Iron shoe positioning method and device, equipment and storage medium

Abstract

The application discloses a positioning method, device and equipment of a shoe, and a storage medium, relates to the technical field of track transportation, and realizes passive shoe identity recognition and position positioning cooperation. Step S1 gives the shoe a unique magnetic code identity; step S2 constructs a magnetic code and spectrum feature template library, and sets a benchmark for identity matching; step S3 collects dynamic magnetic field disturbance and extracts spectrum features; step S4 separates the overlapping signals of multiple shoes, corrects the attitude, and obtains independent spectrum features; step S5 calculates a set of nonlinear distances based on a dipole model; step S6 selects non-collinear sensor distances to construct an equation set, and solves position coordinates by using a least square method; and step S7 compares spectrum features, confirms magnetic codes, and integrates position and identity into positioning information. The whole achieves unique identification and accurate positioning of the shoe, the replacement frequency of the permanent magnet is far lower than that of the power supply, the shoe is suitable for complex environments of railway shunting, and the maintenance cost is greatly reduced.

Description

Technical Field

[0001] This application relates to the field of rail transportation technology, and in particular to a method, apparatus, equipment and storage medium for positioning iron shoes. Background Technology

[0002] Iron shoes are a core tool used in railway shunting operations to prevent locomotives and rolling stock from slipping. Their working principle is to convert the rolling friction between the wheel and the rail into sliding friction to achieve braking. The device consists of a bottom (including the toe and sole) and a head (including a baffle). When the wheel presses down on it, the head slides in conjunction with the wheel. Traditionally, iron shoes are subject to a numbering and registration system, and their technical condition is regularly inspected. They are distinguished by markings: white shirt with red lettering or red shirt with white lettering, indicating the responsible unit.

[0003] In recent years, with increasing levels of intelligent integration, railway shunting has commonly integrated intelligent positioning modules such as GPS, RFID, and inertial measurement units. While these modules offer accurate positioning, they require continuous power or periodic battery replacements, leading to high maintenance costs. Furthermore, the harsh conditions of railway shunting environments, including extreme temperatures, high humidity, and severe vibrations, expose these intelligent positioning modules to prolonged exposure, making them prone to problems such as circuit aging, interface corrosion, battery bulging, poor contact, sensor drift, and mechanical damage.

[0004] In summary, existing intelligent positioning iron shoes have poor environmental adaptability and short service life, making it difficult to meet the requirements for long-term stable operation. Summary of the Invention

[0005] The main objective of this application is to provide a positioning method, device, equipment, and storage medium for iron shoes, in order to solve the problems of existing intelligent positioning iron shoes in the prior art, which have poor environmental adaptability, short service life, and difficulty in meeting the requirements for long-term stable operation.

[0006] The first objective of this invention is to address the problem that traditional track shoes cannot simultaneously achieve unique identification and accurate location positioning under passive conditions in railway shunting operations, making them unsuitable for the complex environment of railway shunting and resulting in high maintenance costs. The specific solution is as follows:

[0007] A method for locating track shoes, wherein there are several track shoes, all used in locomotives and rolling stock in railway shunting operations, each track shoe is embedded with an array of permanent magnets with non-overlapping magnetic poles, and a fluxgate sensor array is laid on both sides of the railway track. The positioning method includes:

[0008] Step S1: Convert each permanent magnet array into a binary encoding sequence and use it as the magnetic encoding sequence for each iron shoe.

[0009] Step S2: Obtain the initial magnetic field signal of each iron shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal through fast Fourier transform. Integrate the magnetic coding sequence of all iron shoes and the standard spectral features to obtain a feature template library.

[0010] Step S3: During the railway shunting operation, all local magnetic field disturbance signals are continuously collected through the fluxgate sensor array, and the spectral characteristics of all local magnetic field disturbance signals are extracted by fast Fourier transform.

[0011] Step S4: Separate the spectral characteristics of all local magnetic field disturbance signals using a hybrid signal separation algorithm, and after correcting the attitude of each iron shoe, obtain the independent magnetic field signal spectral characteristics of each iron shoe.

[0012] Step S5: Based on the three-axis components of each fluxgate sensor, calculate the set of nonlinear distances from each shoe to each fluxgate sensor based on the dipole model using a nonlinear distance estimation algorithm based on the dipole model.

[0013] Step S6: Obtain the nonlinear distance from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them by the least squares method to obtain the position coordinates of the current iron shoe;

[0014] Step S7: Compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and positioning information.

[0015] Beneficial effects of steps S1 to S7:

[0016] It achieves collaborative identification and location positioning for passive iron shoes. The process involves several steps: Step S1 assigns a unique binary magnetic code sequence to each wheel shoe as its identification; Step S2 collects interference-free initial magnetic field signals and extracts standard spectral features, integrating the magnetic code and spectral features to construct a feature template library, establishing a benchmark for identification matching; Step S3 continuously collects local magnetic field disturbance signals and extracts spectral features during shunting to capture magnetic field changes in dynamic environments; Step S4 uses a hybrid signal separation algorithm to separate overlapping spectral features of multiple wheel shoes, combines the inversion of tilt angles using the three-axis components of the fluxgate sensor for attitude correction, and obtains the independent magnetic field signal spectral features of each wheel shoe; Step S5 uses a dipole model nonlinear distance estimation algorithm to calculate the nonlinear distance set from each wheel shoe to each fluxgate sensor using the three-axis components of the sensor, characterizing the nonlinear correlation between magnetic field strength and distance; Step S6 selects at least three non-collinear sensor distances to construct a nonlinear equation set, and solves it using the least squares method to obtain the wheel shoe's position coordinates; Step S7 compares the current wheel shoe's independent spectral features with the feature template library, filters the standard spectral features with the highest similarity using Euclidean distance, confirms the corresponding magnetic code sequence, and finally integrates the position coordinates and magnetic code sequence to form positioning information. This method achieves unique identification and precise location positioning for passive track shoes. Furthermore, the replacement or remagnetizing frequency of the permanent magnets is significantly lower than the replacement or charging frequency of the power supply, enabling the track shoes to adapt to the complex environment of railway shunting and thus significantly reducing maintenance costs. This method achieves coordinated identification and location positioning for passive track shoes. By assigning a unique binary magnetic code sequence to each track shoe, collecting interference-free initial magnetic field signals to construct a feature template library, continuously collecting and processing local magnetic field disturbance signals during shunting, using a hybrid signal separation algorithm and attitude correction to obtain the spectral characteristics of independent magnetic field signals, calculating a nonlinear distance set based on a dipole model nonlinear distance estimation algorithm, constructing a system of nonlinear distance equations to solve for the position coordinates, and finally integrating the position coordinates with the magnetic code sequence to form positioning information. This method achieves unique identification and precise location positioning for passive track shoes, and the replacement or remagnetizing frequency of the permanent magnets is much lower than the replacement or charging frequency of the power supply, significantly reducing maintenance costs and adapting to the complex environment of railway shunting.

[0017] When constructing a feature template library for passive metal shoe identification, factors such as external magnetic field interference, inconsistent signal acquisition posture, and noise and offset in the signal can lead to inaccurate and impure initial magnetic field signals, affecting the accuracy of subsequent feature extraction. This results in a lack of reliable comparison references for identification, reducing the accuracy and reliability of identification. To address this issue, this application further improves upon this by, in step S2, acquiring the initial magnetic field signal of each metal shoe in an interference-free environment, and extracting the standard spectral features of each initial magnetic field signal using Fast Fourier Transform. The magnetic coding sequences of all metal shoes and the standard spectral features are then integrated to obtain the feature template library, including:

[0018] Step S21: Deploy a fluxgate calibration array composed of three-axis coils in an interference-free environment, and form a zero-magnetic environment acquisition condition through magnetic shielding.

[0019] Step S22: Place the permanent magnet array of each iron shoe in the same orientation directly above the fluxgate calibration array, and collect the initial magnetic field time domain signal of each permanent magnet array through the fluxgate calibration array.

[0020] Step S23: Each initial magnetic field time domain signal is preprocessed by bandpass filtering and drift removal preprocessing respectively, and a clean magnetic field time domain signal is obtained based on an initial magnetic field time domain signal.

[0021] Step S24: Calculate a standard frequency domain signal for each pure magnetic field time domain signal using Fast Fourier Transform;

[0022] Step S25: Extract the main frequency component, harmonic phase difference and spectral amplitude of each standard frequency domain signal to obtain a standard spectral feature set based on a frequency domain signal;

[0023] Step S26: The magnetic coding sequence of each iron shoe is structurally integrated with the corresponding standard spectral feature set to obtain a standard magnetic coding feature set based on an iron shoe.

[0024] Step S27: Integrate all standard magnetic coding feature sets of iron shoes to obtain the feature template library.

[0025] Beneficial effects of steps S21 to S27:

[0026] A feature template library is constructed through a modular process to establish a precise benchmark for identifying wheeled shoes. Step S21 involves deploying a three-axis coil fluxgate calibration array and implementing magnetic shielding to create a zero-magnetic environment for data acquisition, eliminating external magnetic field interference. Step S22 places each wheeled shoe's permanent magnet array in the same orientation directly above the array to acquire the initial magnetic field time-domain signal, ensuring consistency and comparability of signal acquisition. Step S23 removes high-frequency noise and DC offset from the signal through bandpass filtering and drift removal preprocessing to obtain a clean magnetic field time-domain signal. Step S24 performs a Fast Fourier Transform on the clean signal to obtain a standard frequency-domain signal, mapping the time-domain information to the frequency-domain feature space. Step S25 extracts the dominant frequency component, harmonic phase difference, and spectral amplitude of the standard frequency-domain signal to form a standard spectral feature set characterizing the magnetic field properties of the wheeled shoes. Step S26 structurally integrates the binary magnetic code sequence of each wheeled shoe with the corresponding standard spectral feature set to generate a standard magnetic code feature set containing identification and magnetic field characteristics. Step S27 summarizes all the standard magnetic code feature sets of all wheeled shoes to construct a feature template library. The sub-steps of this module, through strict environmental control, signal preprocessing, feature extraction, and structured integration, provide a unique and stable comparison reference for the subsequently dynamically acquired magnetic field signals, ensuring the accuracy and reliability of identity recognition and laying the foundation for identity matching in passive iron shoe positioning.

[0027] This improved method constructs a precise feature template library through a series of modular processes, laying a solid foundation for the identification of metal shoes. First, a three-axis coil fluxgate calibration array is deployed and magnetically shielded to create a zero-magnetic environment for data acquisition, effectively eliminating external magnetic field interference. Next, the permanent magnet array of each metal shoe is placed in the same orientation directly above the array to acquire the initial magnetic field time-domain signal, ensuring the consistency and comparability of the acquired signals. Then, bandpass filtering and drift-reduction preprocessing are used to remove high-frequency noise and DC offset from the signal, obtaining a pure magnetic field time-domain signal. The pure signal is then converted to the frequency domain feature space using a fast Fourier transform to obtain a standard frequency domain signal. Next, the dominant frequency component, harmonic phase difference, and spectral amplitude of the standard frequency domain signal are extracted to form a standard spectral feature set that characterizes the magnetic field properties of the metal shoes. Then, the binary magnetic code sequence of each metal shoe is structurally integrated with the corresponding standard spectral feature set to generate a standard magnetic code feature set containing identification and magnetic field characteristics. Finally, the standard magnetic code feature sets of all metal shoes are summarized to construct a feature template library. This improved method, through strict environmental control, signal preprocessing, feature extraction, and structured integration, provides a unique and stable comparison reference for the subsequently dynamically acquired magnetic field signals, ensuring the accuracy and reliability of identity recognition and strongly supporting identity matching in passive iron shoe positioning.

[0028] In railway shunting operations, the magnetic field changes caused by the metal sleds are in a dynamic environment. Traditional methods struggle to capture the magnetic field disturbance signals generated when the sleds pass by accurately and in real time, and cannot convert these continuously changing signals into a form suitable for subsequent processing and analysis. This results in the inability to obtain effective raw magnetic field disturbance information for sled positioning and identification. To address this problem, this application further improves upon this by step S3, where the fluxgate sensor array continuously collects all local magnetic field disturbance signals during the railway shunting operation, and extracts the spectral features of all local magnetic field disturbance signals using a fast Fourier transform, including:

[0029] Step S31: During the railway shunting operation, the fluxgate sensor array is activated to continuously monitor the magnetic field changes when all the iron shoes pass by, and obtain several original local magnetic field disturbance signal streams.

[0030] Step S32: Sample each original local magnetic field disturbance signal stream in real time, and obtain a discrete local magnetic field disturbance time domain signal sequence based on an original local magnetic field disturbance signal stream;

[0031] Step S33: Obtain a perturbation frequency domain signal by performing a fast Fourier transform on each discrete local magnetic field perturbation time domain signal sequence;

[0032] Step S34: Extract the main frequency component, harmonic phase difference and spectral amplitude of each perturbation frequency domain signal, and obtain a perturbation spectral feature set based on a perturbation frequency domain signal.

[0033] Beneficial effects of steps S31 to S34:

[0034] Through dynamic acquisition and feature extraction, the system provides a frequency domain representation of the original magnetic field disturbance information for shunting wheel positioning. Specifically, step S31 activates a fluxgate sensor array to continuously monitor the magnetic field changes as the wheel passes through during shunting, acquiring the original local magnetic field disturbance signal stream and capturing magnetic field anomalies in dynamic environments. Step S32 samples the original signal stream in real time, converting it into a discrete local magnetic field disturbance time-domain signal sequence, transforming continuous changes into processable discrete data. Step S33 uses a Fast Fourier Transform to convert the discrete time-domain sequence into a disturbance frequency-domain signal, achieving a mapping from time-domain information to frequency-domain features. Step S34 extracts the dominant frequency component, harmonic phase difference, and spectral amplitude of the disturbance frequency-domain signal, forming a disturbance spectral feature set. This sub-step captures the magnetic field disturbance caused by the wheel in real time during shunting, and through sampling, transformation, and feature extraction, simplifies the complex time-domain signal into a frequency-domain feature set characterizing the magnetic field properties, providing the original input for subsequent signal separation and identification, ensuring the effective acquisition and preliminary structuring of magnetic field information in dynamic environments. This improved method provides an effective frequency domain representation of the original magnetic field disturbance information for track shunting by employing a series of dynamic acquisition and feature extraction operations. During railway shunting operations, a fluxgate sensor array is first activated to continuously monitor magnetic field changes as the track shunting passes, successfully acquiring the original local magnetic field disturbance signal stream, which can keenly capture magnetic field anomalies in dynamic environments. Next, the original signal stream is sampled in real time and transformed into a discrete local magnetic field disturbance time-domain signal sequence, converting the continuously changing signal into processable discrete data, laying the foundation for subsequent analysis. Then, a Fast Fourier Transform is used to convert the discrete time-domain sequence into a disturbance frequency-domain signal, realizing the mapping from time-domain information to frequency-domain features and uncovering the hidden frequency-domain information in the signal. Finally, the dominant frequency component, harmonic phase difference, and spectral amplitude of the disturbance frequency-domain signal are extracted to form a disturbance spectral feature set. This improved method captures the magnetic field disturbances caused by the track shoe in real time during shunting. Through a series of operations such as sampling, transformation and feature extraction, the complex time-domain signal is simplified into a frequency-domain feature set that can characterize the magnetic field properties. This provides a reliable raw input for subsequent signal separation and identification, ensuring that magnetic field information can be effectively acquired and preliminarily structured in dynamic environments.

[0035] In the dynamic environment of railway shunting operations, when multiple trackers exist simultaneously, the spectral characteristics of their generated local magnetic field disturbance signals overlap and mix. Furthermore, different tracker postures cause deviations in the magnetic field direction, making it difficult to accurately extract the independent and accurate magnetic field signal spectral characteristics of each tracker from the complex mixed signals. This, in turn, affects subsequent distance estimation and precise positioning of the trackers. To address this problem, this application further improves upon this by using a mixed signal separation algorithm in step S4 to separate the spectral characteristics of all local magnetic field disturbance signals. After correcting the posture of each tracker, the independent magnetic field signal spectral characteristics of each tracker are obtained, including:

[0036] Step S41: Input all perturbation spectral feature sets into the PSO-ICA mixed signal separation algorithm to separate several overlapping spectral features of the iron shoes, and obtain several separated perturbation spectral feature sets.

[0037] Step S42: The tilt angle of each shoe is inverted by the spatial distribution characteristics of the three-axis components of the fluxgate sensor array. Attitude correction is performed on all the separated perturbation spectrum feature sets. An attitude-corrected spectrum feature set is obtained based on a separated perturbation spectrum feature set.

[0038] Step S43: Group all the attitude-corrected spectral feature sets according to the uniqueness of the cluster center of each iron shoe to obtain the independent magnetic field signal spectral features of each iron shoe.

[0039] Beneficial effects of steps S41 to S43:

[0040] By performing mixed signal separation, attitude correction, and grouping processing, the independent magnetic field spectrum features of each track shoe are extracted from the overlapping disturbance signals. Step S41 inputs the overlapping disturbance spectrum feature set of multiple track shoes into the PSO-ICA algorithm to separate the mutually interfering spectrum components, obtaining several separated disturbance spectrum feature sets, thus solving the problem of multi-target signal aliasing. Step S42 uses the spatial distribution characteristics of the three-axis components of the fluxgate sensor to invert the tilt angle of the track shoes, performs attitude correction on the separated feature sets, eliminates the magnetic field direction deviation caused by track shoe attitude changes, and obtains the attitude-corrected spectrum feature set. Step S43 groups the corrected feature sets according to the uniqueness of each track shoe cluster center, forming independent magnetic field signal spectrum features for each track shoe. The sub-steps of this section effectively remove the interference and attitude influence of multiple track shoe signals, providing clean and independent spectrum feature input for subsequent distance estimation and positioning, ensuring the identifiability and accuracy of track shoe magnetic field signals in dynamic shunting environments.

[0041] This improved method successfully extracts the independent magnetic field spectrum features of each shoe from overlapping perturbation signals through a series of targeted processing steps. First, step S41 inputs the perturbation spectrum feature set of multiple overlapping shoes into the PSO-ICA hybrid signal separation algorithm. This algorithm, with its powerful signal separation capability, effectively separates mutually interfering spectral components, obtaining several separated perturbation spectrum feature sets, successfully solving the key problem of multi-target signal aliasing. Next, step S42 fully utilizes the spatial distribution characteristics of the three-axis components of the fluxgate sensor to inversely calculate the tilt angle of each shoe, and performs attitude correction on all separated perturbation spectrum feature sets accordingly, eliminating the magnetic field direction deviation caused by changes in shoe attitude, and obtaining attitude-corrected spectrum feature sets. Finally, step S43 groups all attitude-corrected spectrum feature sets according to the uniqueness of each shoe's cluster center, accurately classifying the corrected feature sets to form independent magnetic field signal spectrum features for each shoe. This improved method effectively eliminates interference from multiple track shoe signals and attitude effects, providing a clean, independent, and accurate spectral feature input for subsequent distance estimation and positioning, ensuring that the track shoe magnetic field signal has high identifiability and accuracy in dynamic shunting environments.

[0042] In the dynamic scenario of railway shunting, there is a nonlinear magnetic field correlation between the track shoe and the fluxgate sensor (magnetic field strength is inversely proportional to the cube of the distance), and the sensor's orientation sensitivity makes signal matching susceptible to interference. Traditional methods struggle to accurately establish the spatial correspondence between the independent magnetic field signal of the track shoe and the three-axis components of the sensor, making it impossible to quantify the true spatial distance under magnetic field attenuation. This hinders the reliable construction of subsequent position equations and affects the accuracy of the spatial relationship quantification for passive track shoe positioning. To address this issue, this application further improves upon this by step S5, which calculates the set of nonlinear distances from each track shoe to each fluxgate sensor based on the dipole model using a dipole model nonlinear distance estimation algorithm based on the three-axis components of each fluxgate sensor. This includes:

[0043] Step S51: Acquire the triaxial component data of each fluxgate sensor;

[0044] Step S52: Spatially align the independent magnetic field signal spectrum characteristics of each iron shoe with the triaxial component data of each fluxgate sensor to match the correspondence between the transmit and receive signals of each iron shoe and each fluxgate sensor, and obtain the aligned spectrum characteristics and triaxial component pairing data.

[0045] Step S53: Input the aligned spectral features and triaxial component pairing data into the nonlinear distance estimation algorithm of the dipole model. Based on the inverse cubic relationship between magnetic field strength and distance, and the directional sensitivity characteristics of the fluxgate sensor, calculate the nonlinear distance from each iron shoe to each fluxgate sensor.

[0046] Step S54: The nonlinear distance from each iron shoe to each fluxgate sensor is structurally integrated according to the uniqueness of the cluster center of each iron shoe and the number of the fluxgate sensor, to obtain the set of nonlinear distances from each iron shoe to each fluxgate sensor.

[0047] Beneficial effects of steps S51 to S54:

[0048] By acquiring sensor data, aligning signals, calculating nonlinear distances, and integrating them in a structured manner, the precise measurement and organization of the nonlinear distances from each shoe to each fluxgate sensor are achieved. Step S51 acquires the triaxial component data of each fluxgate sensor, providing a basis for the sensor's directional sensitivity characteristics for distance estimation. Step S52 spatially aligns the independent magnetic field signal spectrum characteristics of the shoe with the triaxial component data of each sensor, matches the correspondence between transmitted and received signals, and obtains the aligned spectrum characteristics and triaxial component paired data to ensure the accuracy of the correlation between the signal and the sensor. Step S53 inputs the paired data into the dipole model nonlinear distance estimation algorithm. Based on the inverse relationship between magnetic field strength and the cube of distance and the sensor's directional sensitivity characteristics, the nonlinear distance from each shoe to each fluxgate sensor is calculated, characterizing the real spatial correlation under magnetic field attenuation. Step S54 integrates the nonlinear distances from each shoe to each fluxgate sensor in a structured manner according to the uniqueness of the shoe cluster center and the fluxgate sensor number, to obtain the set of nonlinear distances from each shoe to each fluxgate sensor. The sub-steps of this section, through precise data association, physical model calculation, and orderly integration, provide reliable raw distance data for the subsequent construction of distance equations to solve for position coordinates, supporting the quantitative expression of spatial relationships in passive tracker positioning. This improved method achieves accurate measurement and orderly organization of the nonlinear distance from each tracker to each fluxgate sensor through a four-step collaborative process of "data acquisition - signal alignment - nonlinear calculation - structured integration". Step S51 acquires the triaxial component data of the sensors, providing a basis for the direction-sensitive characteristics of distance estimation and strengthening the physical basis for spatial correlation. Step S52 matches the independent magnetic field signal spectrum characteristics of the track shoes with the triaxial components of the sensors through spatial alignment, ensuring accurate correspondence between transmitted and received signals and eliminating matching deviations caused by signal aliasing. Step S53 uses a nonlinear distance estimation algorithm based on the dipole model, combined with the inverse cubic relationship of magnetic field strength and the direction-sensitive characteristics of the sensors, to accurately calculate the nonlinear distance from each track shoe to each sensor, characterizing the real spatial correlation under magnetic field attenuation. Step S54 integrates the distance data in a structured manner according to the uniqueness of the track shoe cluster center and the sensor number, forming a set of nonlinear distances from each track shoe to each sensor, providing a standardized and reliable distance input for the subsequent construction of a set of nonlinear distance equations. This improvement, through precise data correlation, physical model calculation, and orderly integration, effectively supports the quantitative expression of spatial relationships in passive track shoe positioning, ensures the accuracy and robustness of position coordinate solutions, and improves the spatial accuracy and reliability of track shoe positioning in dynamic shunting environments.

[0049] In railway shunting operations, the positioning of the track shoe relies on measurement data from fluxgate sensors. However, simply acquiring the nonlinear distance data from the track shoe to each sensor is insufficient to directly determine the track shoe's specific spatial coordinates. Furthermore, due to sensor measurement errors and the complexity of the actual environment, relying solely on distance data from a limited number of sensors makes it difficult to guarantee the accuracy and stability of the positioning, failing to meet the demand for precise track shoe positioning in dynamic shunting environments. To address this issue, this application further improves upon this by step S6, which involves acquiring the nonlinear distances from the current track shoe to at least three non-collinear fluxgate sensors, constructing a set of nonlinear distance equations, and solving these equations using the least squares method to obtain the current track shoe's position coordinates, including:

[0050] Step S61: Obtain the pre-calibrated position coordinates of each fluxgate sensor;

[0051] Step S62: Select the nonlinear distances from the current iron shoe to at least three non-collinear fluxgate sensors from the nonlinear distance set as the nonlinear distance subset of the current iron shoe, and use the pre-calibrated position coordinates corresponding to the selected fluxgate sensors as the sensor position coordinate subset of the current iron shoe.

[0052] Step S63: Construct a set of nonlinear distance equations for the nonlinear distance subset of the current iron shoe and the sensor position coordinate subset based on the dipole model distance formula;

[0053] Step S64: Solve the nonlinear distance equations of the current iron shoe using the least squares method to obtain the current position coordinates of the iron shoe.

[0054] Beneficial effects of steps S61 to S64:

[0055] By acquiring sensor positions, filtering effective distances, and constructing and solving a set of equations, the precise calculation of the current shoe's position coordinates is achieved. Step S61 acquires the pre-calibrated position coordinates of each fluxgate sensor, providing a reference for spatial positioning. Step S62 selects the nonlinear distances from the current shoe to at least three non-collinear sensors as a subset from the nonlinear distance set, and simultaneously acquires the corresponding pre-calibrated position coordinates of the sensors as a subset, ensuring the effectiveness of distance-position matching. Step S63, based on the dipole model distance formula, combines the nonlinear distance subset with the sensor position coordinate subset to construct a set of nonlinear distance equations describing the spatial relationship between the shoe and the sensors. Step S64 solves this set of equations using the least squares method, iteratively optimizing the minimum sum of squared residuals to obtain the current shoe's position coordinates. This sub-step transforms the sensor's measured nonlinear distances into spatial coordinates, and through multi-sensor geometric constraints and nonlinear model solving, provides accurate spatial position data for shoe positioning, supporting the subsequent integration of identity and position information to form a complete positioning result. This improved method achieves precise calculation of the current shoe's position coordinates through a series of orderly and rigorous steps. Step S61 obtains the pre-calibrated position coordinates of each fluxgate sensor, providing a stable and reliable reference for the entire positioning process and ensuring that subsequent calculations are based on an accurate spatial framework. Step S62 carefully selects the nonlinear distances from the current shoe to at least three non-collinear fluxgate sensors as a subset from the nonlinear distance set, and simultaneously obtains the pre-calibrated position coordinates of the corresponding sensors. This selection method fully utilizes the geometric constraints of multiple sensors, ensuring the effectiveness of distance and position matching and effectively avoiding positioning ambiguity caused by sensor collinearity. Step S63, based on the dipole model distance formula, tightly combines the nonlinear distance subset with the sensor position coordinate subset to construct a set of nonlinear distance equations that can accurately describe the spatial relationship between the shoe and the sensors, providing a scientifically reasonable mathematical model for subsequent solutions. Step S64 solves the set of equations using the least squares method, effectively reducing the impact of sensor measurement errors on the positioning results by iteratively optimizing the minimum value of the residual square sum, and finally obtaining the accurate position coordinates of the current shoe. This improved method successfully transforms the nonlinear distance measured by the sensors into accurate spatial coordinates. Through the geometric constraints of multiple sensors and the solution of the nonlinear model, it provides reliable spatial location data for track shunting, which strongly supports the subsequent integration of identity and location information to form a complete positioning result. This significantly improves the accuracy and stability of track shunting positioning in dynamic shunting environments.

[0056] In railway shunting operations, although the magnetic field signal spectrum characteristics and location coordinates of the trackers can be obtained, this information alone is insufficient to determine the trackers' specific identities. Since multiple trackers may exist on-site, and their magnetic field signal spectrum characteristics may share some similarities, and there is a lack of an effective means to accurately associate the trackers' magnetic field characteristics with pre-defined identification markers, it becomes difficult to precisely identify the unique identity of each tracker. Consequently, complete and accurate identification and positioning information cannot be formed, affecting the real-time and accurate monitoring of the trackers' status during shunting operations. To address this problem, this application further improves upon this by step S7, comparing the current tracker's magnetic field signal spectrum characteristics with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrating the current tracker's location coordinates and the magnetic coding sequence corresponding to the highest similarity standard spectrum feature into the current tracker's identification and positioning information, including:

[0057] Step S71: Calculate the Euclidean distance between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes in the feature template library to obtain the set of Euclidean distance values ​​between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes.

[0058] Step S72: Sort all elements in the Euclidean distance value set in ascending order, and obtain the element with the smallest Euclidean distance value, and the magnetic coding sequence corresponding to the element with the smallest distance value in the feature template library, which is the magnetic coding sequence corresponding to the current iron shoe.

[0059] Step S73: The current location coordinates of the iron shoe are structurally integrated with the magnetic code sequence corresponding to the current iron shoe to obtain the identity and location information of the current iron shoe.

[0060] Beneficial effects of steps S71 to S73:

[0061] By comparing Euclidean distances, filtering similarities, and integrating information, the precise association between the current iron shoe's identity and location information is achieved. Step S71 calculates the Euclidean distance between the current iron shoe's independent magnetic field signal spectrum features and all standard iron shoe spectrum features in the feature template library, generating a set of Euclidean distance values ​​representing the similarity between the two, providing a quantitative basis for identity matching. Step S72 sorts this set in ascending order, selects the element with the smallest distance, and obtains its corresponding magnetic coding sequence in the feature template library, thereby confirming the unique identity of the current iron shoe. Step S73 structurally integrates the current iron shoe's position coordinates obtained in step S6 with the aforementioned magnetic coding sequence to form location information containing identity identifiers and spatial locations. The sub-steps of this section ensure precise matching between dynamically acquired iron shoe signals and preset identity benchmarks through rigorous similarity measurement and data association, while reliably binding position coordinates with identity information, ultimately outputting complete passive iron shoe positioning results to support real-time monitoring of iron shoe status during shunting operations. This improved method successfully achieves precise association between the current iron shoe's identity and location information through a series of logically tight and operationally rigorous steps. Step S71 calculates the Euclidean distance between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes in the feature template library. This operation can quantify the degree of difference between the current iron shoe features and each standard feature in the template library, and produce a set of Euclidean distance values ​​representing the similarity between the two, providing a scientific and reliable quantitative basis for subsequent identity matching. Step S72 sorts the set of Euclidean distance values ​​in ascending order and selects the element with the smallest distance. Since the smaller the Euclidean distance, the higher the feature similarity, the standard spectrum feature corresponding to the element with the smallest distance is most similar to the current iron shoe features. Then, the magnetic coding sequence corresponding to the element in the feature template library is obtained to accurately confirm the unique identity of the current iron shoe. Step S73 integrates the current iron shoe position coordinates obtained in step S6 with the determined magnetic coding sequence to organically combine the identity identifier and spatial location information to form complete positioning information containing the identity identifier and spatial location. This improved method ensures that the dynamically acquired track shoe signals can be accurately matched with the preset identity benchmark through a strict similarity measurement and data association mechanism. At the same time, it reliably binds the location coordinates with the identity information and finally outputs a complete passive track shoe positioning result. This provides strong support for real-time monitoring of track shoe status during shunting operations and effectively improves the safety and management efficiency of shunting operations.

[0062] To achieve the above objectives, this application also provides the following technical solutions:

[0063] A positioning device for a metal shoe, the positioning device being applied to the positioning method described above, the positioning device comprising:

[0064] The magnetic coding sequence conversion module for iron shoes is used to convert each permanent magnet array into a binary coding sequence and use it as the magnetic coding sequence for each iron shoe.

[0065] The feature template library acquisition module is used to acquire the initial magnetic field signal of each iron shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal through fast Fourier transform. The magnetic coding sequence of all iron shoes and the standard spectral features are integrated to obtain the feature template library.

[0066] The disturbance signal spectrum feature acquisition module is used to continuously collect all local magnetic field disturbance signals through the fluxgate sensor array during the railway shunting operation, and extract the spectrum features of all local magnetic field disturbance signals through fast Fourier transform.

[0067] The independent magnetic field signal spectrum feature separation module is used to separate the spectrum features of all local magnetic field disturbance signals through a hybrid signal separation algorithm, and after correcting the attitude of each iron shoe, obtain the independent magnetic field signal spectrum features of each iron shoe.

[0068] The nonlinear distance set calculation module is used to calculate the nonlinear distance set from each shoe to each fluxgate sensor based on the dipole model using a nonlinear distance estimation algorithm based on the three-axis components of each fluxgate sensor.

[0069] The iron shoe position coordinate solution module is used to obtain the nonlinear distance from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them by the least squares method to obtain the position coordinates of the current iron shoe.

[0070] The "Iron Shoe Identity and Location Information Acquisition Module" is used to compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and location information.

[0071] To achieve the above objectives, this application also provides the following technical solutions:

[0072] An electronic device includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the positioning method described above.

[0073] To achieve the above objectives, this application also provides the following technical solutions:

[0074] A computer-readable storage medium storing program instructions that, when executed by a processor, enable the positioning method described above. Attached Figure Description

[0075] Figure 1 This is a flowchart illustrating the steps of one embodiment of the positioning method for iron shoes according to this application;

[0076] Figure 2 This is a schematic diagram of the functional modules of a positioning device for a steel shoe according to an embodiment of this application;

[0077] Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of this application;

[0078] Figure 4 This is a schematic diagram of the structure of one embodiment of the storage medium of this application. Detailed Implementation

[0079] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0080] The terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0081] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0082] like Figure 1 As shown, this embodiment provides an example of a positioning method for iron shoes. In this embodiment, there are several iron shoes, all of which are used in locomotives and rolling stock in railway shunting operations. Each iron shoe is embedded with a permanent magnet array with non-overlapping magnetic poles. Magnetic fluxgate sensor arrays are laid on the rail webs on both sides of the railway track.

[0083] Specifically, the positioning method includes the following steps:

[0084] Step S1: Convert each permanent magnet array into a binary encoding sequence and use it as the magnetic encoding sequence for each iron shoe.

[0085] Preferably, the essence of step S1 is to utilize the unique magnetic pole arrangement of the permanent magnet array built into the iron shoe to transform its physical structure into a machine-recognizable binary encoding sequence (i.e., magnetic encoding sequence). The permanent magnet array of each iron shoe consists of several independent magnets arranged according to a fixed geometric pattern (such as a one-dimensional linear array or a two-dimensional grid array). The magnetic pole (north pole N / south pole S) of each magnet has two stable states, corresponding to 1 and 0 of the binary bits. By defining the mapping rules between the magnetic pole states and the binary bits, the physical arrangement of the array can be uniquely converted into a binary sequence, which serves as the basis for the iron shoe's identity authentication.

[0086] Preferably, the specific mapping rule can be set as N pole corresponding to 1 and S pole corresponding to 0, or S pole corresponding to 1 and N pole corresponding to 0, but all of them need to be unified in advance. Let the permanent magnet array be a two-dimensional grid of m rows and n columns, or a one-dimensional linear array of length k, with a total number of magnets N = m × n or N = k. The position of each magnet is denoted as (i, j) or i, where i, j ∈ [1, N]. For the p-th iron shoe (p = 1, 2, ..., P, where P is the total number of iron shoes), the magnetic pole state of its permanent magnet array at the i-th position is s. p (i)(s p (i)=1 represents the N pole, 0 represents the S pole), then the magnetic code sequence B of the iron shoe p For B p =[s p (1),s p (2),...,s p [(N)]2. Where the right subscript 2 represents binary, and the sequence length is N bits, then theoretically 2 to the power of N unique codes can be generated.

[0087] Preferably, a redundancy mechanism can be introduced to prevent code duplication. The minimum array size N is determined based on the total number of iron shoes P, satisfying 2^N>P (e.g., when P=100, N≥7, 2^7=128>100), ensuring redundancy in the code combination to avoid duplication.

[0088] Preferably, for the selection of permanent magnets, high-performance permanent magnet materials such as neodymium iron boron can be used, and the size of a single magnet is uniform, which can be set to 5mm×5mm×2mm to ensure consistent magnetic field strength. An insulating substrate (such as epoxy resin board) is embedded in the bottom of the iron shoe, and N magnet mounting slots are processed according to the design pattern, with a fixed slot spacing (such as 10mm) to avoid magnetic interaction interference.

[0089] Preferably, during the numbering and acquisition process, a gaussmeter with an accuracy of ±0.1mT can be used to scan the magnetic field direction on the surface of the magnets in each iron shoe array, and record the N / S states of N magnets (example: the states of the three magnets in a one-dimensional iron shoe array are [N,S,N], corresponding to s_p(1)=1, s_p(2)=0, s_p(3)=1). Then, according to the mapping rule (N=1,S=0), it is converted into a binary sequence. In the above example, B_p=(1,0,1)_2=5 is generated, where the decimal auxiliary identifier is actually stored as the binary string "101".

[0090] Step S2: Obtain the initial magnetic field signal of each iron shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal through fast Fourier transform. Integrate the magnetic coding sequence of all iron shoes and the standard spectral features to obtain a feature template library.

[0091] Further, in step S2, the initial magnetic field signal of each metal shoe in an interference-free environment is obtained, and the standard spectral features of each initial magnetic field signal are extracted by fast Fourier transform. The magnetic coding sequences of all metal shoes and the standard spectral features are integrated to obtain a feature template library, which specifically includes the following steps:

[0092] Step S21: Deploy a fluxgate calibration array consisting of three-axis coils in an interference-free environment, and form a zero-magnetic environment acquisition condition through magnetic shielding.

[0093] Preferably, the fluxgate calibration array can be a fluxgate calibration array composed of three orthogonal coils with three orthogonal X / Y / Z axes, each coil having N=500 turns and a wire diameter of 0.5mm. The signal acquisition area is 30cm in diameter in the center of the coil. The array shell can be a permalloy magnetic shield with a thickness of 2mm and a relative permeability μr≥50000 to shield against external geomagnetic fields and power frequency interference.

[0094] Preferably, the magnetic field strength in the central region of the array can be monitored by a gaussmeter with an accuracy of ±0.01nT, and the coil current can be adjusted to counteract the residual magnetic field until the ambient magnetic field strength stabilizes at ≤0.1nT (this calibration threshold needs to be better than 1 / 1000 of the magnetic field fluctuation in the railway site environment) to form a zero magnetic environment acquisition condition.

[0095] Step S22: Place the permanent magnet array of each iron shoe in the same orientation directly above the fluxgate calibration array, and collect the initial magnetic field time domain signal of each permanent magnet array through the fluxgate calibration array.

[0096] Preferably, the attitude control of each iron shoe can be achieved by using a mechanical positioning fixture with a repeatability accuracy of ±0.5mm to fix the permanent magnet array at the bottom of the iron shoe directly above the array, with a vertical distance of 10cm to avoid magnetic saturation, ensuring that the attitude is consistent each time it is collected, i.e., pitch angle ≤0.1° and roll angle ≤0.1°.

[0097] Preferably, the sensitivity of the fluxgate sensor can be set to 10pT, the bandwidth to DC-100Hz, and the output of three-axis magnetic field components Bx, By, and Bz. These components are converted into digital signals by a 16-bit ADC with a sampling rate of 1kHz and recorded as the initial magnetic field time-domain signal S(t)=[Bx(t),By(t),Bz(t)]. The sampling duration can be set to 1.024s to correspond to 1024 sampling points, which meets the subsequent FFT resolution requirements and simplifies the calculation.

[0098] Step S23: Each initial magnetic field time domain signal is preprocessed by bandpass filtering and drift removal preprocessing to obtain a clean magnetic field time domain signal based on an initial magnetic field time domain signal.

[0099] Preferably, the bandpass filter is a fourth-order Butterworth bandpass filter with a passband set from 1 Hz to 10 Hz. This calibration is based on the fact that the main frequency of the magnetic field of the permanent magnet in the shoe is concentrated between 2 Hz and 8 Hz, with low-frequency drift below 1 Hz and high-frequency environmental noise above 10 Hz.

[0100] Preferably, the drift-free preprocessing can be performed by calculating the DC offset of the signal and subtracting the DC offset from the original signal to obtain the pure magnetic field time-domain signal.

[0101] Step S24: Calculate a standard frequency domain signal for each pure magnetic field time domain signal using Fast Fourier Transform.

[0102] Preferably, the Fast Fourier Transform can be performed using the radix-2 decimation-time (FFT) algorithm, with a computational complexity of O(NlogN). The input data is the pure magnetic field time-domain signal output by S23 (length N=1024), and the sampling frequency F... s It can be set to 1kHz, with a frequency resolution of Δf=F. s / N=0.9766Hz, close to 1Hz, which meets the main frequency resolution requirements.

[0103] Preferably, the frequency domain signal is in complex form. ,in, For frequency domain signals, Corresponding frequency Then k = 0, 1, 2, ..., 511. For spectral amplitude, For phase.

[0104] Step S25: Extract the main frequency component, harmonic phase difference and spectral amplitude of each standard frequency domain signal to obtain a standard spectral feature set based on a frequency domain signal.

[0105] Preferably, the isomorphic search spectrum amplitude of the dominant frequency component The frequency corresponding to the maximum value is obtained; the harmonic phase difference is obtained by calculating the phase difference between the harmonics that are integer multiples of the main frequency and the fundamental frequency; the spectral amplitude is obtained by extracting the amplitude vector composed of each frequency component in the 1Hz to 10Hz frequency band.

[0106] Step S26: The magnetic coding sequence of each iron shoe is structurally integrated with the corresponding standard spectral feature set to obtain a standard magnetic coding feature set based on an iron shoe.

[0107] Preferably, the structured integration is illustrated in the following JSON format example:

[0108] {

[0109] "magnetic_code": "101011", / / Binary magnetic code sequence

[0110] "standard_feature_set": {

[0111] "main_freq": 3.9, / / Main frequency component (Hz)

[0112] "harmonic_phase_diff": [1.57, -0.79], / / Harmonic phase difference (rad, h=2,3)

[0113] "spectrum_amplitude": [0.2, 0.5, 0.8, ...] / / Spectrum amplitude vector from 1Hz to 10Hz

[0114] },

[0115] "calibration_timestamp": "XXXXYYZZTAABBCCZ" / / Data collection timestamp

[0116] }

[0117] Step S27: Integrate all standard magnetic coding feature sets of iron shoes to obtain the feature template library.

[0118] Beneficial effects of steps S21 to S27:

[0119] A feature template library is constructed through a modular process to establish a precise benchmark for identifying wheeled shoes. Step S21 involves deploying a three-axis coil fluxgate calibration array and implementing magnetic shielding to create a zero-magnetic environment for data acquisition, eliminating external magnetic field interference. Step S22 places each wheeled shoe's permanent magnet array in the same orientation directly above the array to acquire the initial magnetic field time-domain signal, ensuring consistency and comparability of signal acquisition. Step S23 removes high-frequency noise and DC offset from the signal through bandpass filtering and drift removal preprocessing to obtain a clean magnetic field time-domain signal. Step S24 performs a Fast Fourier Transform on the clean signal to obtain a standard frequency-domain signal, mapping the time-domain information to the frequency-domain feature space. Step S25 extracts the dominant frequency component, harmonic phase difference, and spectral amplitude of the standard frequency-domain signal to form a standard spectral feature set characterizing the magnetic field properties of the wheeled shoes. Step S26 structurally integrates the binary magnetic code sequence of each wheeled shoe with the corresponding standard spectral feature set to generate a standard magnetic code feature set containing identification and magnetic field characteristics. Step S27 summarizes all the standard magnetic code feature sets of all wheeled shoes to construct a feature template library. The sub-steps of this module, through strict environmental control, signal preprocessing, feature extraction, and structured integration, provide a unique and stable comparison reference for the subsequently dynamically acquired magnetic field signals, ensuring the accuracy and reliability of identity recognition and laying the foundation for identity matching in passive iron shoe positioning.

[0120] Step S3: During railway shunting operations, all local magnetic field disturbance signals are continuously collected through a fluxgate sensor array, and the spectral characteristics of all local magnetic field disturbance signals are extracted through fast Fourier transform.

[0121] Further, step S3 involves continuously acquiring all local magnetic field disturbance signals through a fluxgate sensor array during railway shunting operations, and extracting the spectral characteristics of all local magnetic field disturbance signals using a fast Fourier transform. This specifically includes the following steps:

[0122] Step S31: During railway shunting operations, the fluxgate sensor array is activated to continuously monitor the magnetic field changes as all the iron shoes pass by, thereby obtaining several original local magnetic field disturbance signal streams.

[0123] Preferably, in order to accurately acquire the original local magnetic field disturbance signal flow, the fluxgate sensor array is laid along the waist of the railway track on both sides at a spacing of 50m. Each group contains 3 sensors, covering a transverse range of ±1m of the track. The sensor model is a high-sensitivity fluxgate with a sensitivity of 5pT, a bandwidth of DC-200Hz, and outputs three-axis magnetic field components (Bx, By, Bz).

[0124] It is worth noting that the fluxgate sensor array does not directly contact the train.

[0125] Preferably, when shunting operation is started, the operation instruction is obtained through the dispatching device API, and the array is activated to enter the listening mode; when the iron shoe enters the sensor coverage area, that is, when the magnetic signal strength is ≥5nT trigger threshold, the magnetic field change is continuously collected to generate the original local magnetic field disturbance signal stream.

[0126] Preferably, for ease of management, the signal stream can be stored in the order of track segment - sensor number - timestamp.

[0127] Step S32: Sample each original local magnetic field disturbance signal stream in real time, and obtain a discrete local magnetic field disturbance time domain signal sequence based on an original local magnetic field disturbance signal stream.

[0128] Preferably, the sampling rate can be set according to the shunting speed (e.g., ≤15km / h, i.e. 4.17m / s) and the sensor coverage range (e.g., lateral ±1m, the iron shoe passage time is about 0.48s). The real-time sampling rate is set to 1.2kHz to satisfy the Nyquist sampling theorem, avoid high-frequency signal aliasing, and the sampling duration is 2 seconds to cover the entire process of the iron shoe from entering to leaving the sensor.

[0129] Preferably, the discretization process can use a 16-bit ADC analog-to-digital converter to generate a discrete local magnetic field perturbation time-domain signal sequence with three-axis components from the analog signal output of the fluxgate, for each original signal stream.

[0130] Step S33: Obtain a perturbation frequency domain signal by performing a fast Fourier transform on each discrete local magnetic field perturbation time domain signal sequence.

[0131] Preferably, the Fast Fourier Transform (FFT) uses a radix-2 time decimation FFT, with a computational complexity of O(Nlog₂N). The discrete-time sequence output in step S32 needs to be padded with zeros to an integer power of 2 to meet the FFT requirements. For example, the length N = 2048, the sampling frequency Fs = 1.2 kHz, and the frequency resolution Δf = Fs / N = 1.2 × 10³ / 2048, which is approximately 0.586 Hz.

[0132] Preferably, the frequency domain signal is represented in the same way as in step S24, and will not be described again here.

[0133] Step S34: Extract the main frequency component, harmonic phase difference and spectral amplitude of each perturbation frequency domain signal, and obtain a perturbation spectral feature set based on a perturbation frequency domain signal.

[0134] Preferably, step S34 is similar to step S25, and will not be described again here.

[0135] Beneficial effects of steps S31 to S34:

[0136] Through dynamic acquisition and feature extraction, the system provides a frequency domain representation of the original magnetic field disturbance information for shunting wheel positioning. Specifically, step S31 activates a fluxgate sensor array to continuously monitor the magnetic field changes as the wheel passes through during shunting, acquiring the original local magnetic field disturbance signal stream and capturing magnetic field anomalies in dynamic environments. Step S32 samples the original signal stream in real time, converting it into a discrete local magnetic field disturbance time-domain signal sequence, transforming continuous changes into processable discrete data. Step S33 uses a Fast Fourier Transform to convert the discrete time-domain sequence into a disturbance frequency-domain signal, achieving a mapping from time-domain information to frequency-domain features. Step S34 extracts the dominant frequency component, harmonic phase difference, and spectral amplitude of the disturbance frequency-domain signal, forming a disturbance spectral feature set. This sub-step captures the magnetic field disturbance caused by the wheel in real time during shunting, and through sampling, transformation, and feature extraction, simplifies the complex time-domain signal into a frequency-domain feature set characterizing the magnetic field properties, providing the original input for subsequent signal separation and identification, ensuring the effective acquisition and preliminary structuring of magnetic field information in dynamic environments.

[0137] Step S4: Separate the spectral characteristics of all local magnetic field disturbance signals using a hybrid signal separation algorithm, and after correcting the attitude of each iron shoe, obtain the independent magnetic field signal spectral characteristics of each iron shoe.

[0138] Further, in step S4, the spectral characteristics of all local magnetic field disturbance signals are separated using a hybrid signal separation algorithm, and the independent magnetic field signal spectral characteristics of each iron shoe are obtained after correcting the attitude of each iron shoe. This specifically includes the following steps:

[0139] Step S41: Input all perturbation spectral feature sets into the PSO-ICA mixed signal separation algorithm to separate several overlapping spectral features of the iron shoes, and obtain several separated perturbation spectral feature sets.

[0140] Preferably, the PSO-ICA hybrid signal separation algorithm is a particle swarm optimization independent component analysis algorithm, combining the global search capability of particle swarm optimization (PSO) with the blind source separation capability of independent component analysis (ICA) to solve the problem of overlapping spectral features of multiple iron shoe magnetic fields. Let the perturbation spectral feature set be a linear mixture of multiple independent signals (formula X=AS, where X is the observed overlapping spectral feature matrix, A is the mixing matrix, and S is the source signal matrix to be separated).

[0141] Preferably, the calculation process is as follows:

[0142] ① Input preprocessing: Denote the set of all disturbance spectrum features output from step S3 as {T} disturb,1 ,T disturb,2 ,...,T disturb,M}, where M is the number of signal streams, stacked as a hybrid matrix X according to sensor channels. N×M (N represents the feature dimension, such as the main frequency, harmonic phase difference, and spectral amplitude, totaling 13 dimensions).

[0143] ② PSO optimization of the ICA separation matrix:

[0144] Particle initialization: Set the number of particles to 20 to balance search efficiency and computational cost. Each particle represents a candidate solution of the ICA separation matrix W, where W is an N×N matrix and its initial value is randomly generated following a standard normal distribution.

[0145] Fitness function: The objective is to maximize the non-Gaussianity of the separated signals, and the negative entropy J(y) is used as the fitness value: J(y) = [E{G(y)} − E{G(v)}] 2 y=Wx is the separation signal, v is a Gaussian variable, G(u)=tanh(au), and a=1.5 is an empirical coefficient, the larger the value, the better the separation effect.

[0146] Iterative update: The ICA separation matrix W is optimized using the PSO velocity-position update formula (learning factor c1=c2=2, inertia weight w=0.7 linearly decreasing) until the fitness value change is less than the threshold 0.0001 to reach convergence or the maximum number of iterations is reached after 50 iterations.

[0147] ③ Output separation result: Under the optimal ICA separation matrix W, the source signal matrix S=WX is separated. After sorting by signal energy, the first K components are taken (K is the estimated number of iron shoes), and the separation perturbation spectrum feature set {S1,S2,...,S...} is obtained. K}

[0148] Step S42: The tilt angle of each shoe is inverted by the spatial distribution characteristics of the three-axis components of the fluxgate sensor array. Attitude correction is performed on all the separated perturbation spectrum feature sets. An attitude-corrected spectrum feature set is obtained based on a separated perturbation spectrum feature set.

[0149] Preferably, the placement posture (tilt angle) of the iron shoe will cause a deviation in the direction of the magnetic field received by the fluxgate sensor. Therefore, let the magnetic field generated by the permanent magnet array in the ideal posture (perpendicular to the ground) be (Bx0, By0, Bz0) in the three-axis components of the sensor; in the actual posture, the tilt angle of the iron shoe around the X / Y / Z axes is (α, β, γ), and a small angle approximation is used. Then, the relationship between the measured three-axis components (Bx, By, Bz) and the ideal components is as follows: ,in, For rotation matrices, at small angles, it simplifies to a diagonal matrix with cross terms. After ignoring higher-order small quantities, α, β, and γ are solved using the least squares method with an accuracy of ±0.5°.

[0150] The next step is to analyze the separated perturbation spectrum feature set S. k (Including triaxial magnetic field component information), the tilt angle (α) is inverted according to the above relationship. k ,β k ,γ k To correct directional deviations in spectral characteristics: ,in, This is the spectral feature set after attitude correction.

[0151] Step S43: Group all the attitude-corrected spectral feature sets according to the uniqueness of the cluster center of each iron shoe to obtain the independent magnetic field signal spectral features of each iron shoe.

[0152] Preferably, the uniqueness grouping of cluster centers can be achieved using the K-means clustering algorithm, with the attitude-corrected spectral feature set as input. The feature dimensions include the main frequency component, harmonic phase difference, and spectral amplitude vector (10 dimensions in total from 1 to 10 Hz), for a total of 13 dimensions.

[0153] The number of clusters can be set to K, which is consistent with the estimated number of iron shoes in step S41. The clustering validity is verified by the silhouette coefficient, and a threshold greater than 0.6 is considered reasonable. The cluster center is calculated by iteratively updating the cluster center by the arithmetic mean of the features within each cluster until the sum of squared errors within the cluster is less than the threshold, for example, 0.001, to reach the convergence judgment.

[0154] It is worth noting that each cluster corresponds to a single iron shoe after clustering. The cluster center (cluster center) is unique. The Euclidean distance between different cluster centers is greater than twice the average intra-cluster distance to avoid misclassification. The attitude-corrected spectral feature sets within the same cluster are grouped together to obtain the independent magnetic field signal spectral features of each iron shoe. The cluster center features of each group are taken as representative, or the mean features within the group are retained.

[0155] Beneficial effects of steps S41 to S43:

[0156] By performing mixed signal separation, attitude correction, and grouping processing, the independent magnetic field spectrum features of each track shoe are extracted from the overlapping disturbance signals. Step S41 inputs the overlapping disturbance spectrum feature set of multiple track shoes into the PSO-ICA algorithm to separate the mutually interfering spectrum components, obtaining several separated disturbance spectrum feature sets, thus solving the problem of multi-target signal aliasing. Step S42 uses the spatial distribution characteristics of the three-axis components of the fluxgate sensor to invert the tilt angle of the track shoes, performs attitude correction on the separated feature sets, eliminates the magnetic field direction deviation caused by track shoe attitude changes, and obtains the attitude-corrected spectrum feature set. Step S43 groups the corrected feature sets according to the uniqueness of each track shoe cluster center, forming independent magnetic field signal spectrum features for each track shoe. The sub-steps of this section effectively remove the interference and attitude influence of multiple track shoe signals, providing clean and independent spectrum feature input for subsequent distance estimation and positioning, ensuring the identifiability and accuracy of track shoe magnetic field signals in dynamic shunting environments.

[0157] Step S5: Based on the three-axis components of each fluxgate sensor, calculate the set of nonlinear distances from each shoe to each fluxgate sensor based on the dipole model using a nonlinear distance estimation algorithm based on the dipole model.

[0158] Further, in step S5, based on the three-axis components of each fluxgate sensor, the nonlinear distance set from each shoe to each fluxgate sensor based on the dipole model is calculated using a nonlinear distance estimation algorithm based on the dipole model. Specifically, this includes the following steps:

[0159] Step S51: Obtain the triaxial component data of each fluxgate sensor.

[0160] Preferably, each sensor in the fluxgate sensor array is denoted as Sj, j=1,2,...,J, where J is the total number of sensors. The output consists of three orthogonal magnetic field components (Bx,j,By,j,Bz,j), in nanotesla. The sensor employs a 5pT sensitivity, DC-200Hz bandwidth fluxgate, and its three-axis components are acquired via a 16-bit ADC at a sampling rate of 1.2kHz, stored as time-series data, with a timestamp accuracy of ±1ms.

[0161] Preferably, the triaxial component data can be calibrated and preprocessed. Before acquisition, the sensor is calibrated for triaxial orthogonality using a Helmholtz coil to ensure that the orthogonality error is ≤0.5°, and temperature drift is compensated with a coefficient of -0.1nT / ℃ to ensure the accuracy of the triaxial component data.

[0162] Step S52: Spatially align the independent magnetic field signal spectrum characteristics of each iron shoe with the triaxial component data of each fluxgate sensor to match the correspondence between the transmit and receive signals of each iron shoe and each fluxgate sensor, and obtain the aligned spectrum characteristics and triaxial component pairing data.

[0163] Preferably, spatial alignment can transform the spectral characteristics of the independent magnetic field signal of the iron shoe into the global coordinate system of the sensor array, with the starting point of the track as the origin, the X-axis along the track direction, the Y-axis laterally, and the Z-axis perpendicular to the ground, and coordinate mapping is achieved through the magnetic field direction vector after attitude correction.

[0164] Preferably, the signal matching can be based on the shunting operation timeline provided by the dispatching device to provide the track of the shunting shoe. The time period when the shunting shoe passes the sensor (i.e., the duration of the magnetic field disturbance signal is about 0.5s) is aligned with the timestamp of the sensor's three-axis component data. The matching rule is that the center of the timestamp of the shunting shoe signal spectrum characteristics coincides with the time window of the sensor's three-axis component data ±0.25s.

[0165] Preferably, the pairing data can be obtained by combining the independent magnetic field spectrum characteristics of the iron shoe with the triaxial component data of the sensor according to the above alignment results, so as to obtain the aligned spectrum characteristics and triaxial component pairing data.

[0166] Step S53: Input the aligned spectral features and triaxial component pairing data into the dipole model nonlinear distance estimation algorithm. Based on the inverse cubic relationship between magnetic field strength and distance, and the directional sensitivity characteristics of fluxgate sensors, calculate the nonlinear distance from each iron shoe to each fluxgate sensor.

[0167] Preferably, the nonlinear distance estimation algorithm for the dipole model is based on the inverse relationship between the magnetic field strength of the magnetic dipole in the far-field region and the cube of the distance:

[0168] .

[0169] in, For non-linear distances, denoted as vacuum permeability, m as magnetic dipole moment, r as distance, and θ as the angle between the magnetic field direction and the dipole axis.

[0170] Specifically, the magnetic dipole moment is calibrated using the initial magnetic field signal under interference-free conditions in step S2, i.e., the magnetic dipole moment of the iron shoe permanent magnet array. Where r0 = 0.1m, B std The initial magnetic field strength is obtained in step S2.

[0171] Preferably, the directional sensitivity characteristics of the fluxgate sensor can be obtained through a triaxial component calibration experiment or by direct query, with the x, y, z axis sensitivity ratio being 1:1.02:0.98.

[0172] Step S54: The nonlinear distance from each iron shoe to each fluxgate sensor is structurally integrated according to the uniqueness of the cluster center of each iron shoe and the number of the fluxgate sensor, to obtain the set of nonlinear distances from each iron shoe to each fluxgate sensor.

[0173] Preferably, JSON format can also be used for structured integration.

[0174] Beneficial effects of steps S51 to S54:

[0175] By acquiring sensor data, aligning signals, calculating nonlinear distances, and integrating them in a structured manner, the precise measurement and organization of the nonlinear distances from each shoe to each fluxgate sensor are achieved. Step S51 acquires the triaxial component data of each fluxgate sensor, providing a basis for the sensor's directional sensitivity characteristics for distance estimation. Step S52 spatially aligns the independent magnetic field signal spectrum characteristics of the shoe with the triaxial component data of each sensor, matches the correspondence between transmitted and received signals, and obtains the aligned spectrum characteristics and triaxial component paired data to ensure the accuracy of the correlation between the signal and the sensor. Step S53 inputs the paired data into the dipole model nonlinear distance estimation algorithm. Based on the inverse relationship between magnetic field strength and the cube of distance and the sensor's directional sensitivity characteristics, the nonlinear distance from each shoe to each fluxgate sensor is calculated, characterizing the real spatial correlation under magnetic field attenuation. Step S54 integrates the nonlinear distances from each shoe to each fluxgate sensor in a structured manner according to the uniqueness of the shoe cluster center and the fluxgate sensor number, to obtain the set of nonlinear distances from each shoe to each fluxgate sensor. The sub-steps of this section provide reliable raw distance data for the subsequent construction of distance equations to solve for position coordinates through precise data association, physical model calculation and orderly integration, supporting the quantitative expression of spatial relationships in passive track shoe positioning.

[0176] Step S6: Obtain the nonlinear distance from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them using the least squares method to obtain the position coordinates of the current iron shoe.

[0177] Further, in step S6, the nonlinear distances from the current iron shoe to at least three non-collinear fluxgate sensors are obtained, and a set of nonlinear distance equations is constructed. The position coordinates of the current iron shoe are then obtained by solving the least squares method. Specifically, this includes the following steps:

[0178] Step S61: Obtain the pre-calibrated position coordinates of each fluxgate sensor.

[0179] Preferably, the origin O can be taken as the starting point of the railway track, with the X-axis along the longitudinal direction of the track (in the direction of increasing mileage), the Y-axis along the transverse direction of the track (in the direction of track gauge, pointing to the outside of the line), and the Z-axis perpendicular to the ground and upward, forming a right-hand rectangular coordinate system.

[0180] Preferably, a total station can be used to measure the installation position of each fluxgate sensor in the field and record the three-dimensional coordinates. The Z-axis direction represents the installation height of the rail web, which can be uniformly set to 1.2m.

[0181] Step S62: Select the nonlinear distances from the current shoe to at least three non-collinear fluxgate sensors from the nonlinear distance set as the nonlinear distance subset of the current shoe, and use the pre-calibrated position coordinates corresponding to the selected fluxgate sensors as the sensor position coordinate subset of the current shoe.

[0182] Preferably, the three non-collinear fluxgate sensors must satisfy the condition that the cross product of the vectors is not zero. At least three non-collinear sensors should be selected to ensure computational feasibility, and four to five sensors can be redundantly selected. Sensors with a distance measurement error ≤0.1m should be selected first.

[0183] Step S63: Construct a set of nonlinear distance equations for the nonlinear distance subset of the current iron shoe and the sensor position coordinate subset based on the dipole model distance formula.

[0184] Preferably, a system of equations is constructed based on the dipole model distance formula to describe the relationship between the position of the track shoe (x, y, z) and the sensor coordinates (x, y, z). j ,y j ,z j Spatial relationship. In the dipole model, the distance r from the magnetic dipole to the sensor is... i,j satisfy:

[0185] .

[0186] in, The measured magnetic field strength is for the sensor.

[0187] Preferably, in actual positioning, minute changes in the magnetic field direction can be ignored to simplify the above equation into a geometric distance equation. .

[0188] Next, squaring both sides of the above equation to eliminate the radical sign yields at least three nonlinear equations:

[0189] .

[0190] The next step is to output the nonlinear distance equation system F(x,y,z)=0, where F is the equation vector.

[0191] Step S64: Solve the nonlinear distance equations of the current iron shoe using the least squares method to obtain the current position coordinates of the iron shoe.

[0192] Preferably, the Gauss-Newton method can be used to solve the nonlinear least squares problem, with the objective of minimizing the sum of squared residuals. The calculation process is as follows:

[0193] ① Initial value setting: The geometric center of the subset of sensor position coordinates is used as the initial guess.

[0194] ② Jacobian matrix construction: Calculate the partial derivative matrix J of the residual function with respect to the unknowns.

[0195] ③ Iterative update: using the formula Δq=(J T J) −1 J T F updates the position vector q=(x,y,z), iterating until the residual change ||Δq|| is less than the convergence threshold of 0.0001 m or until the maximum number of iterations is 20.

[0196] ④ Output: After convergence, the position coordinates (xi,yi,zi) of the current iron shoe i are obtained, with an accuracy of ±0.05m.

[0197] Beneficial effects of steps S61 to S64:

[0198] The system achieves accurate calculation of the current shoe's coordinates through sensor location acquisition, effective distance filtering, equation construction, and solution. Step S61 acquires the pre-calibrated coordinates of each fluxgate sensor, providing a reference for spatial positioning. Step S62 selects the nonlinear distances from the current shoe to at least three non-collinear sensors from the nonlinear distance set as a subset, and simultaneously acquires the corresponding pre-calibrated coordinates of the sensors as a subset to ensure the effectiveness of distance-position matching. Step S63 combines the nonlinear distance subset with the sensor position coordinate subset based on the dipole model distance formula to construct a set of nonlinear distance equations describing the spatial relationship between the shoe and the sensors. Step S64 solves this set of equations using the least squares method, iteratively optimizing the minimum sum of squared residuals to obtain the current shoe's coordinates. This sub-step transforms the sensor-measured nonlinear distances into spatial coordinates, and through multi-sensor geometric constraints and nonlinear model solving, provides accurate spatial location data for shoe positioning, supporting the subsequent integration of identity and location information to form a complete positioning result.

[0199] Step S7: Compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and positioning information.

[0200] Further, in step S7, the magnetic field signal spectrum features of the current iron shoe are compared with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and the position coordinates of the current iron shoe and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity are integrated into the identity and positioning information of the current iron shoe. Specifically, this includes the following steps:

[0201] Step S71: Calculate the Euclidean distance between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes in the feature template library to obtain a set of Euclidean distance values ​​between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes.

[0202] Preferably, Euclidean distance is used to measure the similarity between the current independent magnetic field spectrum features of the iron shoe and the standard spectrum features in the feature template library. The smaller the distance, the higher the similarity. Euclidean distance is the straight-line distance between two points in vector space and is suitable for quantifying the overall differences of multidimensional features.

[0203] Step S72: Sort all elements in the Euclidean distance value set in ascending order, and obtain the element with the smallest Euclidean distance value, as well as the magnetic coding sequence corresponding to the element with the smallest distance value in the feature template library, which is the magnetic coding sequence corresponding to the current iron shoe.

[0204] Preferably, the iron shoe whose spectral features are most similar to the iron shoe in the feature template library corresponding to the entry with the smallest Euclidean distance is the magnetic coding sequence of the current iron shoe.

[0205] Step S73: The current location coordinates of the iron shoe are structurally integrated with the magnetic code sequence corresponding to the current iron shoe to obtain the identity and location information of the current iron shoe.

[0206] Preferably, JSON format can also be used for structured integration, for example:

[0207] {

[0208] "shoe_id": "101011", / / Magnetic code sequence (S72 output)

[0209] "position": [1234.56, 2.30, 1.20], / / Position coordinates (x, y, z) (output in step S6)

[0210] "timestamp": "XXXXYYZZTAABBCCZ", / / Location timestamp (UTC format)

[0211] "confidence": 0.98 / / Confidence (1 - d_{(1)} / d_{max}, where d_{max} is the maximum distance in the template library)

[0212] }

[0213] Beneficial effects of steps S71 to S73:

[0214] By comparing Euclidean distances, filtering similarities, and integrating information, the system achieves precise association between the current iron shoe's identity and location information. Step S71 calculates the Euclidean distance between the current iron shoe's independent magnetic field signal spectrum features and all standard iron shoe spectrum features in the feature template library, generating a set of Euclidean distance values ​​representing the similarity between the two, providing a quantitative basis for identity matching. Step S72 sorts this set in ascending order, selects the element with the smallest distance, and obtains its corresponding magnetic coding sequence in the feature template library, thereby confirming the unique identity of the current iron shoe. Step S73 structurally integrates the current iron shoe's position coordinates obtained in step S6 with the aforementioned magnetic coding sequence to form location information containing identity identifiers and spatial locations. The sub-steps of this module ensure precise matching between dynamically acquired iron shoe signals and preset identity benchmarks through rigorous similarity measurement and data association, while reliably binding position coordinates with identity information, ultimately outputting complete passive iron shoe positioning results to support real-time monitoring of iron shoe status during shunting operations.

[0215] Overall beneficial effects of steps S1 to S7:

[0216] It achieves collaborative identification and location positioning for passive iron shoes. The process involves several steps: Step S1 assigns a unique binary magnetic code sequence to each wheel shoe as its identification; Step S2 collects interference-free initial magnetic field signals and extracts standard spectral features, integrating the magnetic code and spectral features to construct a feature template library, establishing a benchmark for identification matching; Step S3 continuously collects local magnetic field disturbance signals and extracts spectral features during shunting to capture magnetic field changes in dynamic environments; Step S4 uses a hybrid signal separation algorithm to separate overlapping spectral features of multiple wheel shoes, combines the inversion of tilt angles using the three-axis components of the fluxgate sensor for attitude correction, and obtains the independent magnetic field signal spectral features of each wheel shoe; Step S5 uses a dipole model nonlinear distance estimation algorithm to calculate the nonlinear distance set from each wheel shoe to each fluxgate sensor using the three-axis components of the sensor, characterizing the nonlinear correlation between magnetic field strength and distance; Step S6 selects at least three non-collinear sensor distances to construct a nonlinear equation set, and solves it using the least squares method to obtain the wheel shoe's position coordinates; Step S7 compares the current wheel shoe's independent spectral features with the feature template library, filters the standard spectral features with the highest similarity using Euclidean distance, confirms the corresponding magnetic code sequence, and finally integrates the position coordinates and magnetic code sequence to form positioning information. The passive iron shoe achieves unique identification and precise location positioning. Furthermore, the replacement or magnetization frequency of the permanent magnet is significantly lower than the replacement or charging frequency of the power supply, enabling the iron shoe to adapt to the complex environment of railway shunting and thus greatly reducing maintenance costs.

[0217] like Figure 2 As shown, this embodiment provides an example of a positioning device for iron shoes. In this embodiment, the positioning device is applied to the positioning method as described in the above embodiment.

[0218] Specifically, the positioning device includes a magnetic code sequence conversion module 1 for iron shoes, a feature template library acquisition module 2, a disturbance signal spectrum feature acquisition module 3, an independent magnetic field signal spectrum feature separation module 4, a nonlinear distance set calculation module 5, an iron shoe position coordinate solution module 6, and an iron shoe identity positioning information acquisition module 7, which are electrically or signalally connected in sequence.

[0219] The system comprises the following modules: 1) a magnetic coding sequence conversion module for each permanent magnet array, which converts each array into a binary coding sequence and uses it as the magnetic coding sequence for each shoe; 2) a feature template library acquisition module, which acquires the initial magnetic field signal of each shoe under interference-free conditions and extracts the standard spectral features of each initial magnetic field signal using Fast Fourier Transform, integrating the magnetic coding sequences and standard spectral features of all shoes to obtain a feature template library; 3) a disturbance signal spectral feature acquisition module, which continuously collects all local magnetic field disturbance signals through a fluxgate sensor array during railway shunting operations and extracts the spectral features of all local magnetic field disturbance signals using Fast Fourier Transform; and 4) an independent magnetic field signal spectral feature separation module, which separates the spectral features of all local magnetic field disturbance signals using a hybrid signal separation algorithm and assigns each shoe a spectral feature. After attitude correction, the independent magnetic field signal spectrum characteristics of each shoe are obtained; the nonlinear distance set calculation module 5 is used to calculate the nonlinear distance set from each shoe to each magnetic fluxgate sensor based on the dipole model nonlinear distance estimation algorithm using the three-axis components of each fluxgate sensor; the shoe position coordinate solving module 6 is used to obtain the nonlinear distance from the current shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them by the least squares method to obtain the position coordinates of the current shoe; the shoe identity positioning information acquisition module 7 is used to compare the magnetic field signal spectrum characteristics of the current shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the position coordinates of the current shoe and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the identity positioning information of the current shoe.

[0220] Furthermore, the feature template library acquisition module 2 specifically includes a first feature template library acquisition unit, a second feature template library acquisition unit, a third feature template library acquisition unit, a fourth feature template library acquisition unit, a fifth feature template library acquisition unit, a sixth feature template library acquisition unit, and a seventh feature template library acquisition unit that are electrically or signal-connected in sequence; the first feature template library acquisition unit is electrically or signal-connected to the iron shoe magnetic code sequence conversion module 1, and the seventh feature template library acquisition unit is electrically or signal-connected to the disturbance signal spectrum feature acquisition module 3.

[0221] The first feature template library acquisition unit is used to deploy a fluxgate calibration array composed of three-axis coils in an interference-free environment and to form a zero-magnetic environment acquisition condition through magnetic shielding. The second feature template library acquisition unit is used to place the permanent magnet array of each shoe in the same orientation directly above the fluxgate calibration array, and to acquire the initial magnetic field time-domain signal of each permanent magnet array through the fluxgate calibration array. The third feature template library acquisition unit is used to preprocess each initial magnetic field time-domain signal through bandpass filtering and drift removal preprocessing to obtain a clean magnetic field time-domain signal based on an initial magnetic field time-domain signal. The fourth feature template library acquisition unit is used to calculate a standard frequency domain signal for each pure magnetic field time domain signal through Fast Fourier Transform; the fifth feature template library acquisition unit is used to extract the main frequency component, harmonic phase difference and spectral amplitude of each standard frequency domain signal, and obtain a standard spectral feature set based on a frequency domain signal; the sixth feature template library acquisition unit is used to structurally integrate the magnetic coding sequence of each iron shoe with the corresponding standard spectral feature set, and obtain a standard magnetic coding feature set based on an iron shoe; the seventh feature template library acquisition unit is used to integrate the standard magnetic coding feature sets of all iron shoes to obtain the feature template library.

[0222] Furthermore, the disturbance signal spectrum feature acquisition module 3 specifically includes a first signal spectrum feature acquisition unit, a second signal spectrum feature acquisition unit, a third signal spectrum feature acquisition unit, and a fourth signal spectrum feature acquisition unit that are electrically or signal-connected in sequence; the first signal spectrum feature acquisition unit is electrically or signal-connected to the seventh feature template library acquisition unit, and the fourth signal spectrum feature acquisition unit is electrically or signal-connected to the independent magnetic field signal spectrum feature separation module 4.

[0223] The system comprises the following components: a first signal spectrum feature acquisition unit, which activates the fluxgate sensor array during railway shunting operations to continuously monitor the magnetic field changes as all trackers pass by, thereby obtaining several original local magnetic field disturbance signal streams; a second signal spectrum feature acquisition unit, which samples each original local magnetic field disturbance signal stream in real time, and obtains a discrete local magnetic field disturbance time-domain signal sequence based on the original local magnetic field disturbance signal stream; a third signal spectrum feature acquisition unit, which performs a fast Fourier transform on each discrete local magnetic field disturbance time-domain signal sequence to obtain a disturbance frequency-domain signal; and a fourth signal spectrum feature acquisition unit, which extracts the dominant frequency component, harmonic phase difference, and spectral amplitude of each disturbance frequency-domain signal, and obtains a disturbance spectrum feature set based on the disturbance frequency-domain signal.

[0224] Furthermore, the independent magnetic field signal spectrum feature separation module 4 specifically includes a first independent magnetic field signal spectrum feature separation unit, a second independent magnetic field signal spectrum feature separation unit, and a third independent magnetic field signal spectrum feature separation unit that are electrically or signal-connected in sequence; the first independent magnetic field signal spectrum feature separation unit is electrically or signal-connected to the fourth signal spectrum feature acquisition unit, and the third independent magnetic field signal spectrum feature separation unit is electrically or signal-connected to the nonlinear distance set calculation module 5.

[0225] The first independent magnetic field signal spectrum feature separation unit is used to input all disturbance spectrum feature sets into the PSO-ICA hybrid signal separation algorithm to separate the overlapping spectrum features of several iron shoes and obtain several separated disturbance spectrum feature sets. The second independent magnetic field signal spectrum feature separation unit is used to invert the tilt angle of each iron shoe through the three-axis component spatial distribution characteristics of the fluxgate sensor array, perform attitude correction on all separated disturbance spectrum feature sets, and obtain an attitude-corrected spectrum feature set based on a separated disturbance spectrum feature set. The third independent magnetic field signal spectrum feature separation unit is used to group all attitude-corrected spectrum feature sets according to the uniqueness of the cluster center of each iron shoe to obtain the independent magnetic field signal spectrum features of each iron shoe.

[0226] Furthermore, the nonlinear distance set calculation module 5 specifically includes a first nonlinear distance set calculation unit, a second nonlinear distance set calculation unit, a third nonlinear distance set calculation unit, and a fourth nonlinear distance set calculation unit that are electrically or signal-connected in sequence; the first nonlinear distance set calculation unit is electrically or signal-connected to the third independent magnetic field signal spectrum feature separation unit, and the fourth nonlinear distance set calculation unit is electrically or signal-connected to the iron shoe position coordinate solution module 6.

[0227] The system comprises the following components: a first nonlinear distance set calculation unit, which acquires the triaxial component data of each fluxgate sensor; a second nonlinear distance set calculation unit, which spatially aligns the independent magnetic field signal spectrum characteristics of each shoe with the triaxial component data of each fluxgate sensor to match the transmit / receive signal correspondence between each shoe and each fluxgate sensor, thus obtaining aligned spectrum characteristics and triaxial component pairing data; a third nonlinear distance set calculation unit, which inputs the aligned spectrum characteristics and triaxial component pairing data into a dipole model nonlinear distance estimation algorithm, and calculates the nonlinear distance from each shoe to each fluxgate sensor based on the inverse cubic relationship between magnetic field strength and distance, and the directional sensitivity characteristics of fluxgate sensors; and a fourth nonlinear distance set calculation unit, which structurally integrates the nonlinear distances from each shoe to each fluxgate sensor according to the uniqueness of the cluster center of each shoe and the number of the fluxgate sensor, thus obtaining a set of nonlinear distances from each shoe to each fluxgate sensor.

[0228] Furthermore, the iron shoe position coordinate solving module 6 specifically includes a first iron shoe position coordinate solving unit, a second iron shoe position coordinate solving unit, a third iron shoe position coordinate solving unit, and a fourth iron shoe position coordinate solving unit that are electrically or signal-connected in sequence; the first iron shoe position coordinate solving unit is electrically connected to the fourth nonlinear distance set calculation unit and then signal-connected, and the fourth iron shoe position coordinate solving unit is electrically or signal-connected to the iron shoe identity positioning information acquisition module 7.

[0229] The system comprises the following components: a first iron shoe position coordinate solving unit, which obtains the pre-calibrated position coordinates of each fluxgate sensor; a second iron shoe position coordinate solving unit, which selects the nonlinear distances from the current iron shoe to at least three non-collinear fluxgate sensors from the nonlinear distance set as a subset of the nonlinear distances of the current iron shoe, and uses the pre-calibrated position coordinates of the selected fluxgate sensors as a subset of the sensor position coordinates of the current iron shoe; a third iron shoe position coordinate solving unit, which constructs a set of nonlinear distance equations between the subset of nonlinear distances of the current iron shoe and the subset of sensor position coordinates based on the dipole model distance formula; and a fourth iron shoe position coordinate solving unit, which solves the set of nonlinear distance equations of the current iron shoe using the least squares method to obtain the position coordinates of the current iron shoe.

[0230] Furthermore, the "Iron Shoe Identity Location Information Acquisition Module 7" specifically includes a first "Iron Shoe Identity Location Information Acquisition Unit", a second "Iron Shoe Identity Location Information Acquisition Unit", and a third "Iron Shoe Identity Location Information Acquisition Unit" that are electrically or signal-connected in sequence; the first "Iron Shoe Identity Location Information Acquisition Unit" is electrically or signal-connected to the fourth "Iron Shoe Position Coordinate Solving Unit".

[0231] The first iron shoe identity and positioning information acquisition unit is used to calculate the Euclidean distance between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes in the feature template library, so as to obtain a set of Euclidean distance values ​​between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes; the second iron shoe identity and positioning information acquisition unit is used to sort all elements in the Euclidean distance value set in ascending order, and obtain the element with the smallest Euclidean distance value, as well as the magnetic coding sequence corresponding to the element with the smallest distance value in the feature template library, which is the magnetic coding sequence corresponding to the current iron shoe; the third iron shoe identity and positioning information acquisition unit is used to structurally integrate the position coordinates of the current iron shoe with the magnetic coding sequence corresponding to the current iron shoe to obtain the identity and positioning information of the current iron shoe.

[0232] It should be noted that this embodiment is a functional module embodiment based on the above method embodiment. For additional content such as extensions, optimizations, limitations, examples, principle explanations, and beneficial effects of this embodiment, please refer to the above embodiments. This embodiment will not repeat them here.

[0233] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Figure 3 As shown, the electronic device 8 includes a processor 81 and a memory 82 coupled to the processor 81.

[0234] The memory 82 stores program instructions for implementing the federated learning-based collaborative energy-saving method for government data clusters in any of the above embodiments.

[0235] The processor 81 is used to execute program instructions stored in the memory 82 for collaborative energy saving of government data clusters based on federated learning.

[0236] The processor 81 can also be referred to as a CPU (Central Processing Unit). The processor 81 may be an integrated circuit chip with signal processing capabilities. The processor 81 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0237] Furthermore, Figure 4 This is a schematic diagram of the structure of a storage medium according to an embodiment of this application. See also: Figure 4 In this embodiment of the application, the storage medium 9 stores program instructions 91 capable of implementing all the above methods. These program instructions 91 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0238] In the several embodiments provided in this application, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, signal, or other forms.

[0239] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for positioning iron shoes, wherein there are several iron shoes, all used in locomotives and rolling stock in railway shunting operations, each iron shoe is embedded with an array of permanent magnets with non-overlapping magnetic poles, and a fluxgate sensor array is laid on both sides of the railway track, characterized in that... The positioning method includes: Step S1: Convert each permanent magnet array into a binary encoding sequence and use it as the magnetic encoding sequence for each iron shoe. Step S2: Obtain the initial magnetic field signal of each iron shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal through fast Fourier transform. Integrate the magnetic coding sequence of all iron shoes and the standard spectral features to obtain a feature template library. Step S3: During the railway shunting operation, all local magnetic field disturbance signals are continuously collected through the fluxgate sensor array, and the spectral characteristics of all local magnetic field disturbance signals are extracted by fast Fourier transform. Step S4: Separate the spectral characteristics of all local magnetic field disturbance signals using a hybrid signal separation algorithm, and after correcting the attitude of each iron shoe, obtain the independent magnetic field signal spectral characteristics of each iron shoe. Step S5: Based on the three-axis components of each fluxgate sensor, calculate the set of nonlinear distances from each shoe to each fluxgate sensor based on the dipole model using a nonlinear distance estimation algorithm based on the dipole model. Step S6: Obtain the nonlinear distance from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them by the least squares method to obtain the position coordinates of the current iron shoe; Step S7: Compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and positioning information; Step S3: During the railway shunting operation, all local magnetic field disturbance signals are continuously collected through the fluxgate sensor array, and the spectral characteristics of all local magnetic field disturbance signals are extracted by fast Fourier transform, including: Step S31: During the railway shunting operation, the fluxgate sensor array is activated to continuously monitor the magnetic field changes when all the iron shoes pass by, and obtain several original local magnetic field disturbance signal streams. Step S32: Sample each original local magnetic field disturbance signal stream in real time, and obtain a discrete local magnetic field disturbance time domain signal sequence based on an original local magnetic field disturbance signal stream; Step S33: Obtain a perturbation frequency domain signal by performing a fast Fourier transform on each discrete local magnetic field perturbation time domain signal sequence; Step S34: Extract the main frequency component, harmonic phase difference and spectral amplitude of each perturbation frequency domain signal, and obtain a perturbation spectral feature set based on a perturbation frequency domain signal; Step S4: Separate the spectral characteristics of all local magnetic field disturbance signals using a hybrid signal separation algorithm, and after correcting the attitude of each shoe, obtain the independent magnetic field signal spectral characteristics of each shoe, including: Step S41: Input all perturbation spectral feature sets into the PSO-ICA mixed signal separation algorithm to separate several overlapping spectral features of the iron shoes, and obtain several separated perturbation spectral feature sets. Step S42: The tilt angle of each shoe is inverted by the spatial distribution characteristics of the three-axis components of the fluxgate sensor array. Attitude correction is performed on all the separated perturbation spectrum feature sets. An attitude-corrected spectrum feature set is obtained based on a separated perturbation spectrum feature set. Step S43: Group all the attitude-corrected spectral feature sets according to the uniqueness of the cluster center of each iron shoe to obtain the independent magnetic field signal spectral features of each iron shoe.

2. The positioning method according to claim 1, characterized in that, Step S2: Obtain the initial magnetic field signal of each metal shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal using Fast Fourier Transform. Integrate the magnetic coding sequences and standard spectral features of all metal shoes to obtain a feature template library, including: Step S21: Deploy a fluxgate calibration array composed of three-axis coils in an interference-free environment, and form a zero-magnetic environment acquisition condition through magnetic shielding. Step S22: Place the permanent magnet array of each iron shoe in the same orientation directly above the fluxgate calibration array, and collect the initial magnetic field time domain signal of each permanent magnet array through the fluxgate calibration array. Step S23: Each initial magnetic field time domain signal is preprocessed by bandpass filtering and drift removal preprocessing respectively, and a clean magnetic field time domain signal is obtained based on an initial magnetic field time domain signal. Step S24: Calculate a standard frequency domain signal for each pure magnetic field time domain signal using Fast Fourier Transform; Step S25: Extract the main frequency component, harmonic phase difference and spectral amplitude of each standard frequency domain signal to obtain a standard spectral feature set based on a frequency domain signal; Step S26: The magnetic coding sequence of each iron shoe is structurally integrated with the corresponding standard spectral feature set to obtain a standard magnetic coding feature set based on an iron shoe. Step S27: Integrate all standard magnetic coding feature sets of iron shoes to obtain the feature template library.

3. The positioning method according to claim 1, characterized in that, Step S5: Based on the triaxial components of each fluxgate sensor, calculate the set of nonlinear distances from each shoe to each fluxgate sensor using a dipole model nonlinear distance estimation algorithm, including: Step S51: Acquire the triaxial component data of each fluxgate sensor; Step S52: Spatially align the independent magnetic field signal spectrum characteristics of each iron shoe with the triaxial component data of each fluxgate sensor to match the correspondence between the transmit and receive signals of each iron shoe and each fluxgate sensor, and obtain the aligned spectrum characteristics and triaxial component pairing data. Step S53: Input the aligned spectral features and triaxial component pairing data into the nonlinear distance estimation algorithm of the dipole model. Based on the inverse cubic relationship between magnetic field strength and distance, and the directional sensitivity characteristics of the fluxgate sensor, calculate the nonlinear distance from each iron shoe to each fluxgate sensor. Step S54: The nonlinear distance from each iron shoe to each fluxgate sensor is structurally integrated according to the uniqueness of the cluster center of each iron shoe and the number of the fluxgate sensor, to obtain the set of nonlinear distances from each iron shoe to each fluxgate sensor.

4. The positioning method according to claim 1, characterized in that, Step S6: Obtain the nonlinear distances from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them using the least squares method to obtain the current position coordinates of the iron shoe, including: Step S61: Obtain the pre-calibrated position coordinates of each fluxgate sensor; Step S62: Select the nonlinear distances from the current iron shoe to at least three non-collinear fluxgate sensors from the nonlinear distance set as the nonlinear distance subset of the current iron shoe, and use the pre-calibrated position coordinates corresponding to the selected fluxgate sensors as the sensor position coordinate subset of the current iron shoe. Step S63: Construct a set of nonlinear distance equations for the nonlinear distance subset of the current iron shoe and the sensor position coordinate subset based on the dipole model distance formula; Step S64: Solve the nonlinear distance equations of the current iron shoe using the least squares method to obtain the current position coordinates of the iron shoe.

5. The positioning method according to claim 1, characterized in that, Step S7: Compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and positioning information, including: Step S71: Calculate the Euclidean distance between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes in the feature template library to obtain the set of Euclidean distance values ​​between the independent magnetic field signal spectrum features of the current iron shoe and the standard spectrum features of all iron shoes. Step S72: Sort all elements in the Euclidean distance value set in ascending order, and obtain the element with the smallest Euclidean distance value, and the magnetic coding sequence corresponding to the element with the smallest distance value in the feature template library, which is the magnetic coding sequence corresponding to the current iron shoe. Step S73: The current location coordinates of the iron shoe are structurally integrated with the magnetic code sequence corresponding to the current iron shoe to obtain the identity and location information of the current iron shoe.

6. A positioning device for a metal shoe, said positioning device being applied to the positioning method as described in any one of claims 1 to 5, characterized in that, The positioning device includes: The magnetic coding sequence conversion module for iron shoes is used to convert each permanent magnet array into a binary coding sequence and use it as the magnetic coding sequence for each iron shoe. The feature template library acquisition module is used to acquire the initial magnetic field signal of each iron shoe in an interference-free environment, and extract the standard spectral features of each initial magnetic field signal through fast Fourier transform. The magnetic coding sequence of all iron shoes and the standard spectral features are integrated to obtain the feature template library. The disturbance signal spectrum feature acquisition module is used to continuously collect all local magnetic field disturbance signals through the fluxgate sensor array during the railway shunting operation, and extract the spectrum features of all local magnetic field disturbance signals through fast Fourier transform. The independent magnetic field signal spectrum feature separation module is used to separate the spectrum features of all local magnetic field disturbance signals through a hybrid signal separation algorithm, and after correcting the attitude of each iron shoe, obtain the independent magnetic field signal spectrum features of each iron shoe. The nonlinear distance set calculation module is used to calculate the nonlinear distance set from each shoe to each fluxgate sensor based on the dipole model using a nonlinear distance estimation algorithm based on the three-axis components of each fluxgate sensor. The iron shoe position coordinate solution module is used to obtain the nonlinear distance from the current iron shoe to at least three non-collinear fluxgate sensors, construct a set of nonlinear distance equations, and solve them by the least squares method to obtain the position coordinates of the current iron shoe. The "Iron Shoe Identity and Location Information Acquisition Module" is used to compare the magnetic field signal spectrum features of the current iron shoe with the feature template library to obtain the standard spectrum feature with the highest feature similarity, and integrate the current iron shoe's position coordinates and the magnetic coding sequence corresponding to the standard spectrum feature with the highest similarity into the current iron shoe's identity and location information.

7. An electronic device, characterized in that, The method includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the positioning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, enable the positioning method as described in any one of claims 1 to 5.

Images