Nanocrystalline magnetic core current sensor calibration method and system for full-temperature range applications

Abstract

This application provides a calibration method and system for nanocrystalline magnetic core current sensors for applications across the entire temperature range, relating to the field of sensor calibration technology. The method includes acquiring complete calibration data for each sample at various temperature and current points; constructing a batch-specific prior knowledge base; performing actual calibration tests on the target sensor to obtain actual calibration data at sparse test points; and using a hierarchical Bayesian inference method to fuse prior constraints with observational evidence to generate a dedicated calibration model. This application solves the technical problem in existing technologies where individual differences within a batch prevent a unified calibration model from accurately adapting to each sensor, making it difficult to balance calibration efficiency and individual adaptation accuracy. It overcomes the deficiency of a unified model being unable to adapt to individual differences, achieving the technical effect of improving both calibration efficiency and individual adaptation accuracy.

Description

Technical Field

[0001] This application relates to the field of sensor calibration technology, specifically to a calibration method and system for nanocrystalline magnetic core current sensors for applications across the entire temperature range. Background Technology

[0002] Nanocrystalline core current sensors are a type of current measurement device that uses an iron-based nanocrystalline alloy core as the sensing element. After furnace annealing, nanocrystalline alloy materials possess excellent soft magnetic properties such as high permeability, high saturation magnetic induction, and low coercivity, leading to their increasingly widespread application in the field of current sensors. With the ever-increasing demands for accuracy and reliability in current measurement from fields such as new energy vehicles, smart grids, and industrial frequency converters, nanocrystalline core current sensors need to maintain stable measurement accuracy across the entire temperature range.

[0003] Sensor calibration is a crucial step in ensuring measurement accuracy. The fundamental task of calibration is to establish a precise mapping between the sensor's input current and output signal, and to eliminate the influence of various error factors on the measurement results through compensation methods. For nanocrystalline magnetic core current sensors, temperature is one of the most significant factors affecting calibration accuracy. Key parameters such as the permeability of the nanocrystalline magnetic core drift with temperature changes, and this drift exhibits nonlinear characteristics across the entire temperature range. Temperature variations affect the magnetic domain movement of the core, leading to a decrease in permeability and a weakening of magnetic coupling. Therefore, achieving accurate calibration across the entire temperature range is a critical issue that urgently needs to be addressed in the field of nanocrystalline magnetic core current sensor technology.

[0004] In summary, existing technologies suffer from the technical problem that the uniform calibration model cannot accurately adapt to each sensor due to individual differences within a batch, making it difficult for the calibration method to balance calibration efficiency and individual adaptation accuracy. Summary of the Invention

[0005] The purpose of this application is to provide a calibration method and system for nanocrystalline magnetic core current sensors for applications across the entire temperature range, in order to solve the technical problem in the prior art where the uniform calibration model cannot be accurately adapted to each sensor due to individual differences within a batch, making it difficult for the calibration method to balance calibration efficiency and individual adaptation accuracy.

[0006] To achieve the above objectives, this application provides a calibration method and system for nanocrystalline magnetic core current sensors applicable to the entire temperature range.

[0007] Firstly, this application provides a calibration method for nanocrystalline magnetic core current sensors for applications across the entire temperature range. This calibration method is implemented using a calibration system for nanocrystalline magnetic core current sensors for applications across the entire temperature range. The method includes: selecting samples and corresponding heat treatment process parameters from a batch of nanocrystalline magnetic core current sensors; performing complete calibration tests on the samples across the entire temperature range and current range; and obtaining complete calibration data for each sample at each temperature and current point. The complete calibration data is correlated with the heat treatment process parameters to construct a batch prior knowledge base. For the target sensor to be calibrated, based on the batch prior knowledge base, several test points are determined through an active sampling strategy, and the target sensor is subjected to actual calibration tests to obtain the actual calibration data of the target sensor under sparse test points. The common features in the batch prior knowledge base are used as the first layer of prior constraints, and the actual calibration data is used as observation evidence. The hierarchical Bayesian inference method is used to fuse the prior constraints and observation evidence to generate a dedicated calibration model for the target sensor in the full temperature range.

[0008] Secondly, this application also provides a calibration system for nanocrystalline magnetic core current sensors for applications across the entire temperature range, used to execute the calibration method for nanocrystalline magnetic core current sensors for applications across the entire temperature range as described in the first aspect. The calibration system includes: a first testing module, used to select samples and corresponding heat treatment process parameters from a batch of nanocrystalline magnetic core current sensors, and perform complete calibration tests on the samples across the entire temperature range and the entire current range, obtaining complete calibration data for each sample at each temperature point and each current point; and a correlation analysis module, used to correlate the complete calibration data of each sample with the corresponding heat treatment process parameters. The system performs correlation analysis between the fixed data and heat treatment process parameters to construct a batch prior knowledge base. The second testing module is used to determine several test points for the target sensor to be calibrated based on the batch prior knowledge base through an active sampling strategy, and to perform actual calibration tests on the target sensor to obtain actual calibration data of the target sensor under sparse test points. The fusion module is used to use the common features in the batch prior knowledge base as the first layer of prior constraints, and the actual calibration data as observation evidence, and uses a hierarchical Bayesian inference method to fuse the prior constraints and observation evidence to generate a dedicated calibration model for the target sensor in the full temperature range.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: Samples and corresponding heat treatment process parameters are selected from a batch of nanocrystalline magnetic core current sensors. Complete calibration tests are performed on the samples across the entire temperature and current range to obtain complete calibration data for each sample at each temperature and current point. The complete calibration data of each sample is correlated with the heat treatment process parameters to construct a batch prior knowledge base. For the target sensor to be calibrated, based on the batch prior knowledge base, several test points are determined using an active sampling strategy. Actual calibration tests are performed on the target sensor to obtain actual calibration data at sparse test points. Common features in the batch prior knowledge base are used as the first layer of prior constraints, and the actual calibration data is used as observational evidence. A hierarchical Bayesian inference method is employed to fuse the prior constraints and observational evidence to generate a dedicated calibration model for the target sensor across the entire temperature range. By combining batch prior knowledge base constraints with a small amount of active sampling and measurement, and by using hierarchical Bayesian inference to gradually integrate batch common features with individual measured data, a unique calibration model adapted to the individual characteristics of the target sensor can be generated under the condition of a small amount of measured data. This overcomes the defect that a unified model cannot adapt to individual differences and achieves the technical effect of improving calibration efficiency and individual adaptation accuracy.

[0010] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the calibration method for the nanocrystalline magnetic core current sensor applicable to the entire temperature range, as described in this application.

[0013] Figure 2 This is a schematic diagram illustrating the process of obtaining actual calibration data in the calibration method of the nanocrystalline magnetic core current sensor for full-temperature applications in this application.

[0014] Figure 3 This is a schematic diagram of the calibration system for a nanocrystalline magnetic core current sensor intended for use across the entire temperature range, as described in this application.

[0015] Explanation of reference numerals in the attached diagram: First test module 11, correlation analysis module 12, second test module 13, fusion module 14. Detailed Implementation

[0016] This application provides a calibration method and system for nanocrystalline magnetic core current sensors applicable to all temperature ranges. It addresses the technical problem in existing technologies where individual batch variations lead to a uniform calibration model that cannot accurately adapt to each sensor, resulting in a difficulty in balancing calibration efficiency and individual adaptation accuracy. By combining batch prior knowledge base constraints with a small amount of actively sampled measurements, and through hierarchical Bayesian inference to gradually fuse batch common features with individual measured data, a dedicated calibration model adapted to the individual characteristics of the target sensor can be generated even with limited measured data. This overcomes the limitation of uniform models failing to adapt to individual differences, achieving the technical effect of improving both calibration efficiency and individual adaptation accuracy.

[0017] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0018] Example 1, please refer to the appendix. Figure 1 This application provides a calibration method for a nanocrystalline magnetic core current sensor for applications across the entire temperature range. The method is applied to a calibration system for a nanocrystalline magnetic core current sensor for applications across the entire temperature range. The specific steps of the calibration method include: S100: Select samples and corresponding heat treatment process parameters from the batch of nanocrystalline magnetic core current sensors, and perform complete calibration tests on the samples in the full temperature range and full current range to obtain complete calibration data for each sample at each temperature point and each current point.

[0019] Specifically, a certain number of samples are selected from the current batch of nanocrystalline magnetic core current sensors, and the key process parameters of each sample during the heat treatment process are recorded in detail. These parameters include at least the annealing temperature profile, cooling rate, and holding time. For iron-based nanocrystalline alloy materials such as FeCuNbSiB, the annealing temperature is typically set in the range of 510 to 570 degrees Celsius, the holding time is generally between 4 and 10 hours, and the cooling rate is generally controlled between 50 and 100 degrees Celsius per hour. Different annealing temperatures directly affect the precipitation and growth of nanocrystals, thereby changing the permeability and temperature stability of the magnetic core. Therefore, it is essential to record the process parameters in conjunction with subsequent calibration data. Subsequently, a complete calibration test is performed on each selected sample across the entire temperature range and the entire current range. The entire temperature range typically refers to -40 degrees Celsius to 125 degrees Celsius, covering the entire operating temperature range of nanocrystalline magnetic core current sensors in typical application scenarios such as new energy vehicles and smart grids. The entire current range is determined according to the sensor's design range, for example, from zero amperes to the rated maximum current value. The specific calibration test procedure is as follows: The sample sensor is placed in a temperature chamber, and the chamber temperature is gradually set according to a preset temperature point sequence. After sufficient heat preservation at each temperature point to allow the magnetic core temperature to reach thermal equilibrium, a series of known standard current values ​​are applied to the sensor using a high-precision standard current source. Simultaneously, the corresponding output electrical signal of the sensor is recorded, thus obtaining the mapping relationship between the input current and the output signal at that temperature point. Typically, a dozen to twenty temperature sampling points are set across the entire temperature range. The spacing between temperature points is appropriately increased in the low-temperature and high-temperature regions where temperature changes are most drastic. At each temperature point, a dozen to several dozen current test points are set along the entire range, with the spacing between current points appropriately increased in the current range where the sensor response is highly nonlinear. Through this complete point-by-point, temperature-by-temperature scan, complete calibration data for each sample sensor at all temperature and current point combinations is obtained. This data not only contains the full picture of the response characteristics of the nanocrystalline magnetic core across the entire temperature range but also provides a sufficient raw data foundation for subsequently establishing a batch prior knowledge base.

[0020] S200: Correlate the complete calibration data of each sample with the heat treatment process parameters to construct a batch prior knowledge base.

[0021] Specifically, a systematic correlation analysis is performed on the complete calibration data of each sample sensor and its corresponding heat treatment process parameters to identify several key process parameters that have the most significant impact on the calibration characteristics across the entire temperature range. Simultaneously, a quantitative relationship is established between these parameters and the degree of deviation in sensor output across the entire temperature range. Based on this, the results of the correlation analysis are transformed into prior knowledge that can be used for subsequent calibration, i.e., a batch prior knowledge base is constructed. This knowledge base contains at least three levels of content: the first level is the common response characteristics of the nanocrystalline magnetic cores in this batch, namely the average response curves and statistical variance distribution of all samples across the entire temperature range; the second level is a library of correlation rules and correction coefficients between process parameters and calibration characteristics, enabling the direct retrieval of the corresponding initial calibration parameter estimates from the knowledge base once the specific process parameters of a new sensor are known; the third level is the statistical distribution information of inter-individual differences within the batch.

[0022] S300: For the target sensor to be calibrated, based on the batch prior knowledge base, several test points are determined through an active sampling strategy, and the target sensor is subjected to actual calibration tests to obtain the actual calibration data of the target sensor under sparse test points.

[0023] Specifically, for each target sensor to be calibrated on the production line, it is not necessary to perform a full-range, traversal calibration across the entire temperature range. Instead, based on the pre-built batch prior knowledge base, a number of test points with the largest amount of information are intelligently determined through an active sampling strategy. Then, actual calibration tests are performed only on these selected points to obtain the actual calibration data of the target sensor under sparse test points.

[0024] S400: The common features in the batch prior knowledge base are used as the first layer of prior constraints, and the actual calibration data are used as observation evidence. The hierarchical Bayesian inference method is used to fuse the prior constraints and observation evidence to generate a dedicated calibration model for the target sensor in the full temperature range.

[0025] Specifically, after obtaining sparse actual calibration data of the target sensor, the common patterns represented by the batch prior knowledge base are organically combined with the individual characteristics carried by the limited amount of measured data from the target sensor to generate a dedicated full-temperature-range calibration model for that target sensor. A hierarchical Bayesian inference method is employed, which updates the probabilistic framework of parameter cognition based on observational evidence. Its core logic is to use existing prior knowledge as a starting point and measured data as a correction basis to derive a fused new cognition. Hierarchical Bayesian further breaks down the prior knowledge into multiple levels, with each level processing information from different aspects, making information transmission more refined and reasonable. The deviation between the predicted values ​​of the output dedicated calibration model at the full-temperature-range test points and the actual measured values ​​in subsequent full-validation tests achieves accurate capture of the individual characteristics of the target sensor with minimal measured data and provides a reasonable and reliable quantification of uncertainty.

[0026] Furthermore, this application S100 includes: Obtain the preset application temperature range of the nanocrystalline magnetic core current sensor; conduct point-by-point tests on each sample within the full temperature range of the preset application temperature range at predetermined temperature intervals, and conduct point-by-point tests at predetermined current intervals within the rated current range at each temperature point, record the output response value of each sample at each test point, form complete calibration data for each sample, and simultaneously record the heat treatment process parameters of each sample during the preparation of the nanocrystalline magnetic core.

[0027] Specifically, the preset application temperature range is a temperature range determined based on the actual working scenario of the nanocrystalline magnetic core current sensor. For example, in the battery management system of new energy vehicles, the sensor needs to work in an environment of -40 degrees Celsius to 85 degrees Celsius, while in outdoor equipment of smart grids, the working temperature range is usually -25 degrees Celsius to 70 degrees Celsius. After determining the preset application temperature range, point-by-point tests are performed between the lower and upper limits of this temperature range according to a predetermined temperature interval. The temperature interval is generally selected as 10 degrees Celsius or 15 degrees Celsius. However, in the low-temperature and high-temperature ranges where the temperature drift characteristics change drastically, the temperature interval can be increased to 5 degrees Celsius to ensure that all key inflection points of the nonlinear change of the permeability of the nanocrystalline magnetic core with temperature can be captured. At each temperature point, tests are performed point-by-point within the rated current range at predetermined current intervals. The current interval is typically selected as 5% or 10% of the rated current. In the low-current segment where the sensor output response is most nonlinear and in the near-saturation high-current segment, the current interval can be increased to 2% or 3% of the rated current. This allows for precise characterization of the permeability change of the nanocrystalline magnetic core under weak magnetic fields and its saturation tendency under strong magnetic fields. During the test, for each combination of temperature and current points, the sensor's output response value at that point is recorded. This output response value is typically a voltage or current signal, amplified and filtered by the sensor's signal conditioning circuit. After all tests are completed, each sample sensor generates complete calibration data covering all temperature points within the preset application temperature range and all current points within the rated current range. Simultaneously, during the nanocrystalline magnetic core fabrication process, detailed heat treatment parameters for each sample need to be recorded to provide a basis for adjusting prior knowledge during subsequent calibration of new batches of products. The complete calibration data obtained through the above method covers all possible operating points of the sensor under actual working conditions, and can comprehensively reflect the response behavior of the nanocrystalline magnetic core under different temperatures and current excitations. It also provides sufficiently rich and accurate raw data support for the construction of the batch prior knowledge base.

[0028] Furthermore, this application S200 includes: The samples were grouped according to the numerical range of the heat treatment process parameters. The full-range calibration dataset within the same process parameter range was statistically summarized, and the statistical characteristics of the output response values ​​at each temperature and current point were calculated. Using the process parameter range as the index and the statistical characteristics of each test point as the content, a batch prior knowledge base between process parameters and calibration characteristics was constructed.

[0029] Specifically, in the process of constructing the batch prior knowledge base, the selected samples are first grouped according to the numerical range of their heat treatment process parameters. This is because even within the same production batch, the actual heat treatment conditions experienced by different magnetic core samples may have slight differences. For example, the annealing temperature may fluctuate within a range of plus or minus ten degrees Celsius around the set value, and the cooling rate may vary by several degrees Celsius per minute depending on the position in the furnace. After grouping these samples with similar process parameters into the same group, the magnetic properties of the samples within the group are more consistent. For each sample group, the calibration data of all samples within the group at all temperature points across the entire temperature range and all current points across the entire current range are aggregated to form the calibration dataset for that group. This dataset is then statistically processed to calculate statistical characteristics for each temperature point and each combination of current points. These statistical characteristics include two dimensions: central tendency and dispersion. The central tendency characterizes the typical response level of a sample group at the test point; that is, where the output values ​​of multiple sensors within the group are roughly concentrated if repeated tests are performed at the same temperature and current points. The central tendency uses the median or average value. The dispersion characterizes the degree of individual dispersion of a sample group at the test point; that is, how much difference exists in the output responses between different sensors within the group. In this invention, the dispersion is measured using the standard deviation. A larger standard deviation indicates more significant inconsistencies between individuals at that temperature and current point, and higher calibration uncertainty; conversely, a smaller standard deviation indicates good individual consistency.

[0030] For example, a group of ten sample sensors within a certain process parameter range might have their output response values, while all distributed around a central value, fluctuating to some extent at a test point at -40 degrees Celsius and 50% of the rated current. The calculated median value represents the typical output level of this group of sensors at that temperature and current point, while the calculated standard deviation quantifies the dispersion of these ten sensors at that point. A larger standard deviation indicates more significant individual differences in the nanocrystalline cores at that current point in the low-temperature region, requiring more experimental verification in subsequent calibration. After completing the statistical calculations for all temperature and current points within all process parameter groups, a structured batch prior knowledge base is constructed using the process parameter range as the index keyword and the median and standard deviation of each test point within that range as the main content. The batch prior knowledge base is essentially a two-dimensional index table. The horizontal axis consists of different column entries composed of various process parameter ranges, and the vertical axis consists of different row entries composed of combinations of temperature points in the entire temperature range and current points in the entire current range. Each cell stores a pair of statistical characteristics: the median and standard deviation at the test point location within the process parameter range.

[0031] Further details are attached. Figure 2As shown, S300 of this application includes: Based on the batch prior knowledge base, a high uncertainty temperature range and a strong nonlinear current range are obtained; the high uncertainty temperature range and the strong nonlinear current range are combined to form several test points; based on the several test points, the known current value output by the standard current source is applied to the target sensor in sequence, and after the ambient temperature of the target sensor stabilizes at the corresponding temperature point, the actual output response data of the target sensor is recorded to obtain the actual calibration data.

[0032] Specifically, in the implementation of the active sampling strategy, it is first necessary to acquire prior information on two dimensions: the high uncertainty temperature range and the strong nonlinear current range, based on the batch prior knowledge base. After obtaining the information in these two dimensions, representative temperature points in the high uncertainty temperature range and representative current points in the strong nonlinear current range are cross-combined to form several two-dimensional test points. That is, each selected temperature point, in conjunction with each selected current point, constitutes a specific test condition. This combination method can cover all the key areas where the two dimensions of highest uncertainty and strongest nonlinearity intersect with the fewest test points. Based on these formed test points, the target sensor is calibrated sequentially: the target sensor is placed in a temperature chamber, and the temperature chamber temperature is set to the temperature value corresponding to the first test point. A sufficiently long holding time is allowed for the temperature of the nanocrystalline magnetic core inside the sensor to stabilize sufficiently. Typically, the holding time is no less than thirty minutes to ensure that the deviation between the magnetic core temperature and the set temperature of the temperature chamber is less than ±1 degree Celsius. After the temperature stabilizes, a high-precision standard current source is used to apply the known standard current value corresponding to the test point to the sensor. After the sensor output signal stabilizes, its actual output response data is recorded. Then, following the test point list, switch to the next temperature and current point combination in sequence, repeating the above temperature control and measurement process until all selected test points have been tested. At this point, the actual calibration data of the target sensor at the test points is obtained. For example, in a specific instance, the above strategy is used to select five high-uncertainty temperature points and four strong nonlinear current points from twenty temperature points and twenty current points, cross-combining them to form twenty test points. This reduces the number of test points from the original four hundred to one-twentieth, but covers the key areas with the highest uncertainty and strongest nonlinearity. This allows a small amount of measured data to contain sufficiently rich individual characteristic information to support accurate individualized corrections in the subsequent hierarchical Bayesian inference process.

[0033] Furthermore, this application S300 includes: Based on the batch prior knowledge base, the statistical dispersion of calibration parameters at each temperature point is extracted, and several temperature regions with statistical dispersion greater than a preset dispersion threshold are marked as high uncertainty temperature intervals; based on the batch prior knowledge base, the nonlinear characteristic index of the response curve at each current point is extracted, and several current regions with nonlinear characteristic index greater than a preset nonlinear threshold are marked as strong nonlinear current intervals.

[0034] Specifically, when determining the high-uncertainty temperature range, the statistical dispersion of calibration parameters at each temperature point within the process parameter range of the target sensor is first extracted from the batch prior knowledge base. This statistical dispersion, or standard deviation, is a value pre-calculated and stored during the construction of the prior knowledge base. It reflects the dispersion of the output response values ​​of multiple samples within the same process parameter range at the same temperature and current point combination. For each temperature point, all standard deviations across the entire current range need to be synthesized, for example, by calculating the average standard deviation of all current points at that temperature point or taking the maximum standard deviation, as a representative value of the statistical dispersion of calibration parameters at that temperature point. Then, this representative value is compared point-by-point with a preset dispersion threshold. The preset dispersion threshold is an empirical value pre-set based on the overall process stability and sensor accuracy requirements of the current batch, usually a multiple of the average dispersion of all temperature points in the batch or directly specified by engineering experience. When the statistical dispersion at a certain temperature point is greater than the preset dispersion threshold, it indicates that the individual variability of the nanocrystalline magnetic core at that temperature point is significantly higher than the general level, and this temperature point is a point with high calibration uncertainty. All marked high-uncertainty temperature points are clustered and integrated according to temperature coordinates. Consecutive high-uncertainty temperature points are grouped into one temperature range, thus forming one or more high-uncertainty temperature ranges. These ranges typically occur near -40°C at the low-temperature end and above 100°C at the high-temperature end, because the magnetic properties of nanocrystalline magnetic cores are significantly more sensitive to process fluctuations under these extreme temperature conditions. When determining the strongly nonlinear current range, the nonlinear characteristic index of the response curve at each current point within this process parameter range is extracted from the batch prior knowledge base. The response curve refers to the mapping relationship between the input current and the output response across the entire temperature range at that current point. The specific calculation method for the nonlinear characteristic index can be the rate of change of the first derivative or the absolute value of the second derivative of the response curve near that current point, or the percentage deviation between the actual response curve and the ideal straight line at that current point. The larger the nonlinear characteristic index, the more severe the deviation of the sensor's input-output relationship from linearity at that current point, and the greater the difficulty of calibration and compensation. The nonlinear characteristic index of each current point is compared with a preset nonlinear threshold. The preset nonlinear threshold is usually determined according to the linearity requirements in the sensor design specifications. For example, if the linearity of the sensor is required to be better than 0.5% over the entire range, the current point corresponding to the nonlinear characteristic index exceeding 0.5% is marked as a point of excessive nonlinearity. All marked strong nonlinear current points are clustered and integrated according to the current coordinates. Continuously adjacent strong nonlinear current points are merged into a current interval, thus forming a strong nonlinear current interval. These intervals usually correspond to the initial permeability sensitive area of ​​the small current segment and the large current segment close to saturation.For example, in a batch prior knowledge base containing twenty temperature points and twenty current points, after the above comparison, three high uncertainty temperature ranges may be marked, namely -40 to -30 degrees Celsius, 70 to 85 degrees Celsius, and 110 to 125 degrees Celsius, as well as two strongly nonlinear current ranges, namely 0% to 10% of the rated current and 80% to 100% of the rated current. After these ranges are marked, subsequent test point combinations will be preferentially selected from these marked ranges, thereby achieving coverage of all critical areas with high uncertainty and strong nonlinearity with the fewest test points.

[0035] Furthermore, this application S400 includes: The common features in the batch prior knowledge base are used as the first layer input, and the common feature model includes the statistical laws of response characteristics across the entire temperature range. The actual calibration data is used as the second layer input. Based on the common features of the first layer input, the initial parameter range of the target sensor is determined. Within the initial parameter range, the initial parameters are individually adjusted based on the actual response data of the second layer input to obtain the posterior distribution of the target sensor calibration parameters. Based on the posterior distribution, the predicted output values ​​and corresponding confidence intervals of the target sensor at any temperature and any current point across the entire temperature range are generated, forming the dedicated calibration model.

[0036] Specifically, in the execution of hierarchical Bayesian inference, the common features in the batch prior knowledge base are first used as the first layer input. The common feature model contains the statistical laws of response characteristics under all combinations of temperature and current points across the entire temperature range. Specifically, this is reflected in the median and standard deviation data of each test point calculated in the previous steps. The median represents the statistically typical response level of the batch of products, while the standard deviation quantifies the degree of dispersion between individuals at that point. Together, they constitute a prior response surface covering the entire temperature range. The actual calibration data obtained by the target sensor through an active sampling strategy is used as the second layer input. Although this data only covers sparse test points, each data point comes from the key region with the highest prior uncertainty or the strongest nonlinearity, thus possessing sufficient information to characterize the individual characteristics of the target sensor. Subsequently, the initial parameter range of the target sensor is determined based on the common features of the first layer input. The initial parameter range is not a fixed narrow band interval, but a feasible region that dynamically changes with different temperature and current points: in regions with small prior standard deviations and good individual consistency, the initial parameter range is relatively narrow, indicating that the region is less affected by process fluctuations and the prior information is highly reliable; in regions with large prior standard deviations and severe individual dispersion, the initial parameter range is correspondingly widened, indicating that the individual differences in the region are significant and sufficient adjustment space needs to be reserved for subsequent individualized correction. After determining the initial parameter range, individual adjustments are made based on the actual response data of the second-layer input: the measured output of the target sensor at each sparse test point is compared with the prior median value, and the direction and magnitude of the deviation are calculated. Then, using this deviation information as a traction force, the calibration parameters are deflected from the prior median value position to the actual measured value direction within the initial parameter range. The degree of deflection is determined by the prior standard deviation of the test point. A large standard deviation means that the prior constraint in this area is weak, allowing for large corrections; a small standard deviation means that the prior constraint is strong, allowing only small corrections. For areas without measured data, the correction amount is determined by smooth interpolation through the spatial correlation with nearby sparse test points. After the above individualized adjustment process, a single fixed parameter value is no longer output. Instead, the posterior probability distribution of the target sensor calibration parameters is obtained. The posterior distribution is essentially a complete probability description of where the true value of the calibration parameter may fall after combining prior information and measured information, and what the probability of it falling at this position is. Compared with a single point estimate, the posterior distribution retains complete uncertainty information.

[0037] Based on the posterior distribution, the predicted output values ​​and corresponding confidence intervals of the target sensor at any temperature and current point across the entire temperature range are generated. For any given temperature-current combination, the probability density distribution of the predicted output value under that operating condition is calculated by sampling from the posterior distribution. The peak value is taken as the most probable predicted output value, and a probability density concentration interval is defined as the confidence interval. For example, a 95% confidence interval means that the actual output of the target sensor has a 95% probability of falling within this interval. The width of the confidence interval directly reflects the reliability of the calibration model under that operating condition. Finally, these predicted output values ​​and confidence intervals are summarized into structured data, which constitutes the dedicated calibration model for the target sensor. In a typical implementation case, if the target sensor completes the actual measurement at eleven sparse test points through active sampling, the dedicated calibration model output after hierarchical Bayesian inference fusion provides a predicted output value and a confidence interval upper and lower bound for each of the four hundred temperature-current combination points across the entire temperature range. In areas without measured data coverage, the average width of the confidence interval is narrowed by about 40% compared to the initial confidence interval in the prior stage, indicating that the combination of prior knowledge and a small amount of measured data effectively reduces the uncertainty of the full temperature range prediction. Near areas with measured data, the deviation between the predicted output value and the actual measured value in the subsequent full verification test is controlled within one-third of the sensor's accuracy index.

[0038] Furthermore, this application S400 includes: The initial parameter range constitutes the individual adjustment boundary; within the individual adjustment boundary, adjustments are made in the same direction based on the deviation direction between the actual response data of the target sensor and the common characteristics, and the deviation between the actual response data and the common characteristics is proportional to the adjustment amplitude.

[0039] Specifically, when performing individualized adjustments at the individual characteristic level, the initial parameter range constitutes the individual adjustment boundary. The upper and lower limits of this boundary are determined by the standard deviation of the test point in the batch prior knowledge base. Specifically, a feasible region is defined with the median value as the center and several multiples of the standard deviation as the radius. For example, using plus or minus two times the standard deviation as the adjustment boundary means that the adjustment of individual parameters is constrained within 95% of the batch statistical distribution range and cannot deviate outwards indefinitely. Within the individual adjustment boundary, the specific adjustment rule is: to adjust in the same direction based on the deviation direction between the actual response data of the target sensor and the common characteristics. That is, if the measured output value is higher than the prior median value, the calibration parameters are corrected upwards; conversely, if the measured output value is lower than the prior median value, the calibration parameters are corrected downwards. This logic of adjusting in the same direction ensures that the model approximates the true characteristics of the target sensor. Meanwhile, the deviation between the actual response data and the common features is proportional to the adjustment range. In other words, the greater the deviation between the measured value and the median, the larger the correction. This is because a larger deviation indicates that the prior common features at that point are significantly inaccurate in describing the target sensor, requiring stronger correction to compensate. Conversely, when the measured value is close to the median, only a small correction or almost no correction is needed. For example, if the prior median at a certain test point is a unitized output signal with a standard deviation of 5%, and the initial adjustment boundary is set within ±2 times the standard deviation (10%), and the deviation between the measured output value and the median at that point is only 2%, then a suitable correction is calculated proportionally in that direction, so that the adjusted parameter falls about 2% above the median, still far from the boundary. However, if the measured value deviates from the median by 8%, the adjustment range increases accordingly, and the corrected parameter approaches but does not exceed the 10% boundary limit. In other words, the direction of the deviation determines the positive or negative polarity of the adjustment, and the magnitude of the deviation determines the size of the adjustment. Together, they constitute a complete adjustment rule, which makes the individualized correction both directional and appropriate in magnitude. At the same time, the existence of the initial parameter boundary ensures that even if the measured data shows extreme outliers due to some accidental factors, the individualized adjustment will not be overly influenced. The stability and robustness of the entire inference process are thus reliably guaranteed.

[0040] Furthermore, this application S400 includes: The posterior distribution is used to calculate all possible output values ​​and their probabilities at each test point. The output value with the highest probability is extracted as the predicted value, and the probability density distribution interval within a preset range around the predicted value is extracted as the confidence interval. The predicted values ​​and confidence intervals of all temperature points and current points within the entire temperature range are summarized to form the dedicated calibration model.

[0041] Specifically, after obtaining the posterior distribution of the target sensor calibration parameters, the predicted values ​​and confidence intervals in the dedicated calibration model are generated using this posterior distribution. For any given temperature and current point within the entire temperature range, all possible output values ​​and their probabilities under that operating condition are calculated from the posterior distribution. After determining the predicted value, the model expands to the left and right sides from the predicted value until the cumulative probability reaches a preset confidence level, such as 95%, thus defining a probability density distribution interval. This interval is the confidence interval, which physically means that at that temperature and current point, the actual output value of the target sensor has a 95% probability of falling within this interval. The width of the interval directly reflects the degree of uncertainty in the prediction; when the measured data imposes a strong constraint on a certain area, the confidence interval is narrower, and when the constraint is weaker, the confidence interval is wider. It should be noted that the preset range is determined in advance based on the requirements for calibration reliability in actual applications. Typically, a 95% confidence level is used as the standard. However, in some applications with extremely high safety requirements, such as current monitoring in new energy vehicles, the confidence level can be increased to 99%, and the confidence interval will be correspondingly widened to accommodate higher probability coverage. After calculating the predicted value and confidence interval for a single test point, this calculation process is extended to all temperature and current points across the entire temperature range. The same probability extraction and interval delineation are performed on each point, ultimately resulting in a complete data table covering all operating conditions across the entire temperature range. Each point contains both a predicted output value and upper and lower bounds of a confidence interval. This data table constitutes the dedicated calibration model for the target sensor. In subsequent actual measurements, when the sensor operates at any arbitrary temperature and current point, the predicted value of the corresponding point in the dedicated calibration model is directly called as the measurement benchmark, and the width of the confidence interval is used as the error evaluation reference value for the measurement result. This ensures that every measurement result across the entire temperature range has a clear reliability indicator.

[0042] In summary, the calibration method for nanocrystalline magnetic core current sensors for applications across the entire temperature range provided in this application has the following technical advantages: By combining batch prior knowledge base constraints with a small amount of active sampling and measurement, and by using hierarchical Bayesian inference to gradually integrate batch common features with individual measured data, a unique calibration model adapted to the individual characteristics of the target sensor can be generated under the condition of a small amount of measured data. This overcomes the defect that a unified model cannot adapt to individual differences and achieves the technical effect of improving calibration efficiency and individual adaptation accuracy.

[0043] Example 2: Based on the same inventive concept as the nanocrystalline magnetic core current sensor calibration method for full-temperature applications in Example 1, this application also provides a nanocrystalline magnetic core current sensor calibration system for full-temperature applications. Please refer to the appendix. Figure 3The nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range includes: The first test module 11 is used to select samples and corresponding heat treatment process parameters from the batch of nanocrystalline magnetic core current sensors, perform complete calibration tests on the samples in the full temperature range and the full current range, and obtain complete calibration data of each sample at each temperature point and each current point.

[0044] The correlation analysis module 12 is used to perform correlation analysis between the complete calibration data of each sample and the heat treatment process parameters to construct a batch prior knowledge base.

[0045] The second testing module 13 is used to determine several test points for the target sensor to be calibrated based on the batch prior knowledge base through an active sampling strategy, and to perform actual calibration tests on the target sensor to obtain actual calibration data of the target sensor under sparse test points.

[0046] The fusion module 14 is used to take the common features in the batch prior knowledge base as the first layer of prior constraints, take the actual calibration data as observation evidence, and use the hierarchical Bayesian inference method to fuse the prior constraints and observation evidence to generate a dedicated calibration model of the target sensor in the full temperature range.

[0047] Furthermore, the first test module 11 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: Obtain the preset application temperature range of the nanocrystalline magnetic core current sensor; conduct point-by-point tests on each sample within the full temperature range of the preset application temperature range at predetermined temperature intervals, and conduct point-by-point tests at predetermined current intervals within the rated current range at each temperature point, record the output response value of each sample at each test point, form complete calibration data for each sample, and simultaneously record the heat treatment process parameters of each sample during the preparation of the nanocrystalline magnetic core.

[0048] Furthermore, the correlation analysis module 12 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: The samples were grouped according to the numerical range of the heat treatment process parameters. The full-range calibration dataset within the same process parameter range was statistically summarized, and the statistical characteristics of the output response values ​​at each temperature and current point were calculated. Using the process parameter range as the index and the statistical characteristics of each test point as the content, a batch prior knowledge base between process parameters and calibration characteristics was constructed.

[0049] Furthermore, the correlation analysis module 12 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: The statistical characteristics include a central tendency and a dispersion measure; wherein, the central tendency measure is the median of the output response values ​​of a set of samples at the same test point, and the dispersion measure is the standard deviation of the output response values ​​of a set of samples at the same test point.

[0050] Furthermore, the second test module 13 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: Based on the batch prior knowledge base, a high uncertainty temperature range and a strong nonlinear current range are obtained; the high uncertainty temperature range and the strong nonlinear current range are combined to form several test points; based on the several test points, the known current value output by the standard current source is applied to the target sensor in sequence, and after the ambient temperature of the target sensor stabilizes at the corresponding temperature point, the actual output response data of the target sensor is recorded to obtain the actual calibration data.

[0051] Furthermore, the second test module 13 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: Based on the batch prior knowledge base, the statistical dispersion of calibration parameters at each temperature point is extracted, and several temperature regions with statistical dispersion greater than a preset dispersion threshold are marked as high uncertainty temperature intervals; based on the batch prior knowledge base, the nonlinear characteristic index of the response curve at each current point is extracted, and several current regions with nonlinear characteristic index greater than a preset nonlinear threshold are marked as strong nonlinear current intervals.

[0052] Furthermore, the fusion module 14 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: The common features in the batch prior knowledge base are used as the first layer input, and the common feature model includes the statistical laws of response characteristics across the entire temperature range. The actual calibration data is used as the second layer input. Based on the common features of the first layer input, the initial parameter range of the target sensor is determined. Within the initial parameter range, the initial parameters are individually adjusted based on the actual response data of the second layer input to obtain the posterior distribution of the target sensor calibration parameters. Based on the posterior distribution, the predicted output values ​​and corresponding confidence intervals of the target sensor at any temperature and any current point across the entire temperature range are generated, forming the dedicated calibration model.

[0053] Furthermore, the fusion module 14 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: The initial parameter range constitutes the individual adjustment boundary; within the individual adjustment boundary, adjustments are made in the same direction based on the deviation direction between the actual response data of the target sensor and the common characteristics, and the deviation between the actual response data and the common characteristics is proportional to the adjustment magnitude.

[0054] Furthermore, the fusion module 14 in the nanocrystalline magnetic core current sensor calibration system for applications across the entire temperature range is also used for: The posterior distribution is used to calculate all possible output values ​​and their probabilities at each test point. The output value with the highest probability is extracted as the predicted value, and the probability density distribution interval within a preset range around the predicted value is extracted as the confidence interval. The predicted values ​​and confidence intervals of all temperature points and current points within the entire temperature range are summarized to form the dedicated calibration model.

[0055] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The calibration method and specific examples of the nanocrystalline magnetic core current sensor for full-temperature applications in the foregoing embodiment 1 are also applicable to the calibration system of the nanocrystalline magnetic core current sensor for full-temperature applications in this embodiment. Through the foregoing detailed description of the calibration method of the nanocrystalline magnetic core current sensor for full-temperature applications, those skilled in the art can clearly understand the calibration system of the nanocrystalline magnetic core current sensor for full-temperature applications in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0056] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0057] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A calibration method for a nanocrystalline magnetic core current sensor for applications across the entire temperature range, characterized in that, include: Samples and corresponding heat treatment process parameters were selected from batches of nanocrystalline magnetic core current sensors. The samples were fully calibrated in the full temperature range and full current range to obtain complete calibration data for each sample at each temperature and current point. The complete calibration data of each sample are correlated with the heat treatment process parameters to construct a batch prior knowledge base. For the target sensor to be calibrated, based on the batch prior knowledge base, several test points are determined through an active sampling strategy, and the target sensor is actually calibrated to obtain the actual calibration data of the target sensor under sparse test points. The common features in the batch prior knowledge base are used as the first layer of prior constraints, and the actual calibration data are used as observation evidence. The hierarchical Bayesian inference method is used to fuse the prior constraints and observation evidence to generate a dedicated calibration model for the target sensor in the full temperature range.

2. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 1, characterized in that, Samples and corresponding heat treatment parameters were selected from the batch of nanocrystalline magnetic core current sensors. Complete calibration tests were performed on the samples across the entire temperature and current range to obtain complete calibration data for each sample at each temperature and current point, including: Obtain the preset application temperature range for the nanocrystalline magnetic core current sensor; Each sample was tested point by point at predetermined temperature intervals within the full temperature range of the preset application temperature zone. At each temperature point, the sample was tested point by point at predetermined current intervals within the rated current range. The output response value of each sample at each test point was recorded to form complete calibration data for each sample. The heat treatment process parameters of each sample during the preparation of the nanocrystalline magnetic core were also recorded simultaneously.

3. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 1, characterized in that, The complete calibration data of each sample are correlated with the heat treatment process parameters to construct a batch prior knowledge base, including: The samples were grouped according to the numerical range of the heat treatment process parameters. The full-range calibration dataset within the same process parameter range was statistically summarized, and the statistical characteristics of the output response value at each temperature point and each current point were calculated. Using process parameter ranges as indexes and statistical characteristics of each test point as content, a batch prior knowledge base is constructed between process parameters and calibration characteristics.

4. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 3, characterized in that, The statistical characteristics include central tendency and dispersion. The central tendency measure is the median of the output response values ​​of a set of samples at the same test point, and the dispersion measure is the standard deviation of the output response values ​​of a set of samples at the same test point.

5. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 1, characterized in that, For the target sensor to be calibrated, based on the batch prior knowledge base, several test points are determined through an active sampling strategy. Actual calibration tests are then performed on the target sensor to obtain actual calibration data of the target sensor at sparse test points, including: Based on the batch prior knowledge base, the high uncertainty temperature range and the strong nonlinear current range are obtained. The high uncertainty temperature range is combined with the strong nonlinear current range to form several test points; Based on the aforementioned test points, the known current values ​​output by the standard current source are sequentially applied to the target sensor, and after the ambient temperature of the target sensor stabilizes at the corresponding temperature point, the actual output response data of the target sensor is recorded to obtain the actual calibration data.

6. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 5, characterized in that, Based on the aforementioned batch prior knowledge base, the high uncertainty temperature range and the strong nonlinear current range are obtained, including: Based on the batch prior knowledge base, the statistical dispersion of calibration parameters at each temperature point is extracted, and several temperature regions with statistical dispersion greater than a preset dispersion threshold are marked as high uncertainty temperature ranges. Based on the batch prior knowledge base, nonlinear characteristic indicators of the response curves at each current point are extracted, and several current regions with nonlinear characteristic indicators greater than a preset nonlinear threshold are marked as strong nonlinear current intervals.

7. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 1, characterized in that, The common features in the batch prior knowledge base are used as the first layer of prior constraints, and the actual calibration data are used as observational evidence. A hierarchical Bayesian inference method is used to fuse the prior constraints and observational evidence to generate a dedicated calibration model for the target sensor across the entire temperature range, including: The common features in the batch prior knowledge base are used as the first layer input, and the common feature model includes the statistical regularity of response characteristics in the entire temperature range; Use the actual calibration data as the second layer input; Based on the common features of the first layer input, the initial parameter range of the target sensor is determined; Within the initial parameter range, the initial parameters are individually adjusted based on the actual response data of the second layer input to obtain the posterior distribution of the target sensor calibration parameters; Based on the posterior distribution, the predicted output values ​​and corresponding confidence intervals of the target sensor at any temperature and current point within the full temperature range are generated, forming the dedicated calibration model.

8. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 7, characterized in that, When making individual adjustments to the initial parameters, the adjustment range is constrained by the range of the initial parameters, and the specific rules are as follows: The initial parameter range constitutes the individual adjustment boundary; Within the individual adjustment boundary, adjustments are made in the same direction based on the deviation between the actual response data of the target sensor and the common characteristics. The deviation between the actual response data and the common characteristics is proportional to the adjustment range.

9. The calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications as described in claim 7, characterized in that, Based on the posterior distribution, predicted output values ​​and corresponding confidence intervals of the target sensor at any temperature and current point across the entire temperature range are generated, constituting the dedicated calibration model, including: The posterior distribution is used to calculate all possible output values ​​and their probabilities at each test point. The output value with the highest probability is extracted as the predicted value, and the probability density distribution interval within a preset range around the predicted value is extracted as the confidence interval. The predicted values ​​and confidence intervals of all temperature and current points within the entire temperature range are summarized to form the dedicated calibration model.

10. A calibration system for a nanocrystalline magnetic core current sensor for applications across the entire temperature range, characterized in that, The steps for implementing the calibration method for a nanocrystalline magnetic core current sensor for full-temperature applications according to any one of claims 1 to 9, wherein the nanocrystalline magnetic core current sensor calibration system for full-temperature applications comprises: The first testing module is used to select samples and corresponding heat treatment process parameters from the batch of nanocrystalline magnetic core current sensors, perform complete calibration tests on the samples in the full temperature range and the full current range, and obtain complete calibration data of each sample at each temperature point and each current point. The correlation analysis module is used to perform correlation analysis between the complete calibration data of each sample and the heat treatment process parameters to build a batch prior knowledge base; The second testing module is used to determine several test points for the target sensor to be calibrated based on the batch prior knowledge base and through an active sampling strategy, and to perform actual calibration tests on the target sensor to obtain actual calibration data of the target sensor under sparse test points. The fusion module is used to take the common features in the batch prior knowledge base as the first layer of prior constraints, take the actual calibration data as observation evidence, and use the hierarchical Bayesian inference method to fuse the prior constraints and observation evidence to generate a dedicated calibration model for the target sensor in the full temperature range.

Images