A sensor calibration method, device, equipment, readable storage medium and product

Abstract

The application discloses a kind of sensor calibration method, device, equipment, readable storage medium and product, applied to calibration field, comprising: obtaining the full-condition data of sensor to be calibrated;Full-condition data at least includes a group of corresponding data, each group of corresponding data includes reference gap, working temperature, measured material type and the output voltage of sensor to be calibrated;Based on full-condition data, the mapping relationship between output voltage, working temperature, measured material type and reference gap is established;When mapping relationship is established and satisfies preset precision condition, sensor to be calibrated calibration ends;Wherein, reference gap is obtained by reference sensor, sensor to be calibrated and reference sensor are in the same gap position, same measured material and same working temperature environment.The present application significantly improves the calibration efficiency and detection accuracy, and can meet the high-precision and high-reliability requirements of high-speed maglev train gap detection.

Description

Technical Field

[0001] This invention relates to the field of sensor calibration, and in particular to a sensor calibration method, apparatus, device, readable storage medium, and product. Background Technology

[0002] The levitation and guidance sensor (hereinafter referred to as the sensor) is a core component of the levitation and guidance system of high-speed maglev trains and a key device for achieving reliable gap detection. Because the sensor's output characteristics are easily affected by manufacturing processes and temperature drift, accurate detection relationships need to be established through calibration. However, current sensor calibration still relies on manual calibration, which is cumbersome, time-consuming, and susceptible to human error, leading to calibration failures.

[0003] Therefore, how to achieve automated, high-precision, and high-efficiency sensor calibration under multi-factor coupling is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a sensor calibration method, apparatus, device, readable storage medium and product, which solves the problems of low calibration efficiency and poor accuracy under multi-factor coupling in the prior art.

[0005] To address the aforementioned technical problems, this invention provides a sensor calibration method, comprising: Acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, each set of corresponding data including reference gap, operating temperature, type of material being measured and output voltage of the sensor to be calibrated; Based on the full operating condition data, a mapping relationship is established between output voltage, operating temperature, type of material under test and reference gap; When the mapping relationship is established and the preset accuracy condition is met, the calibration of the sensor to be calibrated ends. The reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are located at the same gap position, the same measured material, and the same operating temperature environment.

[0006] Optionally, acquire full-condition data of the sensor to be calibrated, including: The calibration process is started according to the preset calibration parameters. The working temperature of the temperature control environment is controlled by the neural network temperature controller, and the gap displacement between the reference sensor and the material being measured is adjusted by the neural network motion controller to form the reference gap. Relevant data are collected in real time during the calibration process; the relevant data includes reference gap, operating temperature, output voltage, and the type of material being tested; After traversing all preset test material types and operating temperature points, all relevant data collected will be combined to form the full operating condition data.

[0007] Optionally, the operating temperature of the temperature-controlled environment is controlled by a neural network temperature controller, including: The temperature sequence of the current period and the past N periods of the measurement and control environment and the current driving duty cycle of the neural network temperature controller are input into the temperature prediction model to obtain the predicted temperature value of the measurement and control environment. The temperature deviation and the rate of change of the predicted temperature value and the target temperature are calculated. Based on the preset fuzzy rule base and membership function, fuzzy inference is performed on the temperature deviation and the rate of change of the temperature deviation to obtain the fuzzy value of the control change amount. The fuzzy value of the control change is defuzzified to obtain the precise value of the control change of the drive duty cycle; The drive duty cycle is updated based on the precise value of the control change, and the updated drive duty cycle is converted into a pulse width modulation signal and output to the semiconductor cooler to realize closed-loop temperature control of the temperature control environment.

[0008] Optionally, the reference gap is formed by adjusting the gap displacement between the reference sensor and the measured material using a neural network motion controller, including: Obtain the state vector within the current control cycle; the state vector includes the position error of the translation stage, the rate of change of the position error, the control quantity at the previous moment, the target gap position, and the ambient temperature. The state vector is input into the decision network to obtain the action vector. The controller parameters are updated based on the action vector, and the current control quantity is calculated. The current control quantity is output to the servo motor driver through the motion control card to drive the translation stage to move, thereby changing the gap displacement between the reference sensor and the measured material to form the reference gap; The reference sensor is mounted on the translation stage.

[0009] Optionally, after outputting the current control quantity to the servo motor driver via the motion control card to drive the translation stage to move, thereby changing the gap displacement between the reference sensor and the measured material to form the reference gap, the method further includes: Collect the next state vector and calculate the reward value based on the next state vector; The state vector, the action vector, the reward value, and the next state vector are stored as a set of samples in the experience replay pool; When the update conditions are met, samples are collected in batches from the experience replay pool; The evaluation network is updated based on the collected samples, and the decision network is updated using the policy gradient method based on the action value output by the updated evaluation network.

[0010] Optionally, before starting the calibration process according to preset calibration parameters, controlling the operating temperature of the temperature-controlled environment through a neural network temperature controller, and adjusting the gap displacement between the reference sensor and the measured material through a neural network motion controller to form the reference gap, the process further includes: By performing preliminary scans of the sensor to be calibrated and the reference sensor at different gaps, the original response curves of the sensors are obtained. The original response curve of the sensor is subjected to nonlinear analysis to determine the preset calibration parameters.

[0011] Optionally, by performing preliminary scans of the sensor to be calibrated and the reference sensor at different gaps, the original response curves of the sensors are obtained, including: The sensor to be calibrated and the reference sensor are placed in a target operating temperature and a single test material environment. The neural network motion controller drives the sensor to be calibrated and the reference sensor to perform a unidirectional linear scan with a preset step size along the zero gap to the working gap range, and simultaneously collects the original reference gap value and the original voltage value to form the original response curve of the sensor.

[0012] Optionally, nonlinear analysis is performed on the original response curve of the sensor to determine the preset calibration parameters, including: A pre-trained calibration parameter recommendation model is obtained; the calibration parameter recommendation model is trained based on a paired dataset of the sensor's nonlinear characteristics and the optimal calibration parameters. The original response curve of the sensor is input into the calibration parameter recommendation model to extract the nonlinear features of the original response curve of the sensor using the calibration parameter recommendation model, and the preset calibration parameters are output based on the nonlinear features.

[0013] Optionally, the original sensor response curve is input into the calibration parameter recommendation model to extract the nonlinear features of the original sensor response curve using the calibration parameter recommendation model, and the preset calibration parameters are output based on the nonlinear features, including: The original response curve of the sensor is normalized and interpolated to generate a processed curve; The nonlinear features are obtained by extracting features from the processed curve through the convolutional and pooling layers of the calibration parameter recommendation model. The nonlinear features are processed by the fully connected layer of the calibration parameter recommendation model to obtain the preset calibration parameters.

[0014] Optionally, the nonlinear features include at least one or more of the following: curve rate variation features, curve smoothness features, and saturation region shape features.

[0015] Optionally, based on the full-condition data, a mapping relationship is established between the output voltage, operating temperature, type of material under test, and reference gap, including: Based on the output voltage, operating temperature, type of material under test and corresponding reference gap, the initial calibration model is trained to obtain a trained calibration model. The trained calibration model can characterize the nonlinear mapping relationship between the output voltage, operating temperature, type of material being measured, and reference gap of the sensor to be calibrated.

[0016] Optionally, based on the output voltage, operating temperature, type of material under test, and corresponding reference gap, the initial calibration model is trained to obtain a trained calibration model, including: A training dataset is constructed based on the full-condition data. Each sample in the training dataset includes output voltage, operating temperature, type of material under test, and reference gap. The training dataset is divided into a training set and a validation set, and the samples in the training set are input into the initial calibration model. The mean squared error is used as the loss function, and the parameters of the initial calibration model are updated by backpropagation using an adaptive moment estimation optimizer. When the first preset condition is met, the accuracy of the initial calibration model is verified using samples in the validation set. When the second preset condition is met, training is stopped, and the trained calibration model is obtained.

[0017] Optionally, after obtaining the trained calibration model, it may also include: The current operating temperature, current material type, and current output voltage collected by the sensor to be calibrated are input into the trained calibration model, and the corresponding current gap is output.

[0018] Optionally, the preset calibration parameters include at least one or more of the following: a temperature point list, a gap sampling interval list, motion control parameters, temperature control parameters, data acquisition parameters, and material calibration priority parameters.

[0019] Optionally, after acquiring relevant data in real time during the calibration process, the method further includes: The relevant data is subjected to quality assessment, and the preset calibration parameters for subsequent calibration are adjusted in real time based on the quality assessment results; Data is collected based on the adjusted preset calibration parameters.

[0020] Optionally, when the mapping relationship is established and a preset accuracy condition is met, the calibration of the sensor to be calibrated ends, including: When the mapping relationship is established and the allowable error and preset fit are satisfied, the calibration of the sensor to be calibrated is completed.

[0021] The present invention also provides a sensor calibration device, comprising: The full-condition data acquisition module is used to acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, and each set of corresponding data includes the reference gap, operating temperature, type of material being measured and the output voltage of the sensor to be calibrated; The mapping relationship establishment module is used to establish a mapping relationship between output voltage, operating temperature, type of material under test and reference gap based on the full operating condition data; The calibration end module is used to end the calibration of the sensor to be calibrated when the mapping relationship is established and the preset accuracy condition is met. The reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are located at the same gap position, the same measured material, and the same operating temperature environment.

[0022] The present invention also provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to implement the frequency calibration method described above when executing the computer program.

[0023] The present invention also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the frequency calibration method described above.

[0024] The present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the frequency calibration method described above.

[0025] As can be seen from the above technical solution, the present invention acquires full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, each set of corresponding data including reference gap, operating temperature, type of material being measured, and output voltage of the sensor to be calibrated; based on the full-condition data, a mapping relationship is established between output voltage, operating temperature, type of material being measured, and reference gap; when the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated ends; wherein, the reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are in the same gap position, the same material being measured, and the same operating temperature environment. The beneficial effects of this invention are as follows: By synchronously collecting full-condition data of the sensor to be calibrated and the reference sensor at the same gap position, the same measured material, and the same working temperature environment, this invention establishes a high-precision mapping relationship between output voltage, working temperature, measured material type, and reference gap. This effectively solves the problems of cumbersome, inefficient, and easily affected by human factors in traditional manual calibration processes, as well as the inability to take into account the coupled effects of multiple factors such as temperature drift and material characteristics, resulting in insufficient calibration accuracy. This invention achieves automated, high-precision, and high-stability calibration of the suspension guide sensor, significantly improving calibration efficiency and detection accuracy, and can meet the high-precision and high-reliability requirements of high-speed maglev trains for gap detection.

[0026] In addition, the present invention also provides a sensor calibration device, apparatus, readable storage medium, and product, which also have the above-mentioned beneficial effects. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0028] Figure 1 A flowchart of a sensor calibration method provided in an embodiment of the present invention; Figure 2 A flowchart of another sensor calibration method provided in an embodiment of the present invention; Figure 3 A flowchart illustrating another sensor calibration method provided in an embodiment of the present invention; Figure 4 A flowchart illustrating a sensor calibration method provided in an embodiment of the present invention; Figure 5 A flowchart illustrating a sensor calibration architecture provided in an embodiment of the present invention; Figure 6This is a schematic diagram of a sensor calibration device provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of a sensor calibration device provided in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The suspension and guidance sensors, sub-components of the suspension and guidance system of high-speed maglev trains, are key devices for reliable gap detection. The suspension sensor measures the suspension gap, i.e., the vertical distance between the car body and the track; the guidance sensor measures the guidance gap, i.e., the lateral distance between the car body and the track side. Currently, the calibration of these sensors requires manual, repetitive operations, a complex and lengthy process susceptible to calibration failures due to human error, thus increasing manufacturing costs and impacting production cycles and delivery schedules. Therefore, reducing human intervention, achieving automated sensor calibration, and improving the efficiency and accuracy of sensor calibration are urgent technical challenges that need to be addressed.

[0031] To address the above problems, the present invention provides a sensor calibration method, the specific of which can be found in the following embodiments: Example 1: Please refer to Figure 1 , Figure 1 A flowchart illustrating a sensor calibration method provided in an embodiment of the present invention. The method may include: S101: Acquire full-condition data of the sensor to be calibrated.

[0032] Each step in this embodiment can be executed by a designated electronic device, which can be a server, a portable terminal, or other forms. It should be noted that the full-condition data in this embodiment can specifically include at least one set of corresponding data, each set containing a reference gap, operating temperature, type of material being measured, and the output voltage of the sensor to be calibrated. The sensor to be calibrated in this embodiment refers to a gap sensor. For example, it can be a suspension sensor or a guiding sensor. The method for acquiring the full-condition data in this embodiment is limited. The reference gap in this embodiment is the gap value measured by the reference sensor and used as a true reference. It is understood that the accuracy of the reference sensor directly affects the calibration accuracy; therefore, a high-precision reference sensor can be selected. Furthermore, the sensor to be calibrated and the reference sensor are located at the same gap position, using the same material being measured, and under the same operating temperature environment to reduce variable interference. This embodiment does not specifically limit the acquisition method of the full-condition data. For example, step-by-step traversal acquisition can be used; adaptive dynamic acquisition can also be used; or it can be any one or a combination of material priority-based sequential acquisition, temperature range-based segmented acquisition, or data quality-based closed-loop acquisition.

[0033] S102: Based on full-condition data, establish a mapping relationship between output voltage, operating temperature, type of material under test, and reference gap.

[0034] In this embodiment, based on the full-condition data acquired in step S101, a mapping relationship can be established between {output voltage, operating temperature, and type of material under test} and {reference gap}. This embodiment does not limit the specific form of the mapping relationship. For example, it can be a table, such as a mapping table; or it can be a neural network; or it can be a polynomial function; or it can be a piecewise linear function.

[0035] S103: When the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated ends.

[0036] In this embodiment, the preset accuracy condition can be achieving a preset degree of fit, or the error being within the allowable range; or it can be achieving a preset degree of fit and the error being within the allowable range. Furthermore, by simultaneously setting the preset degree of fit and the allowable error range, this embodiment can ensure the accuracy of the mapping relationship from both the overall fitting effect and the single-point measurement accuracy dimensions, thereby achieving automation of the calibration process, qualification of calibration results, and reproducibility.

[0037] As can be seen, the sensor calibration method provided in this embodiment of the invention involves: S101: acquiring full-condition data of the sensor to be calibrated; S102: establishing a mapping relationship between output voltage, operating temperature, type of material being measured, and reference gap based on the full-condition data; S103: when the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated ends. This method establishes a high-precision mapping relationship between output voltage, operating temperature, type of material being measured, and reference gap by simultaneously acquiring full-condition data of the sensor to be calibrated and the reference sensor under the same gap position, the same material being measured, and the same operating temperature environment. This effectively solves the problems of cumbersome, inefficient, and easily affected by human factors in traditional manual calibration processes, as well as the inability to simultaneously account for the coupling effects of multiple factors such as temperature drift and material characteristics, resulting in insufficient calibration accuracy. It achieves automated, high-precision, and high-stability calibration of levitation guidance sensors, significantly improving calibration efficiency and detection accuracy, and meeting the high-precision and high-reliability requirements of high-speed maglev trains for gap detection.

[0038] Example 2: Please refer to Figure 2 , Figure 2 A flowchart illustrating another sensor calibration method provided in an embodiment of the present invention. The method may include: S201: Obtain full-condition data of the sensor to be calibrated.

[0039] This embodiment does not elaborate on this step in detail. For details, please refer to step S101 in embodiment 1.

[0040] S202: Based on the output voltage, operating temperature, type of material under test, and corresponding reference gap, the initial calibration model is trained to obtain a trained calibration model.

[0041] In this embodiment, the trained calibration model can characterize the nonlinear mapping relationship between the output voltage, operating temperature, and measured material type of the sensor to be calibrated, and the reference gap. This step involves using the output voltage, operating temperature, and measured material type of the sensor to be calibrated as model inputs, and the reference gap measured by the reference sensor as the desired output, to train the initial calibration model. This allows the model to adaptively fit the nonlinear characteristics of the sensor under the coupling of multiple physical quantities, ultimately obtaining a calibration model capable of high-precision gap inference. This embodiment does not specifically limit the network structure of the calibration model. For example, a neural network, a deep learning network, or a machine learning network can be used.

[0042] Furthermore, the initial calibration model is trained based on the output voltage, operating temperature, type of material under test, and corresponding reference gap to obtain a trained calibration model. Specifically, this may include: constructing a training dataset based on full-condition data, where each sample in the training dataset includes output voltage, operating temperature, type of material under test, and reference gap; dividing the training dataset into a training set and a validation set, and inputting the samples in the training set into the initial calibration model, using the mean squared error as the loss function, and employing an adaptive moment estimation optimizer to backpropagate and update the parameters of the initial calibration model; when the first preset condition is met, the accuracy of the initial calibration model is verified using samples in the validation set; when the second preset condition is met, training is stopped, and a trained calibration model is obtained.

[0043] It should be noted that this embodiment does not specifically limit the first and second preset conditions, and users can set them according to their actual situation. For example, the first preset condition may be that the number of iterations reaches a preset number of iterations; or it may be that the model loss function converges to a preset threshold; or it may be that the training time reaches a preset time; or it may be that all of the above conditions are met simultaneously. The second preset condition may be that the model validation accuracy meets a preset accuracy index; or it may be that the model performance does not improve significantly after multiple consecutive iterations; or it may be that the model error is less than a preset error threshold; or it may be that all of the above conditions are met simultaneously.

[0044] Because sensor output exhibits nonlinearity, temperature drift, material dependence, and multi-factor coupling errors, it is necessary to correct the original output by establishing an error compensation mapping relationship. Directly compensating for variables such as temperature and material using individual factor corrections ignores the complex relationships of mutual coupling and cross-influence among multiple physical quantities in practical applications, resulting in limited compensation effectiveness and difficulty in meeting high-precision measurement requirements. Therefore, this embodiment uses a deep neural network calibration model to directly establish a mapping from the original input to the gap value. The implementation steps of the calibration model for multi-factor compensation are as follows: (1) Input layer: includes the output voltage V (normalized) of the sensor to be calibrated, the operating temperature T, the type of the material to be measured M (material code, one-hot encoding, which converts the classification label into a vector with only one bit as 1 and the rest as 0, for easy processing by the neural network), for example, 4-dimensional for 4 types of materials, and the ambient humidity H (optional). These inputs constitute a multi-dimensional feature vector.

[0045] (2) Hidden layers: Designed as 4-6 fully connected layers, with the number of neurons in each layer decreasing (e.g., 128-64-32-16). The ReLU (Rectified Linear Unit) activation function is used, and Dropout is added to the intermediate layers to randomly disable some neurons during training, thus preventing overfitting. The ReLU activation function has the following characteristics: (1) It alleviates the gradient vanishing problem: the gradient is always 1 in the positive interval, which can effectively alleviate the gradient vanishing problem in deep networks and is beneficial for training deep models; (2) It is simple and efficient to calculate: it only requires threshold judgment and taking the maximum value, with small computational load and fast training and inference speed; (3) It has sparse output: it will cause some neurons to output 0, resulting in sparse activation, reducing redundant features and improving the generalization ability of the model; (4) It has a faster convergence speed: compared with other activation functions, the ReLU activation function can usually make the model converge faster. It can effectively alleviate the gradient vanishing problem in deep networks and has the advantages of simple computation, fast convergence speed, and good output sparsity, which is conducive to improving model training efficiency and feature representation ability.

[0046] (3) Output layer: one neuron, which outputs the predicted real gap value d_pred (which becomes the actual physical gap after inverse normalization).

[0047] (4) Training process: Training is performed using data collected during full-condition calibration. Each sample in the dataset is: (V, T, M, d_true), where d_true is the reference gap measured by the reference sensor.

[0048] During model training, the parameters of the calibration model can be updated based on the control effect (such as settling time and overshoot) of the current full-condition data, achieving self-evolution. Furthermore, it should be noted that for polynomial mapping relationships, the compensation parameters may be zero bias b, sensitivity k, nonlinear correction coefficients (such as quadratic coefficient a2, cubic coefficient a3), etc., with a relatively small number of parameters (e.g., 5-10). However, the compensation parameters in this embodiment refer to the weights and biases of the entire network of the calibration model, i.e., the parameters of the calibration model.

[0049] For example, if the network structure has a 7-dimensional input, 128-64-32 hidden layers, and a 1-dimensional output, then the number of compensation parameters is 7. 128+128+128 64+64+64 32+32+32 1+1 ≈ 10,000 parameters. These compensation parameters collectively constitute a complex multidimensional mapping function, i.e., the trained calibration model. Understandably, the compensation parameters (i.e., the parameters of the calibration model) will be continuously adjusted. For mapping relationships of polynomial form, the initial compensation parameters (e.g., factory default values) and the final compensation parameters are identical in number, differing only in value. However, for the calibration model, since the network structure is fixed, the number of compensation parameters will remain consistent, but the values ​​will be completely different after training. The initial compensation parameters are set to parameters randomly initialized or pre-trained based on historical data, while the final compensation parameters are parameters finely trained for this sensor.

[0050] During training, the mean squared error (MSE) is used as the loss function, and the Adam (adaptive moment estimator) optimizer is employed. The calibration model automatically learns the complex interactions between output voltage, operating temperature, and materials through nonlinear combinations. For example, the same voltage value corresponds to different gaps under different temperatures and materials; the calibration model can accurately distinguish these differences through nonlinear transformations of the hidden layers. In this embodiment, the mean squared error is the sum of the squares of the differences between the predicted gap value output by the initial calibration model and the reference gap value. Thus, this embodiment, based on training with full-condition data, enables the calibration model to adapt to complex working conditions involving multiple temperatures, materials, and gaps, thus broadening its applicability. Dividing the dataset into training and validation sets allows for simultaneous verification of model accuracy during training, effectively avoiding overfitting and improving model generalization ability and robustness. The training method, combining the mean squared error loss function with an adaptive moment estimation optimizer, results in fast convergence, stable training, and accurate parameter updates. Dual preset conditions control training termination, ensuring high accuracy, strong reliability, and good consistency of the output calibration model. The entire training process is highly automated, requiring no manual intervention or parameter tuning, significantly improving calibration efficiency and quality.

[0051] S203: Input the current operating temperature, current material type, and current output voltage collected by the sensor to be calibrated into the trained calibration model, and output the corresponding current gap.

[0052] The trained calibration model can be directly embedded into the calibrated sensor. During actual measurement, only the real-time acquired output voltage, operating temperature, and type of material being measured need to be input to output an accurate reference gap value, eliminating the need for step-by-step compensation. In other words, by solving and compensating for the nonlinear characteristics under the coupling of multiple physical quantities through the calibration model, a high-precision reference gap under the corresponding operating conditions can be directly output, enabling accurate sensor measurement and resolving the cross-influence of multiple external factors.

[0053] As can be seen, the sensor calibration method provided in this embodiment of the invention proceeds as follows: S201: Acquire full-condition data of the sensor to be calibrated; S202: Train an initial calibration model based on output voltage, operating temperature, type of measured material, and corresponding reference gap to obtain a trained calibration model; S203: Input the current operating temperature, current type of measured material, and current output voltage collected by the sensor to be calibrated into the trained calibration model and output the corresponding current gap. This method trains a calibration model based on full-condition data, which accurately fits the characteristics of a sensor with strong nonlinearity and multi-factor coupling; it can also automatically achieve joint compensation of multiple physical quantities such as temperature, material, and gap, and has strong generalization ability and robustness, adapting to different operating conditions and different sensors; furthermore, the calibrated sensor can be directly embedded into the system for real-time inference, realizing the integration of calibration and application; it also facilitates automated training and accuracy iteration, improving calibration efficiency and reliability.

[0054] Example 3: Please refer to Figure 3 , Figure 3 A flowchart illustrating another sensor calibration method provided in an embodiment of the present invention. The method may include: S301: Obtain the original response curve of the sensor by performing preliminary scans of the sensor to be calibrated and the reference sensor at different gaps.

[0055] This step obtains the original output characteristics of the sensor to be calibrated under different gap inputs. The original response curve of the sensor can intuitively reflect the input-output characteristics, nonlinearity, sensitivity characteristics and zero-point characteristics of the sensor to be calibrated. It is the core basis for characterizing the original measurement law and inherent nonlinear error of the sensor, and provides basic data support for subsequent nonlinear analysis, characteristic identification and calibration parameter determination.

[0056] Furthermore, the above-mentioned preliminary scanning of the sensor to be calibrated and the reference sensor at different gaps to obtain the original sensor response curve can specifically include the following steps: placing the sensor to be calibrated and the reference sensor in a target operating temperature and single measured material environment, driving the sensor to be calibrated and the reference sensor to perform a unidirectional linear scan with a preset step size from zero gap to working gap through a neural network motion controller, and simultaneously acquiring the original reference gap value and the original voltage value to form the original sensor response curve. In this embodiment, during the preliminary scan / coarse scan stage, the temperature and material are fixed (only one material is measured), ensuring that the only variable in this scan is the gap; a high-precision translation stage is used to drive the sensor to perform a unidirectional linear scan from near to far, while simultaneously acquiring the actual gap of the reference sensor and the voltage of the sensor to be measured, ultimately obtaining an accurate original sensor response curve. For example: moving the probe at a uniform speed / step-by-step rapid speed from the smallest gap to the largest gap, continuously acquiring the output voltage; traversing and acquiring data unidirectionally from the largest gap to the smallest gap; sampling point by point within the full range at a fixed step size to form the original response curve; and rapidly traversing at a lower resolution and a larger interval to quickly acquire the overall shape of the curve (slope, zero point, saturation region, noise).

[0057] S302: Perform nonlinear analysis on the original response curve of the sensor to determine the preset calibration parameters.

[0058] This embodiment does not limit the preset calibration parameters, as long as the parameters can characterize the sensor characteristics to be calibrated. This embodiment does not specifically limit the nonlinear analysis method; for example, it could be a neural network fitting analysis method, a deep neural network analysis method, or a deep learning feature analysis method. Nonlinear analysis refers to the process of quantitatively identifying the degree of nonlinearity, error distribution, and distortion characteristics of the sensor's original response curve. Preset calibration parameters refer to the control parameters in the calibration process determined based on the nonlinear analysis results.

[0059] In this embodiment, the control parameters in the preset calibration parameters can be at least one or more of the following: temperature point list, gap sampling interval list, motion control parameters, temperature control parameters, data acquisition parameters, and material calibration priority parameters: (1) Temperature point list: For example, {25, 50, 75, 100}℃ indicates the temperature points at which calibration needs to be performed. The number of temperature points is increased adaptively according to the sensor’s sensitivity to temperature.

[0060] (2) Gap sampling interval list: It is not a fixed interval, but a list. For example, in the range of 0-1mm, the sampling points are [0.1, 0.2, 0.3, 0.5, 0.7, 0.9, 1.0]mm. The sampling interval in the nonlinear interval is smaller, and the sampling interval in the linear interval is larger.

[0061] (3) Motion control parameters: Recommend the motion speed, acceleration, and relevant parameters of the neural network motion controller based on the noise level of the curve.

[0062] (4) Temperature control parameters: such as the relevant parameters of a neural network temperature controller.

[0063] (5) Data acquisition parameters: such as sampling frequency, settling time of each point, number of repeated samplings, etc.

[0064] (6) Material calibration priority parameter: If the curve shows that the sensor is sensitive to the material, it is recommended to calibrate multiple materials; otherwise, you can select only one material for calibration, which can save time; or determine the calibration priority according to the sensitivity of the material.

[0065] It is understandable that the control parameters in the preset calibration parameters will be continuously adjusted subsequently, and the number of initial control parameters may differ from the final control parameters used. This is because the parameters will be dynamically adjusted based on the quality of the real-time acquired data, such as automatically adding or removing sampling points and adjusting temperature points. However, the types of parameters (such as temperature points, sampling points, etc.) are predefined.

[0066] Furthermore, the aforementioned nonlinear analysis of the sensor's original response curve to determine the preset calibration parameters may include: obtaining a pre-trained calibration parameter recommendation model; inputting the sensor's original response curve into the calibration parameter recommendation model to extract the nonlinear features of the sensor's original response curve using the calibration parameter recommendation model, and outputting the preset calibration parameters based on the nonlinear features through parameter matching.

[0067] In this embodiment, a calibration parameter recommendation model is first trained based on a paired dataset of the sensor's nonlinear characteristics and optimal calibration parameters. The calibration parameter recommendation model is then used to extract nonlinear features and determine preset calibration parameters.

[0068] Furthermore, the above-mentioned input of the sensor's original response curve into the calibration parameter recommendation model, in order to extract the nonlinear characteristics of the sensor's original response curve using the calibration parameter recommendation model, and to output preset calibration parameters based on the nonlinear characteristics, may specifically include: Step 311: Normalize and interpolate the original sensor response curve to generate the processed curve.

[0069] Specifically, this step will initially scan the discrete points (d) i V i The data is normalized and interpolated into a fixed-length sequence (e.g., 256 points) to generate a processed curve, which serves as the input to the calibration parameter recommendation model. Where d... i V represents the gap value of the i-th sampling point.i This indicates the output voltage value of the sensor under this gap.

[0070] Step 312: Extract features from the processed curve by using the convolutional and pooling layers of the model with calibrated parameters to obtain nonlinear features.

[0071] Specifically, the calibration parameter recommendation model in this step consists of several 1D (one-dimensional) convolutional and pooling layers. It automatically extracts nonlinear features from the processed curve. These nonlinear features are a set of parameters characterizing the nonlinearity of the sensor's output response curve, including but not limited to changes in the curve's slope (nonlinearity), the smoothness of the curve (noise), and the shape of the saturation region. The nonlinear features are explained in detail below: (1) Change in curve slope (nonlinearity): refers to the degree to which the sensor output voltage changes with the gap, which varies at different locations. This uneven and inconsistent change in slope is the nonlinearity, which reflects the degree to which the sensor output deviates from the ideal straight line.

[0072] (2) Smoothness of the curve (noise): Whether the output curve is smooth, without jitter, spikes, or small fluctuations. Smooth: The curve is continuous and stable, indicating good signal quality. Unsmooth: There are up-and-down jumps, sawtooth patterns, and spikes, indicating high noise and poor stability. The smoothness directly reflects the signal-to-noise ratio and repeatability of the sensor output.

[0073] (3) Shape of the saturation region: After the gap is reduced to a certain extent, the sensor output no longer rises significantly and tends to level off or even stop changing. The shape of the saturation region includes: the inflection point where the rise becomes level off, whether the curve after saturation is straight, slightly rising or slightly falling, the width of the saturation region, and the degree of stability. It reflects the boundary of the sensor's effective measurement range and determines the limit range that the calibration cannot exceed.

[0074] Step 313: Process the nonlinear features through the fully connected layer of the calibration parameter recommendation model to obtain the preset calibration parameters.

[0075] This step flattens the extracted nonlinear features and feeds them into a fully connected layer, which outputs multiple preset calibration parameters. During calibration, the quality of the collected data is assessed, and the preset calibration parameters for subsequent calibrations are adjusted in real time based on the assessment results. The adjusted preset calibration parameters are then used to process the data. For example, if the deviation exceeds a threshold during calibration, a calibration parameter recommendation model can be triggered to reanalyze a portion of the collected data, dynamically adjusting subsequent preset calibration parameters (e.g., if particularly severe nonlinearity is found at a certain temperature point, the sampling density at that temperature point is temporarily increased).

[0076] S303: The calibration process is started according to the preset calibration parameters. The working temperature of the temperature control environment is controlled by the neural network temperature controller, and the gap displacement between the reference sensor and the measured material is adjusted by the neural network motion controller to form a reference gap.

[0077] This step refers to the automatic initiation of the calibration process according to preset calibration parameters. The neural network temperature controller maintains the specified working temperature, and the neural network motion controller precisely adjusts the gap displacement between the reference sensor and the material being measured to obtain the reference gap, providing a reliable benchmark for subsequent calibration.

[0078] Furthermore, the aforementioned control of the operating temperature of the temperature-controlled environment via a neural network temperature controller may specifically include: Step 321: Input the temperature sequence of the current period and the past N periods of the measurement and control environment and the current driving duty cycle of the neural network temperature controller into the temperature prediction model to obtain the predicted temperature value of the measurement and control environment.

[0079] In this embodiment, N is a preset positive integer, ranging from 5 to 20 control cycles. Every control cycle (e.g., 100ms), the temperature sequence [T(t-19), ..., T(t)] for the current period and a past period (e.g., the past 20 cycles), the current drive duty cycle D(t), and the ambient temperature T_env for the current cycle are read and input into the temperature prediction model. The temperature prediction model outputs the predicted temperature values ​​for the next few cycles (e.g., the next 5 cycles) [T_pred(t+1), ..., T_pred(t+5)]. This embodiment does not impose specific limitations on the network structure of the temperature prediction model. For example, the temperature prediction model can be one or more combinations of LSTM (Long Short-Term Memory) neural networks, GRU (Gated Recurrent Unit) neural networks, RNN (Recurrent Neural Network) neural networks, TCN (Temporal Convolutional Network) temporal convolutional networks, CNN (Convolutional Neural Network) convolutional neural networks, MLP (Multilayer Perceptron) neural networks, SVR (Support Vector Regression) models, or attention-based temporal prediction models.

[0080] Step 322: Calculate the temperature deviation and the rate of change of the temperature deviation between the predicted temperature and the target temperature. Based on the preset fuzzy rule base and membership function, perform fuzzy inference on the temperature deviation and the rate of change of the temperature deviation to obtain the fuzzy value of the control change.

[0081] By predicting temperature change trends using a temperature prediction model and combining this with fuzzy decision-making for control, accurate temperature prediction and control of refrigeration and heating devices (such as TECs, Thermoelectric Coolers, and semiconductor refrigerators) can be achieved, ensuring the accuracy of the calibrated sensor temperature. The specific fusion steps are as follows: Based on the temperature deviation E = T_pred(t+5)-T_target between the predicted temperature and the target temperature T_target, and the deviation change rate ΔE=(T_pred(t+5)-T_pred(t+1)) / 4, fuzzy inference is performed. The fuzzy value of ΔD, i.e., the fuzzy value of the control change, is obtained through fuzzy inference.

[0082] The fuzzy set is defined as: negative large (NB), negative small (NS), zero (ZO), positive small (PS), and positive large (PB). The fuzzy rule is: if E is NB and ΔE is NB, then the change in output control quantity ΔD is PB (significantly increasing heating power).

[0083] Step 323: Defuzzify the fuzzy value of the control change to obtain the precise value of the control change of the drive duty cycle.

[0084] This step does not impose specific limitations on the defuzzification process. For example, the centroid method can be used; or the maximum membership averaging method can be used; or the weighted average method can be used, etc. This embodiment can use the centroid method for defuzzification, which can integrate the contributions of all fuzzy rules, making the output drive duty cycle control change value continuous, smooth, and stable, effectively improving the dynamic response performance and steady-state control accuracy of the control system.

[0085] Step 324: Update the drive duty cycle based on the precise value of the control change, and convert the updated drive duty cycle into a pulse width modulation signal, which is then output to the semiconductor cooler to achieve closed-loop temperature control of the temperature-controlled environment.

[0086] This step is the control execution step, updating the drive duty cycle: D(t+1) = D(t) +ΔD, and controlling the TEC (thermal cooler) by outputting a PWM (pulse width modulation) signal through the temperature control module. At the same time, the temperature prediction model uses the actual collected temperature value for online learning and updates the network weights to adapt to the thermal capacity characteristics of different calibration blocks. It should be noted that the thermoelectric cooler in this embodiment is a solid-state refrigeration device for thermoelectric conversion, used to achieve high-precision temperature control of the system. This device has the following characteristics: (1) small size, light weight, no moving parts, high reliability and long life; (2) fast cooling / heating response speed and high temperature control accuracy; (3) temperature can be precisely controlled by adjusting the magnitude and direction of the current; (4) suitable for high-precision, miniaturized and high-stability temperature control scenarios.

[0087] Furthermore, the aforementioned method of adjusting the gap displacement between the reference sensor and the measured material using a neural network motion controller to form a reference gap may include: Step 331: Obtain the state vector within the current control cycle; the state vector includes the position error of the translation stage, the rate of change of the position error, the control quantity at the previous moment, the target gap position, and the ambient temperature.

[0088] In this embodiment, the neural network motion controller can be a PID (Proportional-Integral-Derivative Controller); or it can be a fuzzy PID controller; or it can be a reinforcement learning controller; or it can be a model logic controller. Regardless of the type of controller, its controller parameters can be adjusted in real time to adapt to nonlinear, time-varying system characteristics (such as changes in friction, mechanical deformation caused by temperature, etc.).

[0089] Specifically, in each control cycle (e.g., 1ms), the current position pos_current fed back from the reference sensor is read from the motion control card. The position error e(t) and the rate of change of error de(t) / dt between the current gap position pos_current and the target gap position pos_target are calculated. At the same time, the control quantity (motor drive voltage) u(t-1) of the previous moment is recorded. These are used as the state vector s(t) = [e(t), de(t) / dt, u(t-1), pos_target, T], where T is the operating temperature (used to compensate for thermal expansion).

[0090] Step 332: Input the state vector into the decision network to obtain the action vector, update the controller parameters based on the action vector, and calculate the current control quantity.

[0091] This step uses a PID controller as an example. The state vector s(t) is input into the decision network, which outputs an action vector a(t) = [ΔKp, ΔKi, ΔKd], representing the adjustments to the PID parameters. Then, the controller parameters are updated: Kp_new = Kp_base + ΔKp, and Ki and Kd are updated similarly. Here, Kp_base is a base value that can be preset. Using the adjusted PID parameters, the current control quantity u(t) is calculated.

[0092] Step 333: Output the current control quantity to the servo motor driver through the motion control card to drive the translation stage to move, thereby changing the gap displacement between the reference sensor and the measured material to form a reference gap.

[0093] This step uses a motion control card to output the current control quantity u(t) to the servo motor driver, which then drives the translation stage to move. It can be understood that a reference sensor is mounted on the translation stage; driving the stage changes the gap displacement between the reference sensor and the material being measured, thus creating a reference gap.

[0094] Furthermore, after outputting the current control quantity to the servo motor driver via the motion control card to drive the translation stage to move and change the gap displacement between the reference sensor and the measured material to form a reference gap, the above may further include: acquiring the next state vector and calculating the reward value based on the next state vector; storing the state vector, action vector, reward value, and next state vector as a set of samples in the experience playback pool; when the update condition is met, batch acquiring samples from the experience playback pool; updating the evaluation network based on the acquired samples, and updating the decision network based on the action value output by the updated evaluation network using the policy gradient method.

[0095] In this embodiment, the reward value is used to penalize position errors, sudden changes in control inputs, and motion overshoot. The update condition is not specifically limited; for example, it could be that the cumulative number of samples reaches a threshold, or it could be that a preset time period has been reached. Specifically, this embodiment calculates the reward value r(t) based on the next state vector s(t+1). The specific formula for the reward is: r(t) = -(α... e(t)^2+β (u(t)-u(t-1))^2+γ The overshoot is penalized by three terms: the first term penalizes the error, the second term penalizes the drastic change in the control quantity (smoothness), and the third term penalizes the overshoot. α, β, and γ are weight coefficients, and overshoot is the overshoot value. The state vector (s(t), action vector a(t), reward value r(t), and next state vector s(t+1)) is stored in the experience replay pool. A batch of data is periodically sampled from the experience replay pool to update the evaluation network. Based on the action value output by the updated evaluation network, the decision network is updated using the policy gradient method. In this way, the neural network motion controller continuously learns and optimizes during calibration, enabling the controller parameters to adapt to the system dynamics at different positions and speeds.

[0096] The evaluation network comprehensively assesses the motion control performance during the calibration process, outputting evaluation indicators and state judgment results. The decision network, based on the evaluation network's output and combined with the system's current state and historical information, dynamically adjusts the control parameters to achieve adaptive optimization of the calibration process. The collaborative operation of the evaluation and decision networks separates and coordinates evaluation and decision-making, enhancing the system's adaptability, robustness, and calibration efficiency, making the entire intelligent calibration process more stable, accurate, and intelligent.

[0097] S304: Real-time acquisition of relevant data during calibration; relevant data includes reference gap, operating temperature, output voltage and the type of material being tested.

[0098] During the calibration process, relevant data are collected in real time: (1) Reference gap: measured and output by the reference sensor in real time, serving as the true reference value for calibration; (2) Operating temperature: collected in real time by the temperature sensor, used to reflect the influence of temperature on sensor characteristics; (3) Output voltage: the voltage signal output by the sensor to be calibrated under the current gap, temperature and the material being tested; (4) Type of material being tested: the type of target material being tested used in the current calibration, used to distinguish the influence of different materials on the sensor output.

[0099] Furthermore, after collecting relevant data in real time during the calibration process, the process may also include: conducting a quality assessment of the relevant data and adjusting the preset calibration parameters for subsequent calibrations in real time based on the quality assessment results; and collecting relevant data based on the adjusted preset calibration parameters.

[0100] In this embodiment, during the calibration process, the quality of relevant data such as the real-time acquired reference gap, operating temperature, output voltage, and type of material being tested is assessed to determine whether there are any abnormal fluctuations, missing data, out-of-tolerance data, low signal-to-noise ratio, or poor synchronization. Based on the assessment results, the preset calibration parameters are corrected online in real time, including adjusting the temperature point, gap sampling interval, motion control parameters, data acquisition parameters, or material calibration order. Then, subsequent calibration and data acquisition are performed according to the updated preset calibration parameters, thereby ensuring that the acquired data is stable, effective, and complete, and improving calibration efficiency and the accuracy of the final model.

[0101] S305: After traversing all preset test material types and operating temperature points, all relevant data collected will be combined into full-condition data.

[0102] After the system sequentially switches between all tested material types according to preset calibration parameters and iterates through all preset operating temperature points for each material, completing data acquisition for each operating condition, all valid data collected during the entire calibration process, including reference gap, operating temperature, output voltage, and tested material type, are summarized and integrated to form a complete dataset covering multiple materials, temperatures, and gap coupling conditions—this is the full-condition data. This data comprehensively reflects the sensor's true output characteristics under different operating conditions, providing a complete and reliable data foundation for subsequent mapping relationships.

[0103] S306: Based on full-condition data, establish a mapping relationship between output voltage, operating temperature, type of material under test, and reference gap.

[0104] This embodiment does not elaborate on this step in detail. For details, please refer to step S102 in embodiment 1.

[0105] S307: When the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated is completed.

[0106] This embodiment does not elaborate on this step in detail. For details, please refer to step S103 in embodiment 1.

[0107] As can be seen, the sensor calibration method provided in this embodiment involves: S301 obtaining the original response curve of the sensor by performing a preliminary scan of the sensor to be calibrated and the reference sensor with different gaps; S302: performing nonlinear analysis on the original response curve of the sensor to determine the preset calibration parameters; S303: starting the calibration process according to the preset calibration parameters, controlling the working temperature of the temperature-controlled environment through a neural network temperature controller, and adjusting the gap displacement between the reference sensor and the measured material through a neural network motion controller to form a reference gap; S304: collecting relevant data in real time during the calibration process; the relevant data includes the reference gap, working temperature, output voltage, and type of measured material; S305: after traversing all preset types of measured material and working temperature points, combining all the collected relevant data into full-condition data; S306: establishing a mapping relationship between output voltage, working temperature, type of measured material, and reference gap based on the full-condition data; S307: when the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated is completed. This method determines preset calibration parameters through preliminary scanning and nonlinear analysis, enabling rapid understanding of sensor characteristics, providing high-quality initial values, accelerating calibration convergence, and improving calibration stability and final accuracy. Employing a neural network temperature controller and a neural network motion controller achieves high-precision, high-stability temperature and displacement control, ensuring a stable calibration environment and accurate reference gaps from the outset. This significantly improves the accuracy, consistency, and reliability of sensor calibration, making the final calibration model more practical. Furthermore, a temperature prediction model predicts temperature trends in advance, and fuzzy decision-making adjusts power accordingly, effectively overcoming hysteresis and achieving rapid approach to the target temperature with no overshoot and minimal steady-state fluctuations. Finally, a dual-network reinforcement learning structure of decision network and evaluation network forms an intelligent control closed loop of online learning and real-time iterative optimization.

[0108] For a clearer understanding of this invention, please refer to the following details. Figure 4 , Figure 4 This is a flowchart illustrating a sensor calibration method provided in an embodiment of the present invention, which may specifically include: (1) After the calibration process is started, the intelligent initialization stage is first entered: the characteristics of the original response curve of the sensor are analyzed, and its nonlinearity, temperature sensitivity, material dependence and other nonlinear characteristics are quantitatively identified to provide accurate prior information for subsequent calibration; on this basis, the system automatically recommends the optimal calibration temperature range, gap sampling interval, motion control parameters and other preset calibration parameters to form an initial calibration strategy.

[0109] (2) Subsequently, the system enters the adaptive calibration loop: at the loop entry point, it first determines whether all preset working conditions (including all types of materials under test, working temperature points and gap sampling points) have been completed. If not, the calibration acquisition branch is executed: through the neural network temperature controller, the working temperature of the calibration environment is precisely controlled, and at the same time, the neural network motion controller adjusts the gap displacement between the reference sensor and the material under test to form a reference gap; under each working condition, the output voltage of the sensor to be calibrated, the reference gap, the working temperature and the type of material under test are collected synchronously, and the data integrity, signal-to-noise ratio, synchronization and abnormal fluctuations are evaluated in real time; according to the evaluation results, the subsequent calibration parameters are dynamically adjusted, such as optimizing the sampling interval, adjusting the motion speed or replanning the temperature points, to improve the data validity and acquisition efficiency.

[0110] (3) After all operating conditions have been traversed, the system enters the online learning and verification stage: Based on the full operating condition data, the control strategy and calibration model parameters are optimized online to further improve the accuracy and response speed of temperature control and motion control; then, through cross-validation, the generalization ability and prediction accuracy of the calibration model under different operating conditions are evaluated to verify its reliability; finally, the system generates an intelligent report containing calibration results, error distribution, model performance and uncertainty analysis to provide a comprehensive reference for the subsequent application of the sensor, and the entire calibration process ends.

[0111] Please refer to Figure 5 , Figure 5 A flowchart illustrating a sensor calibration architecture provided in this embodiment of the invention may specifically include: The system consists of three main parts: an industrial control computer, a control and data acquisition cabinet, and a mechanical control platform. The industrial control computer is the brain, enabling intelligent calibration decisions, motion control, temperature control, and multi-factor deep compensation, thus implementing the sensor calibration method provided in this embodiment. The measurement and control hardware is responsible for data acquisition and driving. The mechanical control platform provides high-precision mechanical motion and environment for calibration.

[0112] (1) The mechanical control platform includes: High-precision electrically controlled translation stage: Fixed to the test bench base with bolts. A sensor mounting bracket is installed on the translation stage to secure the sensor to be calibrated. The translation stage is driven by a closed-loop servo motor and ball screw, and has an integrated optical grating ruler for position feedback, achieving high-precision linear motion.

[0113] Multi-material calibration block assembly: Located in front of the translation stage in the direction of movement (i.e., the direction the sensor is facing), and also fixed on the test platform base. It consists of a circular turntable and multiple calibration blocks. The turntable is driven to rotate by a stepper motor to switch between calibration blocks of different materials. For example, iron, aluminum, copper, stainless steel, composite materials, etc., each calibration block corresponds to a type of material being tested.

[0114] High-precision reference sensor: Fixed to the test bench base by an independent bracket, its measuring probe is aligned with the surface of the calibration block (measuring the same position as the sensor to be calibrated, but avoiding interference) to measure the true gap value. Typically, the measuring axis of the reference sensor is parallel to the measuring axis of the sensor to be calibrated, and the measuring points are as close as possible.

[0115] (2) Control and data acquisition cabinet: Motion control card: Connected to the servo motor driver and grating ruler of the translation stage via cable, it receives position feedback and sends control commands to control the movement of the translation stage.

[0116] Data acquisition card: Connects the output signal of the sensor to be calibrated, the output signal of the high-precision gap reference sensor, and the temperature sensor signal on the calibration block via cable to achieve multi-channel synchronous data acquisition.

[0117] Temperature control module: Connects the refrigeration and heating devices and the temperature sensor via cables to achieve temperature control and measurement.

[0118] (3) Industrial control calculator: The industrial control computer has built-in intelligent control software that communicates with motion control cards, data acquisition cards, and temperature control modules via Ethernet or bus (such as PCIe, Peripheral Component Interconnect Express, or high-speed peripheral component interconnection) to send control commands and receive acquired data. The software contains multiple neural network modules that interact with each other through data flow. For example, the intelligent calibration decision module analyzes the raw sensor response curves acquired during the initial scan to determine preset calibration parameters, then calls the neural network motion controller and neural network temperature controller to perform calibration. The acquired full-condition data is used to train a multi-factor deep compensation model (i.e., the calibration model, which can achieve multi-factor deep compensation), while the neural network motion controller and neural network temperature controller themselves also learn and optimize online.

[0119] (4) Interactions in the workflow: At the start of calibration, the industrial control computer uses a motion control card to move the translation stage to its initial position and a temperature control module to control the operating temperature of the sensor to be calibrated. The data acquisition card simultaneously collects the output voltage, reference gap, and operating temperature of the sensor and uploads them to the industrial control computer. Based on the real-time data, the various neural network modules in the intelligent control software of the industrial control computer make decisions, adjust calibration parameters (such as control parameters and compensation parameters), and update the calibration model.

[0120] The sensor calibration device provided in the embodiments of the present invention will be described below. The sensor calibration device described below and the sensor calibration method described above can be referred to in correspondence.

[0121] Please refer to the details. Figure 6 , Figure 6 A schematic diagram of a sensor calibration device provided in an embodiment of the present invention may include: The full-condition data acquisition module 100 is used to acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, and each set of corresponding data includes the reference gap, operating temperature, type of material being measured and the output voltage of the sensor to be calibrated; The mapping relationship establishment module 200 is used to establish a mapping relationship between output voltage, operating temperature, type of material under test and reference gap based on the full operating condition data. The calibration end module 300 is used to end the calibration of the sensor to be calibrated when the mapping relationship is established and the preset accuracy condition is met. The reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are located at the same gap position, the same measured material, and the same operating temperature environment.

[0122] Furthermore, based on the above embodiments, the full-condition data acquisition module 100 may include: The control unit is used to start the calibration process according to the preset calibration parameters, control the working temperature of the temperature control environment through the neural network temperature controller, and adjust the gap displacement between the reference sensor and the material being measured through the neural network motion controller to form the reference gap; A data acquisition unit is used to acquire relevant data in real time during the calibration process; the relevant data includes reference gap, operating temperature, output voltage, and the type of material being measured. The component unit is used to combine all the relevant data collected after traversing all preset test material types and working temperature points to form the full-condition data.

[0123] Furthermore, based on the above embodiments, the control unit may include: The temperature prediction value acquisition subunit is used to input the temperature sequence of the measurement and control environment in the current period and the past N periods and the driving duty cycle of the current neural network temperature controller into the temperature prediction model to obtain the temperature prediction value of the measurement and control environment. The control change fuzzy value acquisition subunit is used to calculate the temperature deviation and the rate of change of the temperature deviation between the predicted temperature value and the target temperature. Based on the preset fuzzy rule base and membership function, fuzzy inference is performed on the temperature deviation and the rate of change of the temperature deviation to obtain the control change fuzzy value. The defuzzing processing subunit is used to defuzzify the fuzzy value of the control change amount to obtain the precise value of the control change amount of the driving duty cycle. The duty cycle drive subunit is used to update the drive duty cycle based on the precise value of the control change, and convert the updated drive duty cycle into a pulse width modulation signal, which is then output to the semiconductor cooler to realize closed-loop temperature control of the temperature control environment.

[0124] Furthermore, based on the above embodiments, the control unit may include: The state vector acquisition subunit is used to acquire the state vector within the current control cycle; the state vector includes the position error of the translation stage, the rate of change of the position error, the control quantity at the previous moment, the target gap position, and the ambient temperature. The current control quantity calculation subunit is used to input the state vector into the decision network to obtain the action vector, update the controller parameters based on the action vector, and calculate the current control quantity. The translation stage drive subunit is used to output the current control quantity to the servo motor driver through the motion control card, drive the translation stage to move, so as to change the gap displacement between the reference sensor and the material being measured, and form the reference gap; The reference sensor is mounted on the translation stage.

[0125] Furthermore, based on the above embodiments, the sensor calibration device may further include: The reward value calculation module is used to output the current control quantity to the servo motor driver through the motion control card, drive the translation stage to move, change the gap displacement between the reference sensor and the measured material, and form the reference gap, then collect the next state vector and calculate the reward value based on the next state vector. The storage module is used to store the state vector, the action vector, the reward value, and the next state vector as a set of samples into the experience replay pool; A batch sample acquisition module is used to acquire samples in batches from the experience replay pool when the update conditions are met. The network update module is used to update the evaluation network based on the collected samples, and to update the decision network using the policy gradient method based on the action value output by the updated evaluation network.

[0126] Furthermore, based on the above embodiments, the sensor calibration device may further include: The sensor raw response curve acquisition module is used to obtain the sensor raw response curve by performing a preliminary scan of the sensor to be calibrated and the reference sensor at different gaps before starting the calibration process according to preset calibration parameters, controlling the working temperature of the temperature control environment through a neural network temperature controller, and adjusting the gap displacement between the reference sensor and the measured material through a neural network motion controller to form the reference gap. The preset calibration parameter determination module is used to perform nonlinear analysis on the original response curve of the sensor to determine the preset calibration parameters.

[0127] Furthermore, based on the above embodiments, the sensor raw response curve acquisition module may include: The sensor raw response curve acquisition unit is used to place the sensor to be calibrated and the reference sensor in a target operating temperature and single measured material environment, and drive the sensor to be calibrated and the reference sensor to perform a unidirectional linear scan with a preset step size along the zero gap to the working gap interval through the neural network motion controller, and simultaneously acquire the raw reference gap value and the raw voltage value to form the sensor raw response curve.

[0128] Furthermore, based on the above embodiments, the preset calibration parameter determination module may include: The calibration parameter recommendation model acquisition unit is used to acquire a pre-trained calibration parameter recommendation model; the calibration parameter recommendation model is trained based on a paired dataset of the sensor's nonlinear characteristics and the optimal calibration parameters. The model output unit is used to input the original response curve of the sensor into the calibration parameter recommendation model, so as to extract the nonlinear features of the original response curve of the sensor using the calibration parameter recommendation model, and output the preset calibration parameters based on the nonlinear features.

[0129] Furthermore, based on the above embodiments, the model output unit may include: The processing subunit is used to normalize and interpolate the original response curve of the sensor to generate a processed curve. The feature extraction subunit is used to extract features from the processed curve through the convolutional and pooling layers of the calibration parameter recommendation model to obtain the nonlinear features. The nonlinear feature processing subunit is used to process the nonlinear features through the fully connected layer of the calibration parameter recommendation model to obtain the preset calibration parameters.

[0130] Furthermore, based on the above embodiments, the nonlinear features include at least one or more of the following: curve rate variation features, curve smoothness features, and saturation region shape features.

[0131] Furthermore, based on the above embodiments, the mapping relationship establishment module 200 may include: The calibration model training module is used to train the initial calibration model based on the output voltage, operating temperature, type of material under test and corresponding reference gap, so as to obtain the trained calibration model. The trained calibration model can characterize the nonlinear mapping relationship between the output voltage, operating temperature, type of material being measured, and reference gap of the sensor to be calibrated.

[0132] Furthermore, based on the above embodiments, the calibration model training module may include: The dataset construction unit is used to construct a training dataset based on the full-condition data. Each sample in the training dataset includes output voltage, operating temperature, type of material under test, and reference gap. The model training unit is used to divide the training dataset into a training set and a validation set, and input the samples in the training set into the initial calibration model. The mean squared error is used as the loss function, and the parameters of the initial calibration model are updated by backpropagation using an adaptive moment estimation optimizer. The model training termination unit is used to verify the accuracy of the initial calibration model using samples in the validation set when a first preset condition is met, and to stop training when a second preset condition is met, thus obtaining the trained calibration model.

[0133] Furthermore, based on the above embodiments, the sensor calibration device may further include: The current gap acquisition module is used to input the current operating temperature, current material type, and current output voltage collected by the sensor to be calibrated into the trained calibration model and output the corresponding current gap.

[0134] Furthermore, based on the above embodiments, the preset calibration parameters include at least one or more of the following: a temperature point list, a gap sampling interval list, motion control parameters, temperature control parameters, data acquisition parameters, and material calibration priority parameters.

[0135] Furthermore, based on the above embodiments, the sensor calibration device may further include: The quality assessment module is used to collect relevant data in real time during the calibration process, assess the quality of the relevant data, and adjust the preset calibration parameters for subsequent calibrations in real time based on the quality assessment results. The preset calibration parameter adjustment module is used to collect relevant data based on the adjusted preset calibration parameters.

[0136] Furthermore, based on the above embodiments, the calibration end module may include: The calibration termination unit is used to complete the calibration of the sensor to be calibrated when the mapping relationship is established and the allowable error and preset fit are met.

[0137] It should be noted that the order of the modules and units in the above-mentioned sensor calibration device can be changed without affecting the logic.

[0138] The sensor calibration device provided in this embodiment of the invention uses a full-condition data acquisition module 100 to acquire full-condition data of the sensor to be calibrated. The full-condition data includes at least one set of corresponding data, each set of corresponding data containing a reference gap, operating temperature, type of material being measured, and output voltage of the sensor to be calibrated. A mapping relationship establishment module 200 is used to establish a mapping relationship between output voltage, operating temperature, type of material being measured, and reference gap based on the full-condition data. A calibration end module 300 is used to end the calibration of the sensor to be calibrated when the mapping relationship is established and a preset accuracy condition is met. The reference gap is obtained from a reference sensor, and the sensor to be calibrated and the reference sensor are in the same gap position, the same material being measured, and the same operating temperature environment. This device establishes a high-precision mapping relationship between output voltage, operating temperature, material type, and reference gap by simultaneously acquiring full-condition data of the sensor to be calibrated and the reference sensor at the same gap position, using the same measured material, and operating temperature. This effectively solves the problems of cumbersome, inefficient, and easily affected by human factors traditional manual calibration processes, as well as insufficient calibration accuracy caused by the coupling effects of multiple factors such as temperature drift and material properties. It realizes automated, high-precision, and high-stability calibration of suspension guidance sensors, significantly improving calibration efficiency and detection accuracy, and can meet the high-precision and high-reliability requirements of high-speed maglev trains for gap detection.

[0139] Furthermore, a calibration model is trained based on full-condition data. This calibration model accurately fits the characteristics of sensors with strong nonlinearity and multi-factor coupling. It can also automatically achieve joint compensation of multiple physical quantities such as temperature, material, and gap, and has strong generalization ability and robustness, adapting to different conditions and different sensors. In addition, the calibration sensor can be directly embedded into the system for real-time inference, realizing the integration of calibration and application. It also facilitates automated training and accuracy iteration, improving calibration efficiency and reliability. Furthermore, by determining preset calibration parameters through preliminary scanning and nonlinear analysis, sensor characteristics can be quickly grasped, providing high-quality initial values, accelerating calibration convergence, and improving calibration stability and final accuracy. Employing neural network temperature controllers and neural network motion controllers achieves high-precision and high-stability temperature and displacement control, ensuring a stable calibration environment and accurate reference gaps from the outset. This significantly improves the accuracy, consistency, and reliability of sensor calibration, making the final calibration model more practically valuable. Moreover, by using a temperature prediction model to predict temperature trends in advance, fuzzy decision-making adjusts power accordingly, effectively overcoming hysteresis and achieving rapid approach to the target temperature with no overshoot and minimal steady-state fluctuations. Finally, a dual-network reinforcement learning structure of decision network and evaluation network forms an intelligent control closed loop of online learning and real-time iterative optimization.

[0140] The sensor calibration device provided in the embodiments of the present invention will be described below. The sensor calibration device described below and the sensor calibration method described above can be referred to in correspondence.

[0141] Please refer to Figure 7 , Figure 7 A schematic diagram of a sensor calibration device provided in an embodiment of the present invention may include: Memory 10 is used to store computer programs; The processor 20 is used to execute computer programs to implement the sensor calibration method described above.

[0142] The memory 10, processor 20, and communication interface 31 all communicate with each other through the communication bus 32.

[0143] In this embodiment of the invention, the memory 10 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment of the invention, the memory 10 may store programs for implementing the following functions: Acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, each set of corresponding data including reference gap, operating temperature, type of material being measured and output voltage of the sensor to be calibrated; Based on full-condition data, a mapping relationship is established between output voltage, operating temperature, type of material under test, and reference gap; When the mapping relationship is established and the preset accuracy conditions are met, the calibration of the sensor to be calibrated is completed. The reference gap is obtained by the reference sensor, and the sensor to be calibrated and the reference sensor are in the same gap position, the same measured material, and the same operating temperature environment.

[0144] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; and the data storage area may store data created during use.

[0145] Furthermore, memory 10 may include read-only memory and random access memory, providing instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores operating systems and operating instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and handling hardware-based tasks.

[0146] Processor 20 can be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic device. Processor 20 can be a microprocessor or any conventional processor. Processor 20 can call programs stored in memory 10.

[0147] Communication interface 31 can be an interface for the communication module, used to connect with other devices or systems.

[0148] Of course, it should be noted that, Figure 7 The structure shown does not constitute a limitation on the sensor calibration device in the embodiments of the present invention. In practical applications, the sensor calibration device may include a ratio Figure 7 More or fewer components as shown, or combinations of certain components.

[0149] It is understood that if the sensor calibration method in the above embodiments is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the current technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, magnetic disk, or optical disk, and other media capable of storing program code.

[0150] Based on this, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the sensor calibration method described above.

[0151] The following describes a computer program product provided by an embodiment of this application. The computer program product described below can be referred to in conjunction with other embodiments described herein.

[0152] A computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the aforementioned disclosed sensor calibration method.

[0153] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0154] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0155] Finally, it should be noted that in this document, relationships such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0156] The present invention provides a detailed description of a sensor calibration method, apparatus, device, readable storage medium, and product. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A sensor calibration method, characterized in that, include: Acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, each set of corresponding data including reference gap, operating temperature, type of material being measured and output voltage of the sensor to be calibrated; Based on the full operating condition data, a mapping relationship is established between output voltage, operating temperature, type of material under test and reference gap; When the mapping relationship is established and the preset accuracy condition is met, the calibration of the sensor to be calibrated ends. The reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are located at the same gap position, the same measured material, and the same operating temperature environment.

2. The sensor calibration method according to claim 1, characterized in that, Obtain full-condition data of the sensor to be calibrated, including: The calibration process is started according to the preset calibration parameters. The working temperature of the temperature control environment is controlled by the neural network temperature controller, and the gap displacement between the reference sensor and the material being measured is adjusted by the neural network motion controller to form the reference gap. Relevant data are collected in real time during the calibration process; the relevant data includes reference gap, operating temperature, output voltage, and the type of material being tested; After traversing all preset test material types and operating temperature points, all relevant data collected will be combined to form the full operating condition data.

3. The sensor calibration method according to claim 2, characterized in that, The operating temperature of the temperature-controlled environment is controlled through a neural network temperature controller, including: The temperature sequence of the current period and the past N periods of the measurement and control environment and the current driving duty cycle of the neural network temperature controller are input into the temperature prediction model to obtain the predicted temperature value of the measurement and control environment. The temperature deviation and the rate of change of the predicted temperature value and the target temperature are calculated. Based on the preset fuzzy rule base and membership function, fuzzy inference is performed on the temperature deviation and the rate of change of the temperature deviation to obtain the fuzzy value of the control change amount. The fuzzy value of the control change is defuzzified to obtain the precise value of the control change of the drive duty cycle; The drive duty cycle is updated based on the precise value of the control change, and the updated drive duty cycle is converted into a pulse width modulation signal and output to the semiconductor cooler to realize closed-loop temperature control of the temperature control environment.

4. The sensor calibration method according to claim 2, characterized in that, Adjusting the gap displacement between the reference sensor and the measured material using a neural network motion controller to form the reference gap includes: Obtain the state vector within the current control cycle; the state vector includes the position error of the translation stage, the rate of change of the position error, the control quantity at the previous moment, the target gap position, and the ambient temperature. The state vector is input into the decision network to obtain the action vector. The controller parameters are updated based on the action vector, and the current control quantity is calculated. The current control quantity is output to the servo motor driver through the motion control card to drive the translation stage to move, thereby changing the gap displacement between the reference sensor and the measured material to form the reference gap; The reference sensor is mounted on the translation stage.

5. The sensor calibration method according to claim 4, characterized in that, After outputting the current control quantity to the servo motor driver via the motion control card to drive the translation stage to move, thereby changing the gap displacement between the reference sensor and the measured material to form the reference gap, the process further includes: Collect the next state vector and calculate the reward value based on the next state vector; The state vector, the action vector, the reward value, and the next state vector are stored as a set of samples in the experience replay pool; When the update conditions are met, samples are collected in batches from the experience replay pool; The evaluation network is updated based on the collected samples, and the decision network is updated using the policy gradient method based on the action value output by the updated evaluation network.

6. The sensor calibration method according to claim 2, characterized in that, Before initiating the calibration process according to preset calibration parameters, controlling the operating temperature of the temperature-controlled environment through a neural network temperature controller, and adjusting the gap displacement between the reference sensor and the measured material through a neural network motion controller to form the reference gap, the process further includes: By performing preliminary scans of the sensor to be calibrated and the reference sensor at different gaps, the original response curves of the sensors are obtained. The original response curve of the sensor is subjected to nonlinear analysis to determine the preset calibration parameters.

7. The sensor calibration method according to claim 6, characterized in that, By performing preliminary scans of the sensor to be calibrated and the reference sensor at different gaps, the original response curves of the sensors are obtained, including: The sensor to be calibrated and the reference sensor are placed in a target operating temperature and a single test material environment. The neural network motion controller drives the sensor to be calibrated and the reference sensor to perform a unidirectional linear scan with a preset step size along the zero gap to the working gap range, and simultaneously collects the original reference gap value and the original voltage value to form the original response curve of the sensor.

8. The sensor calibration method according to claim 6, characterized in that, Nonlinear analysis is performed on the original response curve of the sensor to determine the preset calibration parameters, including: A pre-trained calibration parameter recommendation model is obtained; the calibration parameter recommendation model is trained based on a paired dataset of the sensor's nonlinear characteristics and the optimal calibration parameters. The original response curve of the sensor is input into the calibration parameter recommendation model to extract the nonlinear features of the original response curve of the sensor using the calibration parameter recommendation model, and the preset calibration parameters are output based on the nonlinear features.

9. The sensor calibration method according to claim 8, characterized in that, The original response curve of the sensor is input into the calibration parameter recommendation model to extract the nonlinear features of the original response curve using the calibration parameter recommendation model, and the preset calibration parameters are output based on the nonlinear features, including: The original response curve of the sensor is normalized and interpolated to generate a processed curve; The nonlinear features are obtained by extracting features from the processed curve through the convolutional and pooling layers of the calibration parameter recommendation model. The nonlinear features are processed by the fully connected layer of the calibration parameter recommendation model to obtain the preset calibration parameters.

10. The sensor calibration method according to claim 8, characterized in that, The nonlinear features include at least one or more of the following: curve slope variation features, curve smoothness features, and saturation region shape features.

11. The sensor calibration method according to any one of claims 1 to 10, characterized in that, Based on the full-condition data, a mapping relationship is established between output voltage, operating temperature, type of material under test, and reference gap, including: Based on the output voltage, operating temperature, type of material under test and corresponding reference gap, the initial calibration model is trained to obtain a trained calibration model. The trained calibration model can characterize the nonlinear mapping relationship between the output voltage, operating temperature, type of material being measured, and reference gap of the sensor to be calibrated.

12. The sensor calibration method according to claim 11, characterized in that, Based on the output voltage, operating temperature, type of material under test, and corresponding reference gap, the initial calibration model is trained to obtain a trained calibration model, including: A training dataset is constructed based on the full-condition data. Each sample in the training dataset includes output voltage, operating temperature, type of material under test, and reference gap. The training dataset is divided into a training set and a validation set, and the samples in the training set are input into the initial calibration model. The mean squared error is used as the loss function, and the parameters of the initial calibration model are updated by backpropagation using an adaptive moment estimation optimizer. When the first preset condition is met, the accuracy of the initial calibration model is verified using samples in the validation set. When the second preset condition is met, training is stopped, and the trained calibration model is obtained.

13. The sensor calibration method according to claim 11, characterized in that, After obtaining the trained calibration model, it may also include: The current operating temperature, current material type, and current output voltage collected by the sensor to be calibrated are input into the trained calibration model, and the corresponding current gap is output.

14. The sensor calibration method according to claim 2, characterized in that, The preset calibration parameters include at least one or more of the following: a list of temperature points, a list of gap sampling intervals, motion control parameters, temperature control parameters, data acquisition parameters, and material calibration priority parameters.

15. The sensor calibration method according to claim 2, characterized in that, After collecting relevant data in real time during the calibration process, the process also includes: The relevant data is subjected to quality assessment, and the preset calibration parameters for subsequent calibration are adjusted in real time based on the quality assessment results; Data is collected based on the adjusted preset calibration parameters.

16. The sensor calibration method according to claim 1, characterized in that, When the mapping relationship is established and the preset accuracy condition is met, the calibration of the sensor to be calibrated is completed, including: When the mapping relationship is established and the allowable error and preset fit are satisfied, the calibration of the sensor to be calibrated is completed.

17. A sensor calibration device, characterized in that, include: The full-condition data acquisition module is used to acquire full-condition data of the sensor to be calibrated; the full-condition data includes at least one set of corresponding data, and each set of corresponding data includes the reference gap, operating temperature, type of material being measured and the output voltage of the sensor to be calibrated; The mapping relationship establishment module is used to establish a mapping relationship between output voltage, operating temperature, type of material under test and reference gap based on the full operating condition data; The calibration end module is used to end the calibration of the sensor to be calibrated when the mapping relationship is established and the preset accuracy condition is met. The reference gap is obtained by a reference sensor, and the sensor to be calibrated and the reference sensor are located at the same gap position, the same measured material, and the same operating temperature environment.

18. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the frequency calibration method as described in any one of claims 1 to 16 when executing the computer program.

19. A readable storage medium, characterized in that, The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the frequency calibration method as described in any one of claims 1 to 16.

20. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the frequency calibration method according to any one of claims 1 to 16.

Images