuring points (marked as P1 and P2 … … Pm). Target points (Pg) are in the water, for example, below the ship, where it is not convenient to arrange magnetic field strength measuring devices. A plurality of virtual point magnetic sources (rectangles enclosed by dashed lines) (noted A, B … … C) are placed inside the spatial positions occupied by the hull of the vessel to fit the target magnet (vessel).
In the example of fig. 7, since the positions of the survey points are limited by the field conditions of the port or dock, the arrangement and number of the virtual point magnetic sources are selected according to the relationship between the distance from the survey points to the ship and the size of the ship itself. For example, a magnetic model of the ship is fitted by three virtual point magnetic sources (A, B and C).
And acquiring training samples at each training measuring point, and training three neural networks to obtain magnetic source parameters of the virtual point magnetic source (A, B and C). And then the magnetic field intensity of the virtual point magnetic source A, B and C at the target measuring point Pg is calculated as the estimation of the magnetic field intensity generated by the ship at the position of the measuring point Pg.
It will be appreciated that after the magnetic source parameters of the virtual point magnetic source A, B and C are obtained, the magnetic field strength at any point around the vessel can be estimated. Also, the vessel is not limited to being located in the port or dock where the training samples are obtained, but may be at any location and from the relative positions of the target survey point and the vessel, the magnetic field strength of the target survey point is estimated from the magnetic source parameters of virtual point magnetic sources A, B and C.
Fig. 8 illustrates a schematic diagram of magnet modeling and magnetic field strength prediction using a neural network according to yet another embodiment of the present application.
In the example of fig. 8, the target magnet is fitted with 2 virtual point magnetic sources (denoted as a and B, respectively, not shown). The virtual point magnetic source has a known spatial location.
In order to stabilize the magnetic source parameters of the virtual point magnetic sources estimated by the neural network, a plurality of neural networks are provided for each virtual point magnetic source. Referring to fig. 8, neural networks 810 and 812 are used to estimate the magnetic source parameters of the virtual point magnetic source a, and neural networks 820 and 822 are used to estimate the magnetic source parameters of the virtual point magnetic source B. The magnetic sources of the virtual point magnetic sources, for example, are estimated using the respective outputs of a plurality of neural networks (e.g., neural networks 810 and 812) corresponding to the virtual point magnetic sources. In the example of fig. 8, the outputs of each of the neural networks 810 and 812 are averaged (840) as an estimate of the magnetic source parameter of the virtual point magnetic source a. The outputs of each of the neural networks 820 and 822 are averaged (844) as an estimate of the magnetic source parameters of the virtual point magnetic source B. The consistency or difference of the outputs of each of the plurality of neural networks corresponding to the same virtual point magnetic source is also evaluated as an additional error for training the neural networks. In the example of fig. 8, the mean square error of the outputs of neural networks 810 and 812, respectively, is calculated (842) to evaluate the coherence of their inputs, and the mean square error of the outputs of neural networks 820 and 822, respectively, is calculated (846) to evaluate the coherence of their inputs.
A magnetic field intensity calculating unit 850 for calculating the magnetic field intensity generated by the virtual point magnetic source A at each measuring point according to the estimated magnetic source parameters and position of the virtual point magnetic source A; the magnetic field intensity calculating unit 852 calculates the magnetic field intensity generated by the virtual point magnetic source B at each measuring point according to the magnetic source parameters and the position of the virtual point magnetic source B. The magnetic field intensity calculated by the magnetic field intensity calculating unit 850 and the magnetic field intensity calculated by the magnetic field intensity calculating unit 852 at each measuring point are summed (860), and an estimated value of the magnetic field intensity generated by the virtual point magnetic source A and the virtual point magnetic source B at each measuring point is obtained. There are a plurality of measurement points (denoted as P). The number of stations P with different positions is large to provide enough training samples to train the neural network (810, 812, 820, and 822).
At each station, a measurement of the magnetic field strength generated by the target magnet at each station P is measured. The estimated value of the magnetic field strength at each point P is compared with the measured value, and the difference is fed back as an error to each neural network (810, 812, 820 and 822) to supervise the learning of the neural network (810, 812, 820 and 822). In the example of fig. 8, the mean square error (842) of the output of each of neural networks 810 and 812 and the mean square error (846) of the output of each of neural networks 820 and 822 are summed (862) and fed back to the neural networks (810, 812, 820, and 822) as further errors.
According to the embodiment of fig. 8, the virtual inputs 815 and 825 each follow the same distribution, while multiple neural networks corresponding to the same virtual point magnetic source share weights. For example, neural networks 810 and 812 share weights, while neural networks 820 and 822 share weights. So that the field strength can be calculated using one of the magnetic source parameters calculated by the two networks, or using the average of the two. The virtual input 815 is provided to one of a plurality of neural networks corresponding to the same virtual point magnetic source, and the virtual input 825 is provided to one of a plurality of neural networks corresponding to the same virtual point magnetic source, such that each of the neural networks virtually corresponding to the same virtual point magnetic source receives a different virtual input. In the example of FIG. 8, virtual input 815 is provided to neural networks 810 and 820, while virtual input 825 is provided to neural networks 812 and 822. When training the neural networks (810, 812, 820, and 822), the difference in the output of each of the plurality of neural networks corresponding to the same virtual point magnetic source is set as close as possible to the training target. For example, the difference between the outputs of the respective neural networks corresponding to the same virtual point magnetic source is set to 0 or as close to a predetermined hyper-parameter as possible, and the value of the hyper-parameter is adjusted during implementation so as to achieve the goal that the magnetic source parameters of the virtual point magnetic sources output by the neural networks (810, 812, 820, and 822) are constant (the difference is as small as possible) regardless of the value of the virtual input.
Optionally, for each virtual point magnetic source, other numbers of neural networks are used to estimate the magnetic source parameters thereof, or RNN (Recurrent neural Network) is used to estimate the magnetic source parameters of the virtual point magnetic source, so as to achieve the goal that the magnetic source parameters of each virtual point magnetic source output by the neural networks are fixed regardless of the values of the virtual inputs.
Fig. 9 illustrates a schematic diagram of modeling a target magnet and estimating magnetic field strength according to still another embodiment of the present application.
Fig. 9 shows a spatial coordinate system of the space in which the target magnet is located, with the origin of the coordinate system denoted as O, the horizontal plane being the XOY plane, and the Z-axis being perpendicular to the horizontal plane. The target magnet has a circumscribed cube size of (1,0.1,0.1), the center of the target magnet is located at the origin, and its coordinates are (0,0, 0). By way of example, three virtual point magnetic sources with coordinates (-0.5,0,0), (0,0,0), (0.5,0,0) are used to fit the unknown magnet (the virtual point magnetic sources are represented by black squares in fig. 9), and a fixed value of 1 is used as the virtual input provided to the neural network.
By way of example, the neural network used to estimate the magnetic source parameters is a 3-layer 30-dimensional fully-connected network. The magnetic moment is selected as a magnetic source parameter. The inverse cubic formula provided in magnetic moment to calculate magnetic field strength is used according to the theoretical formula of calculating field strength from measured point coordinates, virtual point magnetic source coordinates and magnetic moment.
Referring to fig. 9, a plurality of training measurement points are set above both sides of the x-axis of the target magnet for acquiring training samples (the training measurement points are represented by black circles in fig. 9), and the acquired sets of < measurement point coordinates, measurement point magnetic field strength > are used as training data. The Mean Square Error (MSE) of the estimated value of the field intensity of the measured points and the actual value is used as the error fed back to the neural network. A plurality of target measurement points (the target measurement points are represented by lower triangles in fig. 9) are provided below the target magnet for testing the fitting result of the model.
Alternatively, when measuring the magnetic field strength at the measuring points, a plurality of magnetometers may be installed at specified positions on the frame of the non-magnetic conductive material to obtain a plurality of magnetic field strength data at different positions. And then multiple sets of data acquisition are performed by continuously translating the frame on the non-magnetic conductive tracks.
Although the present application has been described with reference to examples, which are intended to be illustrative only and not to be limiting of the application, changes, additions and/or deletions may be made to the embodiments without departing from the scope of the application.
Many modifications and other embodiments of the application set forth herein will come to mind to one skilled in the art to which these embodiments pertain having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is to be understood that the application is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.
Claims (10)
1. A target magnet modeling method, comprising:
establishing one or more virtual point magnetic sources representative of the target magnet, the point virtual point magnetic sources having a specified location within the space occupied by the target magnet;
arranging a plurality of training measuring points in the space around the target magnet, wherein the training measuring points have designated positions;
measuring the magnetic field intensity of each training measuring point under the influence of a target magnet;
generating training samples according to the magnetic field intensity of each training measuring point and the positions of the training measuring points to train one or more neural networks; wherein
The one or more virtual point magnetic sources are respectively provided with one or more corresponding neural networks, the neural networks corresponding to the virtual point magnetic sources are trained, and the output of the neural networks represents the magnetic source parameters of the virtual point magnetic sources corresponding to the neural networks;
according to the magnetic source parameters and positions of the one or more virtual point magnetic sources output by the neural network and the positions of the training measuring points in the training sample, calculating the magnetic field intensity generated by the one or more virtual point magnetic sources at the positions of the training measuring points in the training sample as estimated magnetic field intensity;
taking the difference between the estimated magnetic field strength and the magnetic field strength of the training sample as an error for training a neural network corresponding to each of the one or more virtual point magnetic sources;
using the output of the trained neural network as the magnetic source parameter of the virtual point magnetic source corresponding to the output of the trained neural network;
and using the position and the magnetic source parameter of each of the one or more virtual point magnetic sources as a magnetic model of the target magnet.
2. The method of claim 1, wherein
Representing the target magnet with a single virtual point magnet source; or
And setting the one or more virtual point magnetic sources to ensure that 6 times of the distance between any virtual point magnetic source is less than the distance between any training measuring point and any virtual point magnetic source.
3. The method of claim 1 or 2, further comprising:
to train the one or more neural networks, virtual inputs are provided to each of the one or more neural networks, the virtual inputs provided to the one or more neural networks obeying a specified distribution.
4. The method of claim 3, wherein
And training the one or more neural networks to ensure that the estimated magnetic field intensity at each training measuring point calculated according to the magnetic source parameters of the magnetic source of each virtual point obtained from the output of the one or more neural networks is as consistent as possible with the magnetic field intensity of the corresponding training sample.
5. The method of claim 4, wherein
And training the one or more neural networks, and enabling the magnetic source parameters of the magnetic sources of the virtual points obtained according to the output of the one or more neural networks to be consistent as much as possible under the condition of different virtual inputs.
6. The method of claim 5, wherein
The one or more virtual point magnetic sources each have a corresponding plurality of neural networks;
using the statistic value of the output of a plurality of neural networks corresponding to the one or more virtual point magnetic sources respectively as the magnetic source parameter of the virtual point magnetic source represented by the one or more virtual point magnetic sources;
and taking the difference between the outputs of the plurality of neural networks respectively corresponding to the one or more virtual point magnetic sources as a further error of training the neural networks.
7. The method of claim 6, wherein
Providing a plurality of virtual inputs, each subject to the same distribution as each other, the number of the plurality of virtual inputs being the same as the number of the plurality of neural networks to which each of the one or more virtual point magnetic sources corresponds;
providing one of the plurality of virtual inputs to each of a plurality of neural networks to which the one or more virtual point magnetic sources respectively correspond;
each of the plurality of virtual inputs is simultaneously provided to a respective one of a plurality of neural networks for each virtual point magnetic source.
8. The method of claim 7, wherein,
under the condition that the target magnet does not exist, measuring the background magnetic field intensity of each training measuring point;
measuring the real magnetic field intensity of each training measuring point under the condition that the target magnet exists;
and obtaining the magnetic field intensity of each training measuring point under the influence of the target magnet according to the difference between the real magnetic field intensity and the background magnetic field intensity.
9. The magnetic field intensity estimation method of the target measuring point under the influence of the target magnet comprises the following steps:
establishing a magnetic model of a target magnet represented by one or more virtual point magnetic sources according to the method of one of claims 1-8;
and calculating the superposition of the magnetic field intensity generated by all the virtual point magnetic sources of the one or more virtual point magnetic sources at the position of the target measuring point according to the magnetic field intensity formula and the magnetic source parameters and the positions of the one or more virtual point magnetic sources to obtain the magnetic field intensity estimation of the target measuring point under the influence of the target magnet.
10. An information processing apparatus comprising a memory, a processor and a program stored on the memory and executable on the processor, characterized in that the processor implements the method according to one of claims 1 to 9 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210738001.0A CN115130379B (en) | 2022-06-27 | 2022-06-27 | Magnet modeling and magnetic field estimation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210738001.0A CN115130379B (en) | 2022-06-27 | 2022-06-27 | Magnet modeling and magnetic field estimation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115130379A true CN115130379A (en) | 2022-09-30 |
| CN115130379B CN115130379B (en) | 2024-05-24 |
Family
ID=83380104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210738001.0A Active CN115130379B (en) | 2022-06-27 | 2022-06-27 | Magnet modeling and magnetic field estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115130379B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702568A (en) * | 2023-08-04 | 2023-09-05 | 天津天达图治科技有限公司 | Magnetic resonance imaging permanent magnet design method, system, equipment and medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070167703A1 (en) * | 2005-12-30 | 2007-07-19 | Sherman Jason T | Method for determining a position of a magnetic source |
| CN103675719A (en) * | 2013-12-17 | 2014-03-26 | 复旦大学 | Method for determining magnetic induction intensity maximum value by adoption of second-order polynomial curve fitting and implementation system |
| WO2016139135A1 (en) * | 2015-03-02 | 2016-09-09 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V. | Method and assembly for determining the position of a magnetic body by means of magnetic field sensors |
| CN110146845A (en) * | 2019-04-17 | 2019-08-20 | 杭州电子科技大学 | A kind of event driven set time electromagnet source localization method |
| US20190317159A1 (en) * | 2018-04-16 | 2019-10-17 | Commissariat à l'énergie atomique et aux énergies alternatives | Method of calibrating a network of magnetometers |
-
2022
- 2022-06-27 CN CN202210738001.0A patent/CN115130379B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070167703A1 (en) * | 2005-12-30 | 2007-07-19 | Sherman Jason T | Method for determining a position of a magnetic source |
| CN103675719A (en) * | 2013-12-17 | 2014-03-26 | 复旦大学 | Method for determining magnetic induction intensity maximum value by adoption of second-order polynomial curve fitting and implementation system |
| WO2016139135A1 (en) * | 2015-03-02 | 2016-09-09 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V. | Method and assembly for determining the position of a magnetic body by means of magnetic field sensors |
| US20190317159A1 (en) * | 2018-04-16 | 2019-10-17 | Commissariat à l'énergie atomique et aux énergies alternatives | Method of calibrating a network of magnetometers |
| CN110146845A (en) * | 2019-04-17 | 2019-08-20 | 杭州电子科技大学 | A kind of event driven set time electromagnet source localization method |
Non-Patent Citations (4)
| Title |
|---|
| 万成彪;罗诗途;王伟;张琦;翁飞兵;: "磁性目标的单点模量测距方法", 测试技术学报, no. 01, 25 February 2014 (2014-02-25) * |
| 姜浩;周鹰;杨云;: "基于等效磁矩的永磁体磁性测量方法研究", 电子世界, no. 13, 8 July 2016 (2016-07-08) * |
| 张新邦 等: "由磁矩计算磁场强度", 《数字海洋与水下攻防》, vol. 2, no. 2, 15 June 2019 (2019-06-15) * |
| 章尧卿 等: "潜艇空间磁场的建模分析与运用", 《 舰船电子工程》, 20 January 2018 (2018-01-20) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702568A (en) * | 2023-08-04 | 2023-09-05 | 天津天达图治科技有限公司 | Magnetic resonance imaging permanent magnet design method, system, equipment and medium |
| CN116702568B (en) * | 2023-08-04 | 2023-11-10 | 天津天达图治科技有限公司 | Magnetic resonance imaging permanent magnet design method, system, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115130379B (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3314247A2 (en) | System and method for characterizing ferromagnetic material | |
| CN115130379A (en) | Magnet modeling and magnetic field estimation method and device | |
| US8111059B1 (en) | Electric current locator | |
| CN104245499B (en) | System for accuracy control of erection on marine floating dock | |
| EP3537144B1 (en) | Thickness loss detection device, thickness loss detection system, thickness loss detection method, and program | |
| Luo et al. | Calibration strategy and generality test of three-axis magnetometers | |
| CN112180450B (en) | Three-dimensional reconstruction and imaging method of direct current cable based on magnetic field measurement | |
| CN114926633A (en) | High-precision pig weight estimation method based on deep learning | |
| JP2010130220A (en) | System for assisting distribution of radio communication apparatus | |
| JP2020060487A (en) | Sacrificial anode generation current estimation method and device | |
| JP2519657B2 (en) | Lightning detection method and device | |
| CN115144796A (en) | Modeling and magnetic field estimation method and device for residual magnetism and inductive magnetism of magnet in geomagnetic field | |
| CN116127314B (en) | Method for predicting underground density based on self-adaptive multi-scale deep learning network | |
| JP2004020358A (en) | Method of calculating diameter of buried object, such as reinforcement, or the like, buried in structure | |
| Zhao et al. | Construction method of geomagnetic reference map for satellite communication navigation through Kriging method | |
| Birsan et al. | The effect of roll and pitch motion on ship magnetic signature | |
| CN114114438A (en) | Quasi-three-dimensional inversion method for loop source ground-air transient electromagnetic data | |
| Demsar et al. | Time-geography in four dimensions: potential path volumes around 3D trajectories | |
| JP7298637B2 (en) | REDUCTION DETECTION SYSTEM, REDUCTION DETECTION METHOD AND PROGRAM | |
| Jeung et al. | Magnetic dipole modeling combined with material sensitivity analysis for solving an inverse problem of thin ferromagnetic sheet | |
| Moriya et al. | Basic Study on Ultrasonic Remote Leakage Position Estimation Method for Underwater Exploration | |
| Chen et al. | Accurate magnetic object localization using artificial neural network | |
| Grigoriev et al. | The results of marine electromagnetic sounding with a high-power remote source in the Kola Bay in the Barents Sea | |
| JP7384742B2 (en) | Underwater surveying method | |
| Vahdat | Noniterative method to solve 3D EIT forward problem |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |