High-precision zero-flux compensation method

Abstract

The invention discloses a high-precision zero magnetic flux compensation method, which comprises the following steps of: acquiring a detected current signal, and generating a magnetic field detection interface structure after noise reduction, frequency band extraction and magnetic sensitive configuration processing; performing magnetic field analysis, feature matching and resource optimization based on the data to generate efficient compensation parameters; final compensation output is generated through interference identification, compensation recombination and power adaptation; generating a calibration trigger mark through standard value comparison, stability evaluation and threshold judgment; and finally, forming a closed-loop calibration instruction set through type identification, strength calculation and time sequence arrangement, and completing parameter updating and verification. According to the invention, the precision and stability of current measurement are significantly improved, the interference of a complex environment is effectively suppressed, and the characteristics of strong adaptability and low power consumption are achieved.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic measurement technology, and in particular to a high-precision zero-flux compensation method. Background Technology

[0002] In the field of electromagnetic measurement technology, existing solutions related to a high-precision zero flux compensation method are usually based on fixed thresholds or static models to generate compensation parameters, which have limitations such as dynamic response hysteresis, cumulative deviation of multi-ring magnetic sensitive elements, and insufficient ability to suppress magnetic field disturbances.

[0003] Existing methods mostly rely on single-point magnetic field calibration or linear compensation mechanisms. Under the real-time and stability requirements of current measurement scenarios, these methods are prone to compensation lag effects and multi-loop signal coupling interference, making it difficult to achieve stable high-precision zero-flux compensation. Regarding the joint processing of magnetic field inhomogeneity parameters and the output of multi-loop magnetic sensing elements, existing technologies generally lack closed-loop control mechanisms for dynamic feature extraction and adaptive algorithm matching. This makes it difficult to form a complete process of magnetic field parameter acquisition—feature extraction—compensation coefficient generation—power adaptation output under complex operating conditions, resulting in current measurement accuracy being affected by both environmental magnetic field fluctuations and the nonlinear characteristics of the components. Summary of the Invention

[0004] This invention provides a high-precision zero-flux compensation method to address the problem of how to improve current measurement accuracy by integrating a multi-ring magnetic sensing element with an adaptive algorithm based on the magnetic field non-uniformity parameter, while meeting the real-time and stability requirements of zero-flux compensation.

[0005] To address the aforementioned technical problems, this invention provides a high-precision zero-flux compensation method, comprising:

[0006] The process involves acquiring the measured current signal, performing signal noise reduction (including multi-level filtering mechanisms and digital filtering algorithms), frequency band extraction and amplification using bandpass filters and programmable gain amplifiers, and multi-ring magnetic sensor configuration (including the layout design of multiple concentric ring magnetic sensing elements) to generate a magnetic field detection interface structure.

[0007] Based on magnetic field detection interface data, the magnetic field distribution is analyzed using spatial filtering and multivariate statistical analysis methods, feature matching including principal component analysis and dynamic weight adjustment mechanism, and resource optimization processing including compression coding and dynamic adjustment of power consumption model to generate efficient compensation parameters.

[0008] The system receives efficient compensation parameters, performs interference identification using spectrum analysis algorithms and time-domain feature extraction techniques, and employs a weighted fusion method and multi-objective optimization algorithm for compensation and recombination. Power adaptation processing is achieved through a digital power amplifier to generate the final compensated output structure.

[0009] The final compensation output data is obtained, and standard value comparison is performed using a joint time-domain and frequency-domain analysis method. Stability assessment includes calculating error volatility and multivariate regression models, as well as threshold judgment processing. Logical judgment and fuzzy reasoning techniques are used to generate calibration trigger flags.

[0010] The calibration trigger flag is parsed, and the type is identified, including zero-point calibration, gain calibration and nonlinear calibration. The intensity calculation adopts a weighted regression model, and the timing arrangement process adopts a priority-based scheduling algorithm to generate a closed-loop calibration instruction set.

[0011] The system receives the closed-loop calibration instruction set, performs aging coefficient analysis using multivariate regression analysis and machine learning algorithms, updates parameters using a double buffer mechanism, and performs verification write-back processing, including parameter integrity verification and online verification, and generates a closed-loop calibration completion flag.

[0012] Furthermore, the process of generating the magnetic field detection interface structure also includes:

[0013] The measured current signal is acquired, and signal noise reduction processing is performed, including low-pass and band-stop filters in a multi-level filtering mechanism and smoothing processing in a digital filtering algorithm, to obtain a pre-processed current signal.

[0014] The effective frequency band is extracted from the preprocessed current signal. The filter coefficients are optimized using the digital finite impulse response filter structure and window function method in the bandpass filter. The signal is amplified and the standard input signal is generated by adaptive gain adjustment in the programmable gain amplifier.

[0015] The standard input signal is configured with multi-ring magnetic sensitive elements, including the spatial gradient sensing layout design of multiple concentric ring magnetic sensitive sensing elements and synchronous sampling technology, to generate a magnetic field detection interface structure.

[0016] Furthermore, the process of generating efficient compensation parameters also includes:

[0017] The magnetic field detection interface data is acquired, and the multi-ring magnetic field distribution is analyzed by using the linear weighting algorithm in spatial filtering and the covariance matrix eigenvalue decomposition in multivariate statistical analysis to obtain the magnetic field inhomogeneity parameters.

[0018] Key feature values ​​were extracted from the magnetic field inhomogeneity parameters and calculated using the improved Shannon entropy weighting and dynamic correlation index in the principal component analysis method. Adaptive algorithm matching was performed, including historical compensation data feedback in the dynamic weight adjustment mechanism, to generate preliminary compensation coefficients.

[0019] Resource consumption optimization of the initial compensation coefficients includes a sparse representation-based compression algorithm in compression coding and recurrent neural network prediction in dynamic adjustment of the power consumption model, generating efficient compensation parameters.

[0020] Furthermore, the process of generating the final compensated output structure also includes:

[0021] The system receives efficient compensation parameters and performs external interference feature identification by employing Fast Fourier Transform (FFT) technology from spectrum analysis algorithms and Short-Time Energy Calculation and Transient Detection (STM) algorithms from time-domain feature extraction techniques to generate dynamic adjustment factors.

[0022] Priority parameters are extracted from the dynamic adjustment factor using a weighted summation method in a multi-dimensional evaluation algorithm. The compensation current recombination is performed by combining dynamic weight adjustment in a weighted fusion method with a combination of genetic algorithm and particle swarm optimization algorithm in a multi-objective optimization algorithm to generate a real-time compensation signal.

[0023] The real-time compensation signal is processed for power adaptation through the digital control loop, current sensing feedback and temperature compensation module in the digital power amplifier to generate the final compensation output structure.

[0024] Furthermore, the process of generating the calibration trigger flag also includes:

[0025] The final compensation output data is obtained, and the standard value is compared. The root mean square error is calculated and the fast Fourier transform harmonic extraction is performed using the joint analysis method of time domain and frequency domain to generate an error distribution spectrum.

[0026] The out-of-limit nodes are extracted from the error distribution map, and a multi-level threshold filtering algorithm and machine learning classification model are used to conduct stability assessment, including the adaptive sliding window algorithm in calculating error volatility and trend prediction in the multivariate regression model, and a verification conclusion report is generated.

[0027] The threshold determination of the verification conclusion report adopts the multi-condition threshold rule in logical judgment and the comprehensive index decision in fuzzy reasoning technology to generate a calibration trigger flag.

[0028] Furthermore, the process of generating a closed-loop calibration instruction set also includes:

[0029] The calibration trigger flag is analyzed to identify the calibration type, including bias elimination in zero-point calibration, sensitivity adjustment in gain calibration, and curve fitting in nonlinear calibration, and to generate calibration task instructions.

[0030] The parameter offsets, including zero-point drift, gain error, and nonlinear distortion, are extracted from the calibration task instructions. The calibration intensity is calculated by dynamically adjusting the weights of aging trend and environmental change factors in the weighted regression model to generate calibration intensity coefficients.

[0031] The timing arrangement of calibration intensity coefficients is achieved by balancing the urgency of the system with the system load in a priority-based scheduling algorithm, thereby generating a closed-loop calibration instruction set.

[0032] Furthermore, the process of generating the closed-loop calibration completion flag also includes:

[0033] Receive closed-loop calibration instruction set, perform sensor aging coefficient analysis, and use dynamic update mechanism in multivariate regression analysis and machine learning algorithm to generate compensation calibration parameters;

[0034] Key correction terms are extracted from the compensation calibration parameters and evaluated using historical stability and error contribution in the weighted allocation algorithm. The parameter storage area is updated using backup area writing and main storage area switching in the double buffering mechanism to generate a new version of sensor parameters.

[0035] The verification and write-back of the new version of sensor parameters includes verification and comparison in parameter integrity verification and multi-dimensional statistical analysis of indicators in online verification, and generates a closed-loop calibration completion mark.

[0036] Furthermore, the process of extracting key feature values ​​and performing adaptive algorithm matching includes:

[0037] Multidimensional feature analysis of magnetic field inhomogeneity parameters is performed. Principal component analysis is used to reduce the dimensionality of the parameter space, remove redundant information, and retain the main components that have a significant impact on the compensation effect.

[0038] Calculate the parameter covariance matrix, extract the eigenvectors and corresponding eigenvalues, and determine the key features by sorting them according to the size of the eigenvalues.

[0039] Furthermore, the expression for generating the initial compensation coefficients includes:

[0040] Calculate the feature entropy value:

[0041]

[0042] Among them, E 1 ω is the characteristic entropy value; k is the index of the ring magnetic sensitive element; n is the total number of ring magnetic sensitive elements; (k) For improved Shannon entropy weights; M (k) Let M be the magnetic field inhomogeneity parameter matrix of the k-th ring magnetic sensing element; PCA(M (k) ) is for matrix M (k) Principal component scores obtained from principal component analysis;

[0043] E 1 After normalization, the principal component score vector S is generated. 1 :

[0044] S 1 =[S 1(1) ,S 1(2) ,S 1(3) ]

[0045] Among them, S1 The principal component score vector is used, and its first three components are used as key feature basis vectors, i.e., S. 1 =[S 1(1) ,S 1 (2) ,S 1(3) ];

[0046] Define dynamic correlation index:

[0047]

[0048] Among them, C 2 This represents the dynamic correlation index; max is the maximum value function; t is the time index; T is the total sampling window length. Principal component score vector S 1 The a-th and b-th components at time t; τ is the time shift;

[0049] Define the improved Chebyshev norm criterion:

[0050]

[0051] Among them, F 1 For key feature vectors; ||·|| T This represents the improved Chebyshev norm; max is the maximum value function. For F 1 The i-th feature; |·| is the absolute value function; α i The characteristic stability coefficient; ||F 1 || T The weighted criterion value;

[0052] Input to the quadratic cost function:

[0053]

[0054] Among them, J 4 The cost function value; W c This is the compensation coefficient vector; (·) T This is the transpose operation for a vector or matrix; Q is the weight matrix; λ is the smoothing factor. For the feature gradient; ||·|| 2 To represent the square of the L2 norm;

[0055] A dynamic matching process is constructed using an improved Hamiltonian equation:

[0056]

[0057] Among them, H 5 It is the Hamiltonian function; For partial derivative operators; Φ(t) is the gradient vector of the cost function with respect to the compensation coefficient; η is the time-varying coupling matrix; W is the attenuation coefficient; c This is the compensation coefficient vector;

[0058] Adjust via the limiting function:

[0059]

[0060] in, Let be the initial compensation coefficient for the j-th element; j is the index of the compensation coefficient vector. The j-th element of the compensation coefficient vector; ΔW j 5 The term is the increment term; sat(·) is the amplitude limiting function.

[0061] Furthermore, the process of extracting key feature values ​​and performing adaptive algorithm matching also includes:

[0062] The dynamic changes of magnetic field inhomogeneity parameters are identified as periodic or abrupt, and an adaptive sliding window algorithm is used to capture key feature values.

[0063] Key feature values ​​are used as input to the adaptive algorithm and enter the matching module to execute the adaptive algorithm matching process, including using a dynamic weight adjustment mechanism to dynamically adjust the weight coefficients based on the current magnetic field state and historical compensation data.

[0064] The key innovations of this invention include:

[0065] (1) Based on the magnetic field inhomogeneity parameter, an initial compensation coefficient is generated by matching through an adaptive algorithm, which involves dynamic processing of extracting magnetic field feature values ​​and adapting the algorithm.

[0066] (2) A multi-ring magnetic sensitive element collaborative working mechanism is adopted, integrating real-time compensation signal generation and power adaptation processing to enhance the collaborative response capability of the multi-ring magnetic sensitive element.

[0067] (3) Construct a closed-loop control mechanism to realize the dynamic correlation between magnetic field parameter acquisition, feature extraction, compensation coefficient generation and power adaptation output, and improve the integrity of the compensation process.

[0068] The following are its main beneficial effects:

[0069] (1) By extracting key feature values ​​of magnetic field inhomogeneity parameters and matching them with an adaptive algorithm, the generated preliminary compensation coefficient can remain stable in complex magnetic field environments, reducing the occurrence of compensation lag effects.

[0070] (2) The collaborative working mechanism of the multi-ring magnetic sensitive element ensures the stability of the output of the multi-ring magnetic sensitive element and reduces the influence of signal coupling interference through real-time compensation signal power adaptation processing.

[0071] (3) The dynamic correlation of the closed-loop control mechanism enables the acquisition of magnetic field parameters, feature extraction, compensation coefficient generation and power adaptation output to form a complete process, which improves the accuracy and stability of the current measurement system under complex working conditions. Attached Figure Description

[0072] Figure 1 This is a flowchart illustrating a high-precision zero-flux compensation method provided in an embodiment of this application. Detailed Implementation

[0073] Example 1: Refer to Figure 1 This is a flowchart illustrating a high-precision zero-flux compensation method provided in an embodiment of the present invention. The process may include at least steps S100-S600:

[0074] S100: Acquire the measured current signal and perform signal noise reduction processing, including multi-level filtering mechanism and digital filtering algorithm; frequency band extraction and amplification using bandpass filter and programmable gain amplifier; and multi-ring magnetic sensor configuration processing including the layout design of multiple concentric ring magnetic sensing elements to generate magnetic field detection interface structure.

[0075] S200, based on magnetic field detection interface data, performs magnetic field distribution analysis using spatial filtering and multivariate statistical analysis methods, feature matching including principal component analysis and dynamic weight adjustment mechanism, and resource optimization processing including compression coding and dynamic adjustment of power consumption model to generate efficient compensation parameters;

[0076] S300 receives high-efficiency compensation parameters, performs interference identification using spectrum analysis algorithms and time-domain feature extraction technology, and uses a weighted fusion method and multi-objective optimization algorithm for compensation and recombination, as well as power adaptation processing, which is achieved through a digital power amplifier to generate the final compensation output structure.

[0077] S400: Obtain the final compensation output data, perform standard value comparison using time domain and frequency domain joint analysis method, stability assessment including calculation of error volatility and multivariate regression model and threshold judgment processing using logical judgment and fuzzy reasoning technology, and generate calibration trigger flag bit;

[0078] S500 analyzes the calibration trigger flag, performs type identification including zero-point calibration, gain calibration and nonlinear calibration, intensity calculation adopts a weighted regression model and time sequence arrangement processing adopts a priority-based scheduling algorithm to generate a closed-loop calibration instruction set;

[0079] S600 receives the closed-loop calibration instruction set, performs aging coefficient analysis using multivariate regression analysis and machine learning algorithms, and adopts a double buffering mechanism and verification write-back process for parameter updates, including parameter integrity verification and online verification, and generates a closed-loop calibration completion flag.

[0080] Step S100 includes at least steps S110-S130:

[0081] S110. Acquire the measured current signal and perform signal noise reduction processing, including low-pass filter and band-stop filter in the multi-level filtering mechanism and smoothing processing in the digital filtering algorithm, to obtain the pre-processed current signal.

[0082] The measured current signal is acquired, specifically from the primary input terminal of an external current sensing device, serving as the input basis for this step. First, the measured current signal is connected through a preset signal acquisition interface. This interface possesses high-precision sampling capabilities and anti-interference design, enabling real-time capture of current change information while meeting sampling frequency and resolution requirements. Subsequently, a multi-stage filtering mechanism is used to denoise the acquired raw current signal. This includes a low-pass filter to remove high-frequency noise, a band-stop filter to suppress power frequency interference in specific frequency bands, and a digital filtering algorithm to smooth transient interference signals. The denoising process is completed collaboratively by hardware circuitry and embedded software, ensuring the effective preservation of the signal's time and frequency domain characteristics. Further, amplitude normalization and baseline correction are performed on the denoised current signal to eliminate the effects of system bias and zero-point drift, ensuring the accuracy of subsequent processing. Throughout the entire processing flow, the abnormal signal detection module continuously monitors the integrity and stability of the signal waveform. When abnormal fluctuations or sampling distortion are detected, an abnormal recording mechanism is automatically triggered, and the current data segment is marked for subsequent analysis. Finally, the current signal that has undergone the above processing is defined as the preprocessed current signal. This signal is used as the output field name of this step and is passed to the "signal input" of the next step S120 for subsequent processing.

[0083] S120. Extract the effective frequency band from the preprocessed current signal, optimize the filter coefficients using the digital finite impulse response filter structure and window function method in the bandpass filter, amplify the signal, and generate a standard input signal by adaptive gain adjustment in the programmable gain amplifier.

[0084] Based on the preprocessed current signal, further processing to extract the effective frequency band first clarifies the frequency range of the effective frequency band. According to the typical spectral characteristics of the current signal and the application scenario requirements, lower and upper frequency limits are set. A bandpass filter is used to filter the preprocessed current signal by frequency, eliminating interference signals in irrelevant frequency bands to improve the signal-to-noise ratio and processing efficiency. Specifically, the filter design adopts a digital finite impulse response (FIR) filter structure, combined with the window function method to optimize the filter coefficients, ensuring the passband flatness and stopband attenuation performance of the filter. Subsequently, the filtered signal undergoes dynamic range amplification. This amplification process is implemented through a programmable gain amplifier (PGA). The gain value is adaptively adjusted according to the input signal amplitude to avoid signal overload or weakness, ensuring that the signal amplitude meets the input requirements of subsequent processing modules. During gain adjustment, the amplifier output is monitored in real time, and the gain parameters are adjusted according to the feedback loop to complete closed-loop control. The amplified signal is sampled by an analog-to-digital converter (ADC) and converted into a digital signal format. The digital signal undergoes further amplitude correction and timing alignment to resolve timing errors caused by sampling clock deviations. Throughout the processing chain, an abnormal signal detection mechanism continuously operates, recording alarms for sudden changes in signal amplitude or abnormal spectrum. Finally, the standard input signal generated by the processing is used as the output field name for this step and passed to the "input source" in the next step S130 for configuration of the multi-ring magnetic sensitive element.

[0085] S130. Configure the standard input signal with multi-ring magnetic sensitive elements, including spatial gradient sensing layout design and synchronous sampling technology of multiple concentric ring magnetic sensitive sensing elements, to generate magnetic field detection interface structure.

[0086] For the standard input signal, a multi-ring magnetic sensing element configuration is specifically implemented to construct a magnetic field detection interface structure. This multi-ring magnetic sensing element configuration includes a layout design of multiple concentric ring magnetic sensing elements, each with a different detection radius, forming a multi-ring magnetic field sensing hierarchy. First, the standard input signal is used as a driving signal and injected into the excitation port of each ring magnetic sensing element. The excitation signal is regulated by a matching network to ensure signal transmission efficiency and amplitude consistency in each ring element. The sensing port of the multi-ring magnetic sensing element receives the magnetic flux change signal caused by the external magnetic field. This signal undergoes primary amplification and filtering to eliminate environmental noise and crosstalk. Furthermore, for the output signal of each ring element, synchronous sampling technology is used for data acquisition to ensure the time synchronization of signals from different ring elements, facilitating accurate subsequent analysis of the multi-ring magnetic field distribution. The magnetic field detection interface structure also includes a signal conditioning module to perform amplitude normalization and phase correction on the acquired signal, eliminating signal distortion caused by differences in element parameters and wiring. During the configuration process, the system verifies the working status of each ring element through a self-test program, detecting open circuits, short circuits, and abnormal responses. Abnormal information is recorded and fed back to the system management unit. Finally, the magnetic field detection interface of the configured multi-ring magnetic sensitive element is passed as the output field name of this step to the "input interface" of the next step S210 for use by the adaptive compensation algorithm.

[0087] Step S200 includes at least steps S210-S230:

[0088] S210. Obtain magnetic field detection interface data and perform multi-ring magnetic field distribution analysis. Use the linear weighting algorithm in spatial filtering and the covariance matrix eigenvalue decomposition in multivariate statistical analysis to obtain the magnetic field inhomogeneity parameters.

[0089] The magnetic field detection interface data is acquired. Specifically, the magnetic field detection interface is used as the input source, and data acquisition and processing are performed through a multi-ring magnetic field distribution analysis module. First, the system synchronously samples the output signals of the multi-ring magnetic sensitive elements through a high-speed data acquisition unit. The sampling process uses a unified clock source for timing synchronization to eliminate time deviations between different ring magnetic sensitive elements and ensure the time consistency of the sampled data. The sampled data undergoes amplitude normalization and phase correction by a primary signal conditioning module to eliminate signal distortion caused by differences in element parameters and wiring. Subsequently, the analysis module performs spatial filtering on the acquired multi-ring signals. The spatial filter is used to weight and superimpose the signals of different ring magnetic sensitive elements to extract the spatial gradient information of the magnetic field, reflecting the local variation characteristics of the magnetic field in the measured area. Specifically, the spatial filter design is based on a linear weighting algorithm, and the weight parameters are dynamically adjusted according to the detection radius and sensitivity of the ring elements to enhance the response capability to non-uniform magnetic field regions. Further, a multivariate statistical analysis method is used to process the spatial gradient data. By calculating the covariance matrix and its eigenvalue decomposition, key indicators reflecting the non-uniformity of the magnetic field distribution, namely the magnetic field non-uniformity parameter, are extracted. During the analysis process, the system automatically identifies abnormal sampling points, removes sampling noise and abnormal data based on a set threshold, and compensates for missing data using an interpolation algorithm to ensure the continuity and accuracy of the magnetic field inhomogeneity parameter. This magnetic field inhomogeneity parameter, as the output field name of this step, is passed to the "input parameter" of the next step S220 for use by the adaptive algorithm. Simultaneously, this parameter also provides basic data for resource consumption optimization in the subsequent step S230.

[0090] S220. Key feature values ​​are extracted from the magnetic field inhomogeneity parameters. The improved Shannon entropy weighting and dynamic correlation index in the principal component analysis method are used for calculation. Adaptive algorithm matching is performed, including historical compensation data feedback in the dynamic weight adjustment mechanism, to generate preliminary compensation coefficients.

[0091] The system extracts key feature values ​​from the magnetic field inhomogeneity parameters. Specifically, it first performs multi-dimensional feature analysis on the parameters, using Principal Component Analysis (PCA) to reduce the dimensionality of the parameter space, eliminating redundant information and retaining the main components that significantly affect the compensation effect. This process calculates the parameter covariance matrix, extracts eigenvectors and corresponding eigenvalues, and determines key features based on the size of the eigenvalues, ensuring that the extracted features accurately reflect the changing trend of magnetic field inhomogeneity. Furthermore, combining time series analysis techniques, the system identifies the periodicity and abrupt changes in the dynamic changes of the magnetic field inhomogeneity parameters, employing an adaptive sliding window algorithm to capture key feature values ​​and reflect real-time changes in the magnetic field state. These key feature values ​​serve as input to the adaptive algorithm, entering the matching module to execute the adaptive matching process. The matching process specifically includes using a dynamic weight adjustment mechanism to dynamically adjust the weight coefficients based on the current magnetic field state and historical compensation data, achieving real-time optimization of the compensation coefficients. During the matching process, the system judges the validity of the key feature values ​​based on preset thresholds, filtering out abnormal or invalid data to ensure the stability and accuracy of the compensation coefficients. This process gradually approaches the optimal compensation coefficient through iterative calculation and feedback adjustment. The preliminary compensation coefficient, as the output field name of this step, is passed to the "calculation input" of the next step S230, providing data support for resource consumption optimization and laying the foundation for subsequent dynamic compensation parameter adjustments.

[0092] In another embodiment, in S220, the first input data is the magnetic field inhomogeneity parameter matrix output in step S210, denoted as M. (k) Where k∈[1,8] represents 8 ring magnetic sensitive elements. Specifically, through the multivariate statistical analysis module, the improved Shannon entropy weighted principal component analysis (PCA) method is used to decouple the parameters and calculate the characteristic entropy value (formula ①):

[0093]

[0094] in:

[0095] E 1 The feature entropy value reflects the information density of the parameter matrix and serves as the basis for feature selection.

[0096] k is the index of the ring-shaped magnetic element;

[0097] n is the total number of ring-shaped magnetic sensitive elements;

[0098] ω (k) The improved Shannon entropy weights are calculated from the sensitivity index of the ring element extracted from the magnetic field detection interface data;

[0099] M(k) The magnetic field non-uniformity parameter matrix of the k-th annular magnetic sensitive element output in step S210;

[0100] PCA(M (k) ) is for matrix M (k) Principal component scores obtained from principal component analysis.

[0101] Processing procedure and output:

[0102] Calculated E 1 After normalization, the principal component score vector S is generated. 1 Take the first three components as the key feature basis vectors, denoted as:

[0103] S 1 =[S 1(1) ,S 1(2) ,S 1(3) ]

[0104] in:

[0105] S 1 The principal component score vector is generated from the normalized output of formula ①, with its first three components serving as the key feature basis vectors, i.e., S. 1 =[S 1(1) ,S 1(2) ,S 1(3) ].

[0106] The temporal correlation of principal component score vectors is analyzed using the cross-correlation algorithm, and a dynamic correlation index is defined (Formula ②):

[0107]

[0108] in:

[0109] C 2 It is a dynamic correlation index used to identify periodic features;

[0110] max is the maximum value function, which takes the maximum summation among all values ​​of τ;

[0111] t is the time index, representing the sampling time point;

[0112] T is the total sampling window length, which is the total number of sampling periods;

[0113] Principal component score vector S 1 At time t, the a-th and b-th components, a, b ∈ [1, 3];

[0114] τ is the time shift, with a value range of [-5, 5] and a sampling period, configured by the threshold and strategy table.

[0115] This step outputs: Key feature vector F 1 Including entropy weight ω (k) Principal component score S 1 and dynamic correlation C 2 This serves as the input for the next step, the "feature fusion module".

[0116] Furthermore, the key feature vector F that receives the above output 1 This step uses a dynamic weight adjustment mechanism to perform feature fusion, enhancing feature stability, and defines an improved Chebyshev norm criterion (Formula ③):

[0117]

[0118] in:

[0119] F 1 The key feature value vector contains features such as entropy weights, principal component scores, and time series correlations, with a dimension of 5.

[0120] ‖·‖ T This represents the improved Chebyshev Norm (CN), which is the maximum of the weighted absolute values ​​among all elements.

[0121] `max` is the maximum value function;

[0122] For F 1 The i-th dimension feature, i∈[1,5];

[0123] |·|: Absolute value function;

[0124] α i The feature stability coefficient is calculated from the reciprocal of the variance of the corresponding feature in the historical compensation data.

[0125] ‖F 1 || T : Weighted criterion value, used to filter effective features.

[0126] The weighted criterion, together with the real-time magnetic field state parameters, is input into the quadratic cost function (Formula ④):

[0127]

[0128] in:

[0129] J 4 The cost function value;

[0130] W c The compensation coefficient vector is initialized from the baseline parameters in the strategy table.

[0131] (·) T This is a transpose operation for vectors or matrices;

[0132] Q is the weight matrix, and the entropy value E is calculated from it. 1 Diagonalization generation;

[0133] λ is a smoothing factor, with a value range of [0.1, 1.0], which is dynamically configured based on the threshold and strategy table;

[0134] The feature gradient is calculated as the difference in features between adjacent sampling windows.

[0135] ‖·‖ 2 Let represent the square of the L2 norm.

[0136] This step outputs: the optimized cost function value J. 4 and compensation coefficient vector W c This is used for subsequent matching engine processing.

[0137] Furthermore, based on the cost function J of the above output... 4 An improved Hamiltonian equation is used to construct a dynamic matching process, thereby achieving closed-loop optimization of the compensation coefficient (Formula ⑤):

[0138]

[0139] in:

[0140] H 5 The Hamiltonian function describes the energy state of the system;

[0141] This is the partial derivative operator, representing the differentiation with respect to time;

[0142] Let be the gradient vector of the cost function with respect to the compensation coefficients;

[0143] Φ(t) is a time-varying coupling matrix, derived from the spatial gradient data of the multi-ring magnetic field distribution analytical module;

[0144] η is the attenuation coefficient, a fixed value of 0.05, defined by the strategy table;

[0145] W c This is the compensation coefficient vector.

[0146] After verifying the system stability using Lyapunov functions, preliminary compensation coefficients W are generated. f Its elements are adjusted using a limiting function:

[0147]

[0148] in:

[0149] W j f Let be the initial compensation coefficient for the j-th element;

[0150] j is the index of the compensation coefficient vector;

[0151] W j c This is the j-th element of the compensation coefficient vector;

[0152] ΔW j 5 For incremental terms;

[0153] sat(·) is the amplitude limiting function, and the boundary value is defined by the strategy table to ensure that the compensation coefficient is within a reasonable range.

[0154] This step outputs: Preliminary compensation coefficient vector W f , which is a key parameter for S230 “computation input”.

[0155] Input-output relationships between each step and closed-loop description of the technical path

[0156] The key feature vector F is extracted using the improved Shannon entropy weighted PCA (Formula ①) and cross-correlation analysis (Formula ②). 1 This vector contains entropy weights, principal component scores, and time series correlations, and serves as input to the feature fusion module to enhance feature stability.

[0157] Based on the feature fusion results, a weighted criterion (Formula ③) is calculated and combined with the real-time magnetic field state to construct a quadratic cost function (Formula ④), outputting the cost value J. 4 and compensation coefficient vector W c This is used by the matching engine.

[0158] The compensation coefficients are optimized using the Hamiltonian dynamic matching equation (Equation ⑤) in a closed loop to generate the initial compensation coefficients W. f The parameter rationality is ensured by the limiting function, and the output field "preliminary compensation coefficient" is passed to the "calculation input" of S230 to complete the closed loop from feature extraction to compensation parameter generation.

[0159] S230. Optimize the resource consumption of the preliminary compensation coefficients, including the sparse representation-based compression algorithm in compression coding and the recurrent neural network prediction in the dynamic adjustment of the power consumption model, to generate efficient compensation parameters.

[0160] The system optimizes resource consumption for the initial compensation coefficients. Specifically, it employs a distributed computing architecture and caching mechanism to process the compensation coefficient sequence. First, the initial compensation coefficients are compressed and encoded to reduce data transmission and storage resource consumption. The compression process uses a sparse representation-based compression algorithm, combined with the spatiotemporal correlation of the compensation coefficients, to achieve efficient encoding. Subsequently, a power consumption model is used to dynamically adjust the update frequency and computational complexity of the compensation parameters. Based on system operating status and environmental changes, the allocation of computing resources is optimized to reduce overall power consumption. Furthermore, the system combines a prediction model to predict the changing trend of the compensation parameters. Temporal prediction algorithms such as Recurrent Neural Networks (RNNs) are used to predict and adjust future compensation parameters, preparing the compensation sequence in advance and reducing the real-time computation burden. During the optimization process, the system continuously corrects the prediction error through a feedback mechanism to ensure the accuracy and timeliness of the compensation parameters. The optimization result forms an efficient compensation parameter sequence with high computational efficiency and compensation accuracy. This efficient compensation parameter is used as the output field name of this step and passed to the "adjustment benchmark" in the next step S310, providing a basis for dynamic compensation parameter adjustment and supporting the real-time response of the subsequent dynamic compensation control module.

[0161] Step S300 includes at least steps S310-S330:

[0162] S310: Receives efficient compensation parameters and performs external interference feature identification using Fast Fourier Transform technology in spectrum analysis algorithms and Short-Time Energy Calculation and Transient Detection algorithms in time-domain feature extraction techniques to generate dynamic adjustment factors;

[0163] The system receives the high-efficiency compensation parameters, specifically, the high-efficiency compensation parameters are derived from the "adjustment reference" in step S230 and serve as input to the dynamic compensation parameter adjustment module. First, the high-efficiency compensation parameters are loaded into the dynamic compensation control unit via a high-speed data bus interface. This unit integrates a multi-channel signal processor and an interference identification module. The dynamic compensation control unit ensures that the timing of the high-efficiency compensation parameters is consistent with the current system state data through a preset signal synchronization mechanism. Subsequently, the dynamic compensation control unit initiates the external interference feature identification process. Specifically, it uses a spectrum analysis algorithm combined with time-domain feature extraction technology to collect and analyze environmental electromagnetic interference signals in real time. This collection is accomplished through a multi-point sensor array, whose layout covers key interference sources around the measured area, ensuring complete spatial coverage of the interference signal. The spectrum analysis section uses Fast Fourier Transform (FFT) technology combined with an adaptive window function to accurately capture the frequency distribution and amplitude variation characteristics of the interference signal. Time-domain feature extraction uses short-time energy calculation and transient detection algorithms to identify the duration and suddenness characteristics of the interference signal. The system further utilizes machine learning models to classify the collected interference data, distinguishing different types of interference sources and their impact levels, and dynamically updating the interference database. Based on the above analysis results, the dynamic compensation control unit calculates a dynamic adjustment factor. This factor maps external interference characteristics to the adjustment space of compensation parameters through a weighting function. The weighting function design considers the combined effects of interference frequency, amplitude, and duration, and generates an adaptive adjustment factor in conjunction with the current compensation state of the system. The calculation process of the adjustment factor includes normalization, threshold determination, and smoothing filtering to avoid drastic fluctuations in compensation parameters caused by transient interference. The dynamic adjustment factor, as the output field name of this step, is passed to the "control input" of the next step S320 for compensation current reconstruction.

[0164] S320. Priority parameters are extracted from the dynamic adjustment factor. The weighted summation method in the multi-dimensional evaluation algorithm is used to perform compensation current recombination. The dynamic weight adjustment in the weighted fusion method and the combination of the genetic algorithm and particle swarm optimization algorithm in the multi-objective optimization algorithm are used to generate real-time compensation signals.

[0165] Priority parameters are extracted from the dynamic adjustment factors. Specifically, the dynamic compensation parameter adjustment module further performs priority analysis on the dynamic adjustment factors, assigning priority weights to different adjustment factors based on the type, intensity, and impact of the interference signal on the system compensation effect. The extraction of priority parameters employs a multi-dimensional evaluation algorithm, combining statistical characteristics, time-series trends, and historical compensation effect feedback to form a comprehensive evaluation index. This algorithm normalizes each index and calculates the final priority score using a weighted summation method. Subsequently, based on the priority score, the system reorganizes the compensation current. Specifically, a weighted fusion method is used to superimpose and adjust multiple dynamic adjustment factors with the basic compensation parameters. The compensation current reorganization process employs a multi-objective optimization algorithm to balance compensation accuracy and system stability, dynamically adjusting the amplitude and phase of the compensation signal. The optimization algorithm combines genetic algorithms and particle swarm optimization algorithms, adjusting for real-time computing resources. During the reorganization process, the system monitors the current waveform characteristics of the compensation signal in real time and uses a feedback control mechanism to adjust the weighting coefficients, ensuring that the compensation signal meets preset electromagnetic compatibility and power limits. The generated real-time compensation signal is filtered and shaped by the digital signal processing module to eliminate high-frequency noise and phase distortion that may be introduced during the reconstruction process. The real-time compensation signal, as the output field name of this step, is passed to the "output source" of the next step S330 for use in power adaptation processing.

[0166] S330: Power adaptation processing of real-time compensation signal is performed through digital control loop and current sensing feedback and temperature compensation module in digital power amplifier to generate final compensation output structure.

[0167] The real-time compensation signal undergoes power adaptation processing. Specifically, the dynamic compensation parameter adjustment module receives the real-time compensation signal from the "output source" of the S320 and inputs it into the power adaptation unit. The power adaptation unit dynamically adjusts the power output level of the compensation signal based on system load status and environmental change parameters. The adjustment process is first implemented through a digital power amplifier (DPA), which, combined with a digital control loop, adjusts the amplification factor in real time. The digital control loop includes a current sensing feedback and temperature compensation module, which monitors the output power and ambient temperature, and dynamically adjusts the amplifier gain based on a preset adjustment curve. Power adaptation processing further includes power fluctuation adjustment, using a voltage regulation module to stabilize the input power supply and eliminate the influence of power supply noise on the compensation signal. The voltage regulation module employs a multi-stage filtering and soft-start mechanism to avoid transient fluctuations in the output current. The power adaptation unit also integrates overload protection and short-circuit detection functions, monitoring the current and voltage parameters of the output signal in real time, and automatically triggering a protection mechanism in case of abnormalities to prevent system damage. Finally, the compensated signal, after power adaptation processing, forms the final compensated output structure, which conforms to the system interface specification and supports the input requirements of the subsequent measurement error verification and analysis module. This final compensated output, as the output field name of this step, is passed to the "verification input" of the next step S410, and is also available for use by the subsequent measurement error verification, analysis, and calibration module.

[0168] Step S400 includes at least steps S410-S430:

[0169] S410. Obtain the final compensation output data, compare it with the standard value, and use the root mean square error calculation and fast Fourier transform harmonic extraction in the joint analysis method of time domain and frequency domain to generate an error distribution spectrum.

[0170] To acquire the final compensation output data, specifically, the final compensation output described in step S330 is used as input. First, the compensation signal is transmitted to the measurement error verification and analysis module via a high-speed data acquisition interface. This interface possesses high-precision sampling capabilities and a wide dynamic range, enabling real-time capture of the amplitude, phase, and spectral characteristics of the compensation signal. During acquisition, the system uses a unified clock source to synchronize the sampling timing, eliminating the impact of timing deviations on error analysis. Subsequently, the acquired compensation signal data is preprocessed, including noise reduction filtering and amplitude normalization. Specifically, a multi-order digital filter combination is used to filter out environmental noise and high-frequency interference components, ensuring the integrity of signal characteristics. Further, the system compares the compensation signal with preset standard values ​​using a joint time-domain and frequency-domain analysis method. Time-domain analysis includes calculating the root mean square error (RMSError) and transient response differences of the signal, while frequency-domain analysis uses Fast Fourier Transform (FFT) to extract the harmonic content and spectral distribution characteristics of the signal. During the comparison process, the system automatically identifies abnormal waveforms and abrupt change nodes, uses a threshold judgment mechanism to filter out sampling points with errors exceeding limits, and marks them as abnormal nodes. Anomaly node information is recorded in the error log database for subsequent stability analysis and calibration strategy development. Finally, based on the above analysis results, an error distribution map is generated, which displays the amplitude and frequency characteristics of the compensation error in both time series and spatial distribution. The error distribution map, as the output field name of this step, is passed to the "Analysis Source" in the next step S420 and can be used by the subsequent measurement error verification analysis and calibration module.

[0171] S420. Extract out-of-limit nodes from the error distribution map, use a multi-level threshold filtering algorithm and machine learning classification model to conduct stability assessment, including the adaptive sliding window algorithm in calculating error volatility and trend prediction in the multivariate regression model, and generate a verification conclusion report.

[0172] The system extracts out-of-limit nodes from the error distribution map. Specifically, the system first filters the error distribution map data, using a multi-level threshold filtering algorithm to identify nodes whose error amplitude and frequency exceed a preset safety range. This filtering algorithm, combined with statistical analysis and machine learning classification models, can distinguish between inherent system errors and anomalous errors, improving the accuracy of out-of-limit node identification. Subsequently, a stability assessment process is executed for the identified out-of-limit nodes. The stability assessment includes calculating the error volatility, duration, and recurrence frequency, and analyzing the temporal evolution trend of the error using an adaptive sliding window algorithm. Furthermore, the system combines environmental parameters and historical compensation data, using a multivariate regression model to predict the future development trend of the error and determine the stability level of the compensation system. During the assessment process, anomalous data points are removed by an anomaly detection module to ensure the reliability of the assessment results. The assessment results are output in the form of an automatically generated verification conclusion report. The report details the location, frequency, and possible causes of the out-of-limit error nodes, including error statistics and stability scores. The verification conclusion report, as the output field name of this step, is passed to the "judgment basis" of the next step S430 and is used by the subsequent calibration instruction generation and transmission module.

[0173] S430. The threshold judgment of the verification conclusion report adopts the multi-condition threshold rule in logical judgment and the comprehensive index decision in fuzzy reasoning technology to generate a calibration trigger flag bit.

[0174] The system performs threshold judgment on the verification conclusion report. Specifically, it first parses the error statistics and stability score in the report and makes a judgment based on preset threshold rules. These rules cover the upper limit of error amplitude, the limit of error duration, and the lower limit of stability score; the rule parameters can be dynamically adjusted through the system configuration file. During the judgment process, the system uses logical judgment and fuzzy reasoning techniques to comprehensively determine the calibration trigger decision based on multiple indicators. If the judgment result meets the calibration conditions, the system generates a calibration trigger flag, which includes a trigger timestamp, trigger reason, and related error information. This flag is transmitted to the calibration command generation module via a secure communication protocol to ensure the integrity and real-time nature of the information. If the trigger conditions are not met, the system records the judgment result and related data for subsequent trend analysis and parameter adjustment. The calibration trigger flag, as the output field name of this step, is transmitted to the "command source" of the next step S510 and can be called by the calibration command generation and transmission module.

[0175] Step 500 includes at least steps S510-S530:

[0176] S510: Analyze the calibration trigger flag, identify the calibration type including bias elimination in zero-point calibration, sensitivity adjustment in gain calibration, and curve fitting in nonlinear calibration, and generate calibration task instructions.

[0177] The calibration trigger flag is parsed. Specifically, the calibration trigger flag originates from the "instruction source" in step S430 and serves as input to the calibration instruction generation and transmission module. First, the calibration trigger flag is received through a secure communication interface. This interface supports encrypted transmission and data integrity verification, ensuring the security and accuracy of the signal during transmission. The calibration trigger flag includes a trigger timestamp, trigger reason, and related error information. The system first verifies the integrity of the flag's format and content, confirming that it conforms to the preset protocol standard. Subsequently, the parsing module, based on the trigger reason field, calls the predefined calibration type identification logic, combining the system's current state and historical calibration records to determine the required calibration type. Calibration types include, but are not limited to, zero-point calibration, gain calibration, and nonlinear calibration, each corresponding to different calibration task requirements. During the parsing process, the system uses a rule engine and state machine model to automatically identify and classify multiple calibration types, supporting multi-condition composite judgment. Based on the identification results, the system generates a calibration task instruction, which includes key information such as task identifier, execution priority, parameter range, and execution time window. The generated calibration task instruction is formatted to conform to the system instruction set specification, facilitating parsing and execution by subsequent modules. Throughout the entire processing chain, any abnormal calibration trigger flags or identification failures are recorded in the system log and trigger an abnormal alarm mechanism for maintenance personnel to diagnose. The calibration task instruction, as the output field name of this step, is passed to the "execution basis" of the next step S520 for use in calculating calibration intensity. At the same time, this instruction also provides basic data for the generation of subsequent closed-loop calibration instruction sets.

[0178] S520. Extract parameter offsets, including zero-point drift, gain error, and nonlinear distortion indicators, from the calibration task instruction. Perform calibration intensity calculation by dynamically adjusting the weights of aging trend and environmental change factors in the weighted regression model to generate calibration intensity coefficients.

[0179] The system extracts parameter offsets from the calibration task instruction. Specifically, the calibration task instruction generated in step S510 is used as input to the calibration intensity calculation module. First, the system parses the parameter range and task identifier in the calibration task instruction to identify the types of sensor parameters to be adjusted and their offset magnitudes. The parameter offsets include multi-dimensional deviation indicators such as zero-point drift, gain error, and nonlinear distortion. Subsequently, the system constructs a mapping relationship between parameter offsets and compensation requirements using historical calibration data and real-time measurement error information. The mapping process uses a weighted regression model, dynamically adjusting the weight coefficients based on sensor aging trends and environmental changes to achieve accurate estimation of the offsets. Further, the calibration intensity calculation calculates the intensity coefficient of the calibration task based on the mapping results, combined with system power consumption limits and execution resource status. The intensity coefficient reflects the magnitude and frequency of the calibration operation, specifically achieved by adjusting the calibration parameter update step size and execution count. During the calculation process, the system sets upper and lower limits to prevent over-calibration from causing system instability. The calibration intensity calculation module also integrates a feedback mechanism to monitor the calibration effect in real time and adjust the intensity parameters. Abnormal data or calculation anomalies are recorded and trigger warnings to ensure the safety and effectiveness of the calibration process. The calibration intensity coefficient is used as the output field name of this step and is passed to the "control parameter" of the next step S530 for use in the generation of closed-loop calibration instruction set, while also supporting the execution optimization of the subsequent sensor parameter self-calibration module.

[0180] S530. The timing arrangement of calibration intensity coefficients is performed by balancing the urgency level and system load in a priority-based scheduling algorithm to generate a closed-loop calibration instruction set.

[0181] The calibration intensity coefficients are time-series orchestrated. Specifically, the calibration intensity coefficients output in step S520 are used as input to the closed-loop calibration instruction set generation module. First, the system plans the timing of calibration tasks based on the calibration intensity coefficients, combined with the current system status and historical calibration execution records. The timing orchestration employs a priority-based scheduling algorithm, balancing the urgency of calibration tasks with system load to ensure calibration operations are executed within appropriate time windows. The system supports multi-task parallel and serial execution modes, dynamically adjusting execution strategies for different calibration types and intensities. Furthermore, during orchestration, the system considers the sensor's working cycle and environmental change rhythm to avoid calibration operations interfering with normal measurements. The timing orchestration module integrates an anomaly detection mechanism, monitoring task execution status in real time and automatically adjusting subsequent plans for delayed or failed tasks. Subsequently, the system generates a closed-loop calibration instruction set based on the timing arrangement. This instruction set includes specific calibration command sequences, execution parameters, timestamps, and calibration feedback path information. The instruction set uses a unified format encoding, supporting remote distribution and local storage, facilitating access and management by various system modules. The generated closed-loop calibration instruction set undergoes multi-level verification, including format verification, logical consistency verification, and security authentication, ensuring the integrity and validity of the instruction set. Abnormal or conflicting instructions are automatically removed and recorded, triggering a system alarm. The closed-loop calibration instruction set, as the output field name of this step, is passed to the "input instruction" of the next step S610 for execution by the sensor parameter self-calibration module, while also supporting the realization of the overall system calibration closed loop.

[0182] Step S600 includes at least steps S610-S630:

[0183] S610: Receives closed-loop calibration instruction set, performs sensor aging coefficient analysis, and generates compensation calibration parameters by using dynamic update mechanism in multivariate regression analysis and machine learning algorithm.

[0184] The system receives the closed-loop calibration instruction set, specifically, the closed-loop calibration instruction set originating from the "input instruction" in step S530, serving as input to the sensor parameter self-calibration module. First, the closed-loop calibration instruction set is loaded into the self-calibration control unit via a secure communication interface. This interface supports encrypted transmission and data integrity verification, ensuring the security and accuracy of the instruction set during transmission. The self-calibration control unit performs format parsing and content verification on the received instruction set, confirming that the instruction set conforms to preset protocol specifications and logical consistency, eliminating format errors, duplicate instructions, and conflicting commands. Subsequently, the system initiates the sensor aging coefficient parsing process according to the task sequence in the instruction set. Specifically, the parsing module extracts aging model parameters matching the current sensor model and operating environment from the sensor's historical performance database, and calculates the sensor's aging coefficient by combining real-time measurement data and environmental status information. The aging coefficient reflects the degree of performance degradation of the sensor caused by factors such as time, temperature, humidity, and mechanical stress, and is dynamically updated using multivariate regression analysis and machine learning algorithms. During the parsing process, the system performs anomaly detection on the input data, and corrects abnormal or missing data using interpolation and filtering methods to ensure the accuracy and continuity of the aging coefficient calculation. The aging coefficient analysis results serve as the foundational data, entering the compensation calibration parameter generation module. This module combines the aging coefficient with the current calibration task requirements, employing a multi-objective optimization algorithm to generate compensation calibration parameters. These parameters include zero-point correction values, gain adjustment coefficients, and nonlinear compensation curves, specifically implemented through parameter mapping tables and model fitting functions. During generation, the system performs boundary constraints and stability checks on the compensation parameters to prevent over-adjustment that could lead to system instability. The compensation calibration parameters, as the output field names of this step, are passed to the "update source" in the next step S620 for sensor parameter updates, and also support the verification write-back process in subsequent step S630.

[0185] S620: Extract key correction terms from compensation calibration parameters, evaluate historical stability and error contribution in weight allocation algorithm, update parameter storage area using backup area writing and main storage area switching in dual buffering mechanism, and generate new version of sensor parameters.

[0186] Key correction items are extracted from the compensation calibration parameters. Specifically, the compensation calibration parameters are used as input, and key correction items are extracted through a parameter parsing module. This module selects the parameters that have the greatest impact on sensor performance based on preset correction strategies and priority rules, including zero-point offset, sensitivity drift, and nonlinear error correction items. During the extraction process, the system uses a weighted allocation algorithm to dynamically adjust the priority of correction items by combining the historical stability of the parameters and the current contribution of measurement errors. Subsequently, the key correction items are passed to the parameter storage area update module, which performs parameter update operations on the sensor's internal memory or external configuration storage unit through a secure write interface. The update process adopts a double buffering mechanism: the new version of the sensor parameters is first written to the backup area, and then switched to the main storage area after the write is completed, ensuring the atomicity and data integrity of the update process. During the write operation, the system monitors the write status and error codes in real time. When an anomaly occurs, it automatically rolls back to the previous stable version and records the anomaly log. Furthermore, the update module performs parameter consistency verification, including verification and CAPTCHA detection, to confirm the correctness of the new version of the parameters. The new version of the sensor parameters is used as the output field name of this step and passed to the "Write Object" of the next step S630 for verification and write-back, and then fed back to step S110 to realize closed-loop control of configuration update.

[0187] S630. Verify and write back the parameters of the new version sensor, including verification and comparison in parameter integrity verification and multi-dimensional statistical analysis of indicators in online verification, and generate a closed-loop calibration completion mark.

[0188] The new version of the sensor parameters is verified and written back. Specifically, the new version of the sensor parameters output in step S620 are used as input to the verification and write-back module. This module first performs parameter integrity verification by comparing checksums and redundancy check codes to confirm that the parameter data has not been tampered with or damaged. Subsequently, the system performs online verification of the parameter application effect, specifically including real-time sampling of the sensor output signal and comparing the performance changes before and after calibration. The verification process uses multi-dimensional indicators, including the reduction in zero-point drift, the improvement in gain linearity, and the reduction in nonlinear error, combined with statistical analysis methods to calculate the verification score. If the verification result meets the preset qualification standard, the system automatically writes the new version of the sensor parameters back to the sensor configuration storage area, completing the formal activation of the calibration parameters. The write-back operation uses an encrypted writing protocol to prevent the parameters from being illegally modified. Further, the module generates a closed-loop calibration completion flag, which includes a calibration completion timestamp, a calibration task identifier, and a verification result summary. Digital signature technology is used to ensure the authenticity and non-repudiation of the flag. The closed-loop calibration completion flag is transmitted to the "configuration update" step S110 through the internal communication bus, triggering the signal acquisition module to load the latest sensor parameters, realizing the automatic update of the calibration closed loop. If the verification fails, the system automatically rolls back to the previous valid parameter version and records detailed anomaly information for maintenance personnel to analyze. An alarm mechanism is also triggered simultaneously in case of an anomaly, prompting the system to require manual intervention. The closed-loop calibration completion flag is used as the output field name for this step and is passed to the "Configuration Update" step S110, realizing closed-loop control of the sensor parameter self-calibration process.

Claims

1. A high precision zero flux compensation method, characterized in that, The method comprises the following steps: Obtain the measured current signal, and perform signal noise reduction processing including multi-stage filtering mechanism and digital filtering algorithm, frequency band extraction and amplification using band-pass filter and programmable gain amplifier, and multi-ring magnetic sensitive configuration processing including layout design of multiple concentric ring-shaped magnetic sensitive sensing elements, to generate a magnetic field detection interface structure; Based on the magnetic field detection interface data, perform magnetic field distribution analysis using spatial filtering processing and multivariate statistical analysis method, feature matching including principal component analysis and dynamic weight adjustment mechanism, and resource optimization processing including compression encoding and dynamic adjustment of power consumption model, to generate efficient compensation parameters; Receive the efficient compensation parameters, perform interference identification using spectral analysis algorithm and time domain feature extraction technology, compensation reorganization using weighted fusion method and multi-objective optimization algorithm, and power adaptation processing through digital power amplifier, to generate the final compensation output structure; Obtain the final compensation output data, and perform standard value comparison using time domain and frequency domain joint analysis method, stability evaluation including calculation of error fluctuation rate and multivariate regression model, and threshold judgment processing using logic judgment and fuzzy reasoning technology, to generate a calibration trigger flag; Analyze the calibration trigger flag, perform type identification including zero point calibration, gain calibration and non-linear calibration, intensity calculation using weighted regression model, and time sequence arrangement processing using priority-based scheduling algorithm, to generate a closed-loop calibration instruction set; Receive the closed-loop calibration instruction set, perform aging coefficient analysis using multivariate regression analysis and machine learning algorithm, parameter update using double buffering mechanism, and verification and write-back processing including parameter integrity verification and online verification, to generate a closed-loop calibration completion flag.

2. The method of claim 1, wherein, The process of generating the magnetic field detection interface structure further comprises: Obtain the measured current signal, and perform signal noise reduction processing including low-pass filter and band-stop filter in the multi-stage filtering mechanism, and smoothing processing in the digital filtering algorithm, to obtain a preprocessed current signal; Extract the effective frequency band from the preprocessed current signal using the digital finite impulse response filter structure in the band-pass filter and the window function method to optimize the filter coefficients, and perform signal amplification through gain self-adaptive adjustment in the programmable gain amplifier, to generate a standard input signal; Perform multi-ring magnetic sensitive element configuration on the standard input signal, including spatial gradient sensing layout design of multiple concentric ring-shaped magnetic sensitive sensing elements and synchronous sampling technology, to generate the magnetic field detection interface structure.

3. The method of claim 1, wherein, The process of generating the efficient compensation parameters further comprises: Obtain the magnetic field detection interface data, and perform multi-ring magnetic field distribution analysis using linear weighting algorithm in spatial filtering processing and covariance matrix eigenvalue decomposition in multivariate statistical analysis method, to obtain a magnetic field non-uniformity parameter; Extract the key eigenvalue from the magnetic field non-uniformity parameter using the improved Shannon entropy weighting in the principal component analysis method and the dynamic correlation index calculation, and perform adaptive algorithm matching including historical compensation data feedback in the dynamic weight adjustment mechanism, to generate a preliminary compensation coefficient; Optimize the resource consumption of the preliminary compensation coefficient including the sparse representation-based compression algorithm in compression encoding and the recurrent neural network prediction in dynamic adjustment of power consumption model, to generate the efficient compensation parameters.

4. The method of claim 1, wherein, The process of generating the final compensation output structure further comprises: Receiving high-efficiency compensation parameters, performing external interference feature recognition adopts fast Fourier transform technology in spectrum analysis algorithm and short-time energy calculation and transient detection algorithm in time domain feature extraction technology to generate dynamic adjustment factor; Extracting priority parameters from dynamic adjustment factor adopts weighted summation method in multi-dimensional evaluation algorithm, performing compensation current reorganization adopts dynamic weight adjustment in weighted fusion method and genetic algorithm combined with particle swarm optimization algorithm in multi-objective optimization algorithm to generate real-time compensation signal; Performing power adaptation processing on real-time compensation signal generates final compensation output structure through digital control loop in digital power amplifier, current sensing feedback and temperature compensation module.

5. The method of claim 1, wherein, The process of generating calibration trigger flag also includes: Obtaining final compensation output data, performing standard value comparison adopts root mean square error calculation and fast Fourier transform harmonic extraction in time and frequency domain joint analysis method to generate error distribution map; Extracting over-limit nodes from error distribution map adopts multi-level threshold filtering algorithm and machine learning classification model, performing stability evaluation includes adaptive sliding window algorithm in error fluctuation rate calculation and trend prediction in multivariate regression model to generate verification conclusion report; Performing threshold judgment on verification conclusion report generates calibration trigger flag using multi-condition threshold rule in logic judgment and comprehensive index decision in fuzzy reasoning technology.

6. The method of claim 1, wherein, The process of generating closed-loop calibration instruction set also includes: Analyzing calibration trigger flag, performing calibration type recognition includes bias elimination in zero-point calibration, sensitivity adjustment in gain calibration and curve fitting in non-linear calibration to generate calibration task instruction; Extracting parameter offset from calibration task instruction includes zero-point drift, gain error and non-linear distortion index, performing calibration intensity calculation adopts dynamic weight adjustment of aging trend and environmental change factors in weighted regression model to generate calibration intensity coefficient; Performing time sequence arrangement on calibration intensity coefficient generates closed-loop calibration instruction set using emergency level and system running load balance in priority-based scheduling algorithm.

7. The method of claim 1, wherein, The process of generating closed-loop calibration completion flag also includes: Receiving closed-loop calibration instruction set, performing sensor aging coefficient analysis adopts multivariate regression analysis and dynamic update mechanism in machine learning algorithm to generate compensation calibration parameters; Extracting key correction items from compensation calibration parameters adopts weight distribution algorithm in historical stability and error contribution evaluation, performing parameter storage area update adopts backup area writing in double buffering mechanism and main storage area switching to generate new version sensor parameters; Verifying and rewriting new version sensor parameters includes checksum comparison in parameter integrity verification and multi-dimensional index statistical analysis in online verification to generate closed-loop calibration completion flag.

8. The method of claim 3, wherein, The process of extracting key feature values and performing adaptive algorithm matching includes: Performing multi-dimensional feature analysis on magnetic field non-uniformity parameters, adopting principal component analysis method to reduce dimension of parameter space, eliminating redundant information and retaining main components that significantly affect compensation effect; Calculating parameter covariance matrix, extracting feature vectors and corresponding eigenvalues, determining key features according to eigenvalue size sorting.

9. The method of claim 3, wherein, The expression of generating the preliminary compensation coefficient includes: Calculate the characteristic entropy value: wherein E 1 is the characteristic entropy value; k is the index of the annular magnetic sensitive element; n is the total number of annular magnetic sensitive elements; ω (k) is the improved Shannon entropy weight; M (k) is the magnetic field inhomogeneity parameter matrix of the kth annular magnetic sensitive element; PCA(M (k) ) is the principal component score obtained by performing principal component analysis on the matrix M (k) . E 1 After normalization, the principal component score vector S 1 : S 1 = [S 1(1) , S 1(2) , S 1(3) ] where S 1 is the main component score vector, whose first three components are taken as the key feature basis vectors, i.e., S 1 = [S 1(1) , S 1(2) , S 1 (3) ] ; Define the dynamic correlation degree index: where C 2 is the dynamic correlation index; max is the maximum function; t is the time index; T is the total sampling window length; is the principal component score vector S 1 is the a, b component at time t; τ is the time shift amount; Define the improved Chebyshev norm criterion: where F 1 is the vector of eigenvalues;‖·‖ T denotes the modified Chebyshev norm; max is the maximum function; is the i-th eigenvalue of F 1 ; |·| is the absolute value function; α i is the eigenvalue stability coefficient;‖F 1 ‖ T is the weighted criterion value; Input to the quadratic form cost function: where J 4 is the cost function value; W c is the compensation coefficient vector; (·) T is the transpose operation of a vector or matrix; Q is a weight matrix; λ is a smoothing factor; is the feature gradient; ‖·‖ 2 is the square of the L2 norm; Adopt the improved Hamilton equation to construct the dynamic matching process: where H 5 is the Hamiltonian function; is the partial derivative operator; is the gradient vector of the cost function with respect to the compensation coefficients; Φ(t) is the time-varying coupling matrix; η is the decay coefficient; W c is the compensation coefficient vector; Adjust through the limiting function: wherein is a preliminary compensation coefficient for the jth element; j is an index of the compensation coefficient vector; is the jth element of the compensation coefficient vector; is an increment term; sat(·) is a saturation function.

10. The method of claim 3, wherein, The process of extracting the key characteristic value and performing adaptive algorithm matching also includes: Periodicity and suddenness identification is performed on the dynamic change of the magnetic field non-uniformity parameter, and an adaptive sliding window algorithm is used to capture the key characteristic value; The key characteristic value is input into the matching module as the input of the adaptive algorithm, and the adaptive algorithm matching process is executed, including dynamically adjusting the weight coefficient according to the current magnetic field state and historical compensation data by using the dynamic weight adjustment mechanism.

Images