Upconversion fluorescence wide temperature range high-precision temperature measurement method

Abstract

This invention relates to the field of temperature measurement technology, and in particular to a high-precision upconversion fluorescence wide-temperature-range temperature measurement method. By setting a switching temperature threshold, a fluorescence afterglow lifetime temperature measurement mode, which is highly sensitive to temperature, is used in the low and medium temperature ranges to ensure high accuracy. In the high temperature range, it automatically and seamlessly switches to a fluorescence radiation temperature measurement mode, utilizing the thermal radiation effect of the fluorescent material itself for measurement. This effectively compensates for the shortcomings of the fluorescence lifetime method in the high-temperature range, realizing continuous and high-precision temperature measurement from deep low temperature to high temperature. At the same time, a unified signal preprocessing and temperature prediction model is used to process the signals in both modes, making the system architecture simple and the data processing flow unified. The temperature measurement mode is automatically selected by judging the temperature prediction value corresponding to the intensity of the upconversion fluorescence signal, without manual intervention, realizing intelligent and fully automatic measurement.

Description

Technical Field

[0001] This invention relates to the field of temperature measurement technology, and in particular to a high-precision temperature measurement method for a wide temperature range of upconversion fluorescence. Background Technology

[0002] Fluorescent fiber optic thermometry, based on the detection principle of the correlation between the properties of fluorescent substances and temperature, has significant application value in various fields such as industry and scientific research. The core of this technology is to utilize the fluorescence quenching effect or the characteristic of fluorescence lifetime changing with temperature of fluorescent substances. Light signals are transmitted through optical fibers, and temperature is detected after photoelectric conversion and signal processing. Currently, the mainstream fluorescent fiber optic thermometry scheme is down-conversion fluorescence thermometry, which uses high-energy, short-wavelength ultraviolet / blue light to excite rare-earth materials, generating low-energy, long-wavelength fluorescence, and then calculates the temperature value based on the correlation between fluorescence lifetime and temperature.

[0003] However, existing downconversion fluorescence thermometry techniques have significant drawbacks: high-energy, short-wavelength ultraviolet / blue light used as excitation light easily excites background fluorescence in industrial media such as optical fibers, ceramic tubes, and couplers in the optical path, generating strong background noise interference. This problem directly leads to a decrease in the signal-to-noise ratio during the temperature measurement process, causing the effective temperature information in the fluorescence signal to be masked by noise, thus severely affecting the temperature measurement accuracy and failing to meet the requirements of high-precision temperature measurement scenarios. Furthermore, when a higher upper limit of temperature measurement is required, it is often necessary to turn to radiation thermometry. However, traditional radiation thermometry generally uses non-contact radiation thermometry (such as infrared thermometers) that relies on detecting the thermal radiation propagating through the air from the target object. Since the signal propagates through the air, it is susceptible to environmental interference and depends on the surface emissivity of the measured object, posing challenges to measurement accuracy and stability. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a high-precision upconversion fluorescence method with a wide temperature range, capable of achieving continuous measurement with high precision across a wide temperature range from low to high temperatures.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A high-precision upconversion fluorescence wide-temperature-range temperature measurement method includes the following steps: S1. Obtain fluorescence afterglow curve data at multiple calibration temperature points to form a training dataset; S2. Normalize and preprocess each fluorescence afterglow curve in the training dataset to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve. S3. Establish a temperature prediction model, and use multiple first fluorescence feature vectors and their corresponding calibration temperature points as training data to train the temperature prediction model to obtain a mapping model. S4. Obtain an upconversion fluorescence signal by exciting the fluorescent material, and determine whether the temperature value at which the intensity of the upconversion fluorescence signal is located is higher than a preset switching temperature threshold. S5. If the temperature value at which the intensity of the upconversion fluorescence signal is located is lower than or equal to the switching temperature threshold, then the fluorescence afterglow thermometry mode is executed: The upconversion fluorescence signal is converted into a first electrical signal, and the first electrical signal is then converted from analog to digital to obtain a first light intensity array; If the temperature at which the intensity of the upconversion fluorescence signal is located is higher than the switching temperature threshold, then the fluorescence radiation thermometry mode is executed: The infrared radiation signal generated by the fluorescent material under thermal radiation is acquired, converted into a second electrical signal, and the second electrical signal is converted from analog to digital to obtain a second light intensity array. S6. Perform normalization preprocessing on the first light intensity array or the second light intensity array obtained in step S5 to generate the second fluorescence feature vector; S7. Input the second fluorescence feature vector into the mapping model obtained in step S3 to obtain the temperature value at the measured location.

[0006] The beneficial effects of this invention are as follows: This solution sets a switching temperature threshold. In the low and medium temperature range (e.g., below 300°C), it adopts a fluorescence afterglow lifetime measurement mode that is extremely sensitive to temperature, ensuring high accuracy. In the high temperature range (e.g., above 300°C), it automatically and seamlessly switches to fluorescence radiation temperature measurement mode, utilizing the thermal radiation effect of the fluorescent material itself for measurement. This effectively compensates for the shortcomings of the fluorescence lifetime method in the high temperature range, realizing continuous and high-precision temperature measurement from deep low temperature (e.g., -196°C) to high temperature (e.g., 600°C).

[0007] In the high-temperature radiation temperature measurement mode, this scheme directly detects the infrared radiation signal generated by the fluorescent material in the fluorescent probe due to high temperature, and transmits the signal directly to the processing unit through optical fiber. This "contact probe, optical fiber guided signal" method fundamentally avoids the loss, environmental interference and background radiation effects caused by the transmission of radiation signal in the air in traditional non-contact radiation temperature measurement. At the same time, it reduces the dependence on the emissivity of the surface of the measured object and significantly improves the measurement stability and reliability in high-temperature complex environments.

[0008] This solution employs a unified signal preprocessing and temperature prediction model to process signals under both modes, resulting in a simple system architecture and a unified data processing flow. By judging the temperature prediction value corresponding to the intensity of the upconversion fluorescence signal, the temperature measurement mode is automatically selected without manual intervention, achieving intelligent and fully automatic measurement. Attached Figure Description

[0009] Figure 1 This is a flowchart of the steps of the upconversion fluorescence wide-temperature-range high-precision temperature measurement method of the present invention; Figure 2 This is the optical path diagram of the upconversion fluorescence wide-temperature-range high-precision temperature measurement method of the present invention. Detailed Implementation

[0010] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0011] Please refer to Figure 1 A high-precision upconversion fluorescence wide-temperature-range temperature measurement method, characterized by comprising the following steps: S1. Obtain fluorescence afterglow curve data at multiple calibration temperature points to form a training dataset; S2. Normalize and preprocess each fluorescence afterglow curve in the training dataset to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve. S3. Establish a temperature prediction model, and use multiple first fluorescence feature vectors and their corresponding calibration temperature points as training data to train the temperature prediction model to obtain a mapping model. S4. Obtain an upconversion fluorescence signal by exciting the fluorescent material, and determine whether the temperature value at which the intensity of the upconversion fluorescence signal is located is higher than a preset switching temperature threshold. S5. If the temperature value at which the intensity of the upconversion fluorescence signal is located is lower than or equal to the switching temperature threshold, then the fluorescence afterglow thermometry mode is executed: The upconversion fluorescence signal is converted into a first electrical signal, and the first electrical signal is then converted from analog to digital to obtain a first light intensity array; If the temperature at which the intensity of the upconversion fluorescence signal is located is higher than the switching temperature threshold, then the fluorescence radiation thermometry mode is executed: The infrared radiation signal generated by the fluorescent material under thermal radiation is acquired, converted into a second electrical signal, and the second electrical signal is converted from analog to digital to obtain a second light intensity array. S6. Perform normalization preprocessing on the first light intensity array or the second light intensity array obtained in step S5 to generate the second fluorescence feature vector; S7. Input the second fluorescence feature vector into the mapping model obtained in step S3 to obtain the temperature value at the measured location.

[0012] As can be seen from the above description, the beneficial effects of the present invention are as follows: This solution sets a switching temperature threshold. In the low and medium temperature range (e.g., below 300°C), it adopts a fluorescence afterglow lifetime measurement mode that is extremely sensitive to temperature, ensuring high accuracy. In the high temperature range (e.g., above 300°C), it automatically and seamlessly switches to fluorescence radiation temperature measurement mode, utilizing the thermal radiation effect of the fluorescent material itself for measurement. This effectively compensates for the shortcomings of the fluorescence lifetime method in the high temperature range, realizing continuous and high-precision temperature measurement from deep low temperature (e.g., -196°C) to high temperature (e.g., 600°C).

[0013] In the high-temperature radiation temperature measurement mode, this scheme directly detects the infrared radiation signal generated by the fluorescent material in the fluorescent probe due to high temperature, and transmits the signal directly to the processing unit through optical fiber. This "contact probe, optical fiber guided signal" method fundamentally avoids the loss, environmental interference and background radiation effects caused by the transmission of radiation signal in the air in traditional non-contact radiation temperature measurement. At the same time, it reduces the dependence on the emissivity of the surface of the measured object and significantly improves the measurement stability and reliability in high-temperature complex environments.

[0014] This solution employs a unified signal preprocessing and temperature prediction model to process signals under both modes, resulting in a simple system architecture and a unified data processing flow. By judging the temperature prediction value corresponding to the intensity of the upconversion fluorescence signal, the temperature measurement mode is automatically selected without manual intervention, achieving intelligent and fully automatic measurement.

[0015] Furthermore, the normalization preprocessing in step S2 specifically includes the following steps: Map the light intensity value of each fluorescence afterglow curve to the interval [0, 1]; The mathematical expression for normalization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The minimum light intensity value among all sampling points of the fluorescence afterglow curve. For the first The maximum light intensity value among all sampling points of the fluorescence afterglow curve. After normalization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0016] As can be seen from the above description, by mapping the light intensity value of the fluorescence afterglow curve to the [0,1] interval, the influence of the large difference in the peak value of the fluorescence curve at different temperatures is effectively eliminated. This avoids the peak size contributing too much to the feature and masking the core temperature-sensitive feature of decay rate. It also makes the preprocessed fluorescence feature vector have a uniform scale, providing standardized and high-quality data input for the subsequent training of the temperature prediction model, which helps to improve the model training effect and temperature measurement accuracy.

[0017] Furthermore, step S3 specifically includes: S31. Divide the dataset consisting of the multiple first fluorescence feature vectors and their corresponding calibration temperature points into a training set, a validation set, and a test set. S32. Using the first fluorescence feature vector in the training set as input data and the corresponding calibration temperature point as the training target, the model parameters of the temperature prediction model are updated through the optimization algorithm to obtain the preliminary trained temperature prediction model. S33. Using the first fluorescence feature vector in the validation set as input data and the corresponding calibration temperature point as validation target, evaluate the performance of the pre-trained temperature prediction model under different hyperparameters, and determine the optimal hyperparameter combination based on the evaluation results. S34. Using the first fluorescence feature vector in the test set as input data and the corresponding calibration temperature point as the test target, the performance of the temperature prediction model using the optimal hyperparameter combination is tested, and the model that passes the test is used as the completed mapping model.

[0018] As described above, dividing the dataset into training, validation, and test sets, optimizing model parameters using the training set, determining the optimal hyperparameter combination using the validation set, and validating model performance using the test set constitutes a complete model training and evaluation system. This process ensures that the temperature prediction model not only has a good fit on the training data but also maintains excellent generalization ability on the test data that was not used in the training, avoiding overfitting or underfitting, and further improving the reliability and stability of the temperature measurement method.

[0019] Furthermore, in step S31, the division ratio of the dataset is as follows: The training set accounts for 70%, the validation set accounts for 20%, and the test set accounts for 10%.

[0020] As described above, 70% of the training set provides sufficient data support for model parameter learning, ensuring that the model fully learns the mapping relationship between fluorescence feature vectors and temperature values; 20% of the validation set can effectively achieve hyperparameter tuning, balancing the model's fitting ability and generalization ability; and 10% of the test set can objectively evaluate the model's actual application performance, avoiding poor model training results or distorted evaluation results due to improper dataset partitioning, thus providing a quantitative standard for ensuring model performance.

[0021] Furthermore, if there are extreme outliers in the fluorescence afterglow curve in step S2, the normalization preprocessing method is replaced with the standardization processing method, and the mean of the fluorescence afterglow curve is 0 and the standard deviation of the fluorescence afterglow curve is 1. The mathematical expression for standardization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The average light intensity of the fluorescence afterglow curves. Standard deviation After standardization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0022] As can be seen from the above description, when there are extreme outliers in the fluorescence afterglow curve, standardization is used instead of normalization to make the mean of the fluorescence afterglow curve 0 and the standard deviation of the fluorescence afterglow curve 1. This effectively reduces the impact of extreme outliers on the data distribution and avoids outliers interfering with model training and temperature prediction results.

[0023] Furthermore, the normalization preprocessing operation in step S2 is the same as the normalization preprocessing operation in step S6.

[0024] As described above, the normalization preprocessing operations in steps S2 and S6 are completely identical, ensuring that the first fluorescence feature vector in the training phase and the second fluorescence feature vector in the actual temperature measurement phase maintain consistency in data scale, dimension, and processing logic. This consistency avoids differences in feature vector characteristics due to different preprocessing methods, ensuring that the second fluorescence feature vector input to the model during actual temperature measurement has the same data distribution characteristics as the input data during model training. This guarantees that the mapping model can accurately output the temperature value at the measured location, improving the reliability and consistency of the method.

[0025] Furthermore, in step S2, before performing normalization preprocessing on each fluorescence afterglow curve in the training dataset, the following steps are also included: A moving average filter operation is performed on each fluorescence afterglow curve in the training dataset to obtain a smooth curve; The early attenuation segment of the smooth curve is extracted as valid data; The valid data is normalized to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve.

[0026] As described above, before normalization preprocessing, moving average filtering and effective data truncation steps were added. Moving average filtering can smooth the high-frequency noise of the fluorescence afterglow curve and reduce the influence of irrelevant factors such as circuit noise and ambient light interference, resulting in a smoother curve that is closer to the true fluorescence decay law. Truncation of the early decay segment as effective data is because the temperature-sensitive information of fluorescence afterglow is mainly concentrated in the early decay segment, while the later signal is close to the background noise. Truncation can reduce redundant data, reduce the computational load of the model, and highlight the core temperature-sensitive features, further improving the model training efficiency and temperature measurement accuracy.

[0027] Furthermore, in step S6, before performing normalization preprocessing on the first or second light intensity array obtained in step S5, the following steps are also included: The first or second light intensity array obtained in step S5 is subjected to a moving average filtering operation to obtain a smoothed array. The early decay segment of the smoothed array is extracted as valid data; The valid data is normalized to generate a second fluorescence feature vector.

[0028] As described above, adding a moving average filter and effective data extraction step before the light intensity array preprocessing in the actual temperature measurement stage ensures that the signal processing flow during actual temperature measurement remains consistent with that during the training stage, enabling the second fluorescence feature vector to accurately reflect the fluorescence attenuation characteristics corresponding to the measured temperature. This step further enhances the standardization of signal processing, reduces the interference of noise and redundant data on the actual temperature measurement results, and ensures that the trained mapping model can stably perform in practical applications.

[0029] Furthermore, in step S1, the temperature range of the calibration temperature point is -196℃ to 600℃.

[0030] As described above, the temperature range for the clearly defined calibration temperature points is -196℃ to 600℃. This range covers the temperature measurement needs of most industrial production, scientific research experiments, and other scenarios, demonstrating broad applicability. By selecting calibration temperature points within this wide temperature range to construct a training dataset, the trained mapping model can accurately capture the mapping relationship between fluorescence features and temperature within this temperature range, meeting the temperature measurement needs of different scenarios and improving the practicality of this solution.

[0031] Furthermore, the temperature prediction model in step S3 is either a ridge regression model or a support vector regression model.

[0032] As described above, the ridge regression model, by regularizing and penalizing the squared absolute value of the parameter, can avoid overfitting and is suitable for fluorescence feature-temperature mapping scenarios with strong linear relationships. The support vector regression model, on the other hand, can effectively handle nonlinear relationships and is suitable for scenarios with complex fluorescence decay characteristics. The availability of these two models allows this approach to adapt to different fluorescent material properties and temperature measurement scenarios, further expanding the applicability of this method.

[0033] Please refer to Figure 1 and Figure 2 As shown, Embodiment 1 of the present invention is as follows: Please refer to Figure 1 A high-precision upconversion fluorescence wide-temperature-range temperature measurement method includes the following steps: S1. Obtain fluorescence afterglow curve data at multiple calibration temperature points to form a training dataset; In step S1, the temperature range of the calibration temperature point is -196℃ to 600℃.

[0034] S2. Normalize and preprocess each fluorescence afterglow curve in the training dataset to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve. The normalization preprocessing in step S2 specifically includes the following steps: Map the light intensity value of each fluorescence afterglow curve to the interval [0, 1]; The mathematical expression for normalization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The minimum light intensity value among all sampling points of the fluorescence afterglow curve. For the first The maximum light intensity value among all sampling points of the fluorescence afterglow curve. After normalization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0035] The normalization preprocessing operation in step S2 is the same as the normalization preprocessing operation in step S6.

[0036] In step S2, before performing normalization preprocessing on each fluorescence afterglow curve in the training dataset, the following steps are also included: A moving average filter operation is performed on each fluorescence afterglow curve in the training dataset to obtain a smooth curve; The early attenuation segment of the smooth curve is extracted as valid data; The valid data is normalized to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve.

[0037] S3. Establish a temperature prediction model, and use multiple first fluorescence feature vectors and their corresponding calibration temperature points as training data to train the temperature prediction model to obtain a mapping model. Step S3 specifically includes: S31. Divide the dataset consisting of the multiple first fluorescence feature vectors and their corresponding calibration temperature points into a training set, a validation set, and a test set. S32. Using the first fluorescence feature vector in the training set as input data and the corresponding calibration temperature point as the training target, the model parameters of the temperature prediction model are updated through the optimization algorithm to obtain the preliminary trained temperature prediction model. S33. Using the first fluorescence feature vector in the validation set as input data and the corresponding calibration temperature point as validation target, evaluate the performance of the pre-trained temperature prediction model under different hyperparameters, and determine the optimal hyperparameter combination based on the evaluation results. S34. Using the first fluorescence feature vector in the test set as input data and the corresponding calibration temperature point as the test target, the performance of the temperature prediction model using the optimal hyperparameter combination is tested, and the model that passes the test is used as the completed mapping model.

[0038] In step S31, the division ratio of the dataset is as follows: The training set accounts for 70%, the validation set accounts for 20%, and the test set accounts for 10%.

[0039] If there are extreme outliers in the fluorescence afterglow curve in step S2 (extreme outliers are abrupt changes; a sampling point should not change much from the sampling points before and after it. For example, if the previous sampling value was 100, and this time it was 90~110, which is normal, but this time it became 20, which is an abrupt change and belongs to extreme outliers), then the normalization preprocessing method will be replaced with the standardization processing method, and the mean of the fluorescence afterglow curve will be 0, and the standard deviation of the fluorescence afterglow curve will be 1. The mathematical expression for standardization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The average light intensity of the fluorescence afterglow curves. Standard deviation After standardization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0040] S4. Obtain an upconversion fluorescence signal by exciting the fluorescent material, and determine whether the temperature value at which the intensity of the upconversion fluorescence signal is located is higher than a preset switching temperature threshold. S5. If the temperature value at which the intensity of the upconversion fluorescence signal is located is lower than or equal to the switching temperature threshold, then the fluorescence afterglow thermometry mode is executed: The upconversion fluorescence signal is converted into a first electrical signal, and the first electrical signal is then converted from analog to digital to obtain a first light intensity array; If the temperature at which the intensity of the upconversion fluorescence signal is located is higher than the switching temperature threshold, then the fluorescence radiation thermometry mode is executed: The infrared radiation signal generated by the fluorescent material under thermal radiation is acquired, converted into a second electrical signal, and the second electrical signal is converted from analog to digital to obtain a second light intensity array. S6. Perform normalization preprocessing on the first light intensity array or the second light intensity array obtained in step S5 to generate the second fluorescence feature vector; In step S6, before performing normalization preprocessing on the first or second light intensity array obtained in step S5, the following steps are also included: The first or second light intensity array obtained in step S5 is subjected to a moving average filtering operation to obtain a smoothed array. The early decay segment of the smoothed array is extracted as valid data; The valid data is normalized to generate a second fluorescence feature vector.

[0041] S7. Input the second fluorescence feature vector into the mapping model obtained in step S3 to obtain the temperature value at the measured location.

[0042] The temperature prediction model in step S3 is either a ridge regression model or a support vector regression model.

[0043] The expression for the ridge regression model is: ; in, It is a bias term. , ... It is the feature weight. , ... It is the input fluorescence characteristic.

[0044] Parameter meaning: Each weight It intuitively reflects the corresponding fluorescence characteristics. The "contribution" to the temperature prediction results. If A positive value indicates that the larger the characteristic value, the higher the predicted temperature; if... A negative value means that the larger the feature value, the lower the predicted temperature. Bias term It is the temperature baseline value when all characteristic values ​​are zero.

[0045] The core parameters of a support vector regression model include three categories: One is the support vector, which is a key sample point selected from the training set and directly determines the decision boundary of the model; Secondly, there are kernel function parameters, such as the bandwidth γ of the radial basis function and the order of the polynomial kernel. These parameters determine how to map the low-dimensional fluorescence feature space to a high-dimensional space in order to better handle nonlinear relationships. Thirdly, there is the penalty coefficient C, which is used to control the model's tolerance for prediction errors.

[0046] Parameter significance: Support vectors are the carriers of the key "fluorescence feature-temperature" mapping relationship in the training data that the model "memorizes." By calculating the similarity with the fluorescence features of new inputs, they provide the core basis for temperature prediction. The kernel function parameter adjusts the mapping method of the feature space, enabling the model to adapt to fluorescence decay characteristics of varying complexity. The penalty coefficient C balances the model's fit to the training data and its generalization ability to new data; a C value that is too large may lead to overfitting, while a C value that is too small may lead to underfitting.

[0047] The expression for the support vector regression model is as follows: ; in, To predict temperature values, This is the bias compensation term; Kernel function is ,in, For the first A dataset of fluorescence features from 10 samples. This is a dataset of fluorescence features acquired in real time. This is a regularization parameter used to balance temperature measurement accuracy and model generalization ability, and needs to be optimized according to the characteristics of fluorescence data. The specific implementation steps of the above-mentioned upconversion fluorescence wide-temperature-range high-precision temperature measurement method are as follows: I. Preparation Stage: 1. To construct the upconversion fluorescence excitation and signal acquisition optical path, please refer to... Figure 2The optical path includes a 980nm wavelength light-emitting diode, two band-stop filters (band-stop filter 1 and band-stop filter 2; band-stop filter 1 allows light with wavelengths above 700nm to be reflected perpendicularly, while allowing all other wavelengths to pass through; band-stop filter 2 allows light with wavelengths above 700nm to be reflected at a 45° angle, while allowing all other wavelengths to pass through), optical fiber (dual-core optical fiber, with a chalcogenide core and a quartz core; the chalcogenide core is used to transmit radiation, and the quartz core is used to transmit fluorescence), upconversion phosphor at the sensor end (mainly NaYF4-based fluoride), photodiode (center wavelength 660nm), circuit board, and processor; 2. Prepare a platinum resistance thermometer with an accuracy of <0.1℃ to calibrate the actual temperature value at the calibration temperature point; 3. Determine the calibration temperature range as -196℃ to 600℃, and select multiple calibration temperature points evenly within this range (e.g., select points evenly at 10℃ intervals, for a total of 26 calibration temperature points).

[0048] II. Training Dataset Construction: 1. For each calibrated temperature point, the actual temperature value of that temperature point is obtained by calibration using a platinum resistance thermometer; 2. The processor controls a 980nm wavelength light-emitting diode to emit near-infrared light. This light is reflected by a 45° filter (i.e., band-stop filter 2) and then transmitted through an optical fiber to the upconversion phosphor at the end of the sensor, exciting the upconversion phosphor to produce 660nm fluorescence. 3. The 660nm fluorescence is reflected back to the photoelectric conversion module along the optical fiber, and then filtered by two stages of filters (band-stop filter 1 and band-stop filter 2) before entering the photodiode; 4. The photodiode converts the fluorescence signal into an electrical signal and transmits it to the processor. The processor performs analog-to-digital conversion through the ADC module to obtain the fluorescence afterglow curve data at the calibration temperature point. 5. Collect multiple sets of repeated fluorescence afterglow curve data at each calibration temperature point (for example, collect 50 sets of repeated fluorescence afterglow curve data at each calibration temperature point, and each set of fluorescence afterglow curve data is obtained by sampling at a sampling frequency of 1MHz for 10ms, containing 10,000 "timestamp-light intensity value" data pairs), and construct the training dataset corresponding to "fluorescence afterglow curve data-calibration temperature value".

[0049] 6. Associate each group of fluorescence afterglow curves with the corresponding precise temperature value (calibrated by a high-precision thermometer, such as a platinum resistance thermometer, with an error of <0.1℃) to form a labeled dataset D = {(X1, T1), (X2, T2), ..., (Xm, Tm)}, where Xk is the light intensity array of the k-th group of fluorescence afterglow curves (dimension 1×N, where N is the number of sampling points), and Tm is the corresponding temperature.

[0050] III. Training Dataset Preprocessing: 1. Perform a moving average filter operation (such as a window moving average filter with a window size of 5) on each fluorescence afterglow curve in the training dataset to remove high-frequency noise and obtain a smooth curve; 2. Extract the early decay segment of the smooth curve (e.g., the first 3ms) as valid data, and remove redundant data that is close to background noise in the later stages; 3. Check for extreme outliers in the valid data: If no extreme outliers exist, a normalization method is used to map the light intensity values ​​of the valid data to the interval [0, 1]. The mathematical expression for normalization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The minimum light intensity value among all sampling points of the fluorescence afterglow curve. For the first The maximum light intensity value among all sampling points of the fluorescence afterglow curve. After normalization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0051] If there are extreme outliers in the fluorescence afterglow curve in step S2, the normalization preprocessing method will be replaced with the standardization method, and the mean of the fluorescence afterglow curve will be 0, and the standard deviation of the fluorescence afterglow curve will be 1. The mathematical expression for standardization is: ; in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The average light intensity of the fluorescence afterglow curves. Standard deviation After standardization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

[0052] 4. After the above preprocessing steps, the first fluorescence feature vector corresponding to each fluorescence afterglow curve is generated.

[0053] IV. Temperature Prediction Model Training: 1. Dataset partitioning: The corresponding data of "first fluorescence feature vector - calibration temperature value" are divided into training set (70%), validation set (20%) and test set (10%) in a ratio of 7:2:1. 2. Initial model training: Select ridge regression or support vector regression as the temperature prediction model, use the first fluorescence feature vector in the training set as the input data, and the corresponding calibration temperature point as the training target. Update the model parameters through optimization algorithm (such as Adam optimizer) to obtain the initially trained temperature prediction model. 3. Hyperparameter tuning: The first fluorescence feature vector in the validation set is used as the input data, and the corresponding calibration temperature point is used as the validation target. The performance of the preliminary training model is evaluated under different hyperparameter combinations (e.g., using mean squared error (MSE) as the evaluation index). The optimal hyperparameter combination is determined based on the evaluation results. 4. Model performance test: The first fluorescence feature vector in the test set is used as the input data, and the corresponding calibration temperature point is used as the test target. The performance of the temperature prediction model with the optimal hyperparameter combination is tested. If the test results meet the preset temperature measurement accuracy requirements, the model is used as the trained mapping model and implanted into the embedded processor.

[0054] V. Actual Temperature Measurement Stage: During actual measurement, the following steps should be performed: 1. Signal acquisition and pattern prediction: The fluorescent material inside the temperature probe is excited by a pulsed light source to generate an upconversion fluorescence signal. The system analyzes the intensity of this signal in real time and, based on a preset algorithm or lookup table, quickly determines whether the temperature value corresponding to the intensity of the upconversion fluorescence signal is higher than a preset switching temperature threshold (e.g., 300℃). This switching temperature threshold is a key parameter pre-calibrated experimentally based on the characteristics of the fluorescent material used (such as the fluorescence lifetime decay curve with temperature). It marks the turning point where the accuracy of the fluorescence afterglow thermometry mode begins to decrease significantly, while the fluorescence radiation thermometry mode becomes applicable.

[0055] 2. Dual-mode signal acquisition and conversion: Based on the above predictions, the system automatically selects and executes the corresponding temperature measurement mode: If the temperature value is less than or equal to the switching temperature threshold, the system enters the fluorescence afterglow temperature measurement mode. In this mode, the photodetector receives the fluorescence afterglow signal (passive emission) generated by the excitation of the fluorescent material. This signal is converted into a first electrical signal and processed by an analog-to-digital converter to form a first light intensity array characterizing the dynamic process of afterglow decay.

[0056] If the temperature exceeds the switching temperature threshold, the system enters the fluorescence radiation temperature measurement mode. In this mode, the excitation light source can be turned off or its intensity significantly reduced. The photodetector primarily receives the infrared radiation signal (its center wavelength is typically in the infrared band) actively emitted by the fluorescent material itself due to thermal excitation in a high-temperature environment. This signal is unrelated to the excitation light and is the material's own thermal radiation. This infrared radiation signal is guided directly from the probe to the detector via a high-temperature resistant optical fiber, effectively avoiding loss and interference during propagation in the air. It is then converted into a second electrical signal and, after analog-to-digital conversion, forms a second light intensity array that primarily reflects the radiation intensity.

[0057] 3. Feature extraction and temperature calculation: The first or second light intensity array obtained in the above steps is subjected to normalization preprocessing consistent with the training phase to eliminate the influence of absolute signal intensity fluctuations and generate a second fluorescence feature vector for real-time query.

[0058] Finally, this second fluorescence feature vector is input into the temperature mapping model that has been established and optimized during the training phase. The model automatically matches the corresponding temperature mapping relationship based on the input feature pattern and finally outputs the accurate temperature value of the measured location.

[0059] The advantages of the upconversion fluorescence thermometry method designed in this scheme compared to the existing patented fluorescence intensity comparison method are as follows: Existing patents, such as application number CN202210666408.7, entitled "High-Temperature Fiber Optic Temperature Measurement Device Based on Fluorescence Intensity Ratio and Thermal Radiation Temperature Characteristics," employ two independent temperature-sensing probes. One probe collects fluorescence signals, filters the emitted light through an optical module, and then connects the fluorescence to a spectrometer to measure the fluorescence intensity at two different wavelengths for comparison and temperature calculation. The other probe handles high-temperature thermal radiation, measuring the radiation intensity through an optical power meter. Two probes and two independent optical fibers are installed simultaneously at the same temperature measurement point, leading to the spectrometer and optical power meter respectively. This design is complex, inconvenient to install, and unsuitable for various temperature measurement scenarios. The temperature measurement host requires two independent optical interfaces for each temperature measurement channel. Simultaneously, the spectrometer and optical power meter measure the fluorescence intensity ratio and radiation intensity respectively. Data analysis relies on a computer connected to the spectrometer and optical power meter, resulting in a large size and extremely high cost, hindering large-scale industrialization.

[0060] This solution uses a single dual-core optical fiber, with both cores within the same protective layer, connected to a single fluorescent probe. The fluorescent powder enables both low-temperature lifetime measurement and high-temperature radiation measurement, resulting in a small probe. Simultaneously, the acquisition unit requires only one optical interface per temperature measurement point, and internally uses a band-stop filter to filter the emitted wavelength. A module composed of a fluorescent photodiode and an infrared radiation detector is used to achieve low-temperature fluorescence lifetime detection and high-temperature thermal radiation detection. Its compact size makes it suitable for mass production and commercialization. In contrast, the fluorescence intensity comparison method requires more complex optical paths to separate two different wavelengths of light in the fluorescence or uses a spectrometer, which is costly and bulky. Therefore, this solution combines the fluorescence lifetime method and the thermal radiation method, using a dual-detector module with a high-speed processor to acquire and calculate accurate temperatures. Its small size and simple structure facilitate secondary integration and commercialization.

[0061] In summary, the upconversion fluorescence wide-temperature-range high-precision temperature measurement method provided by this invention, by setting a switching temperature threshold, adopts a fluorescence afterglow lifetime temperature measurement mode that is extremely sensitive to temperature in the low and medium temperature range (such as below 300°C) to ensure high accuracy; in the high temperature range (such as above 300°C), it automatically and seamlessly switches to the fluorescence radiation temperature measurement mode, using the thermal radiation effect of the fluorescent material itself for measurement, thereby effectively making up for the shortcomings of the fluorescence lifetime method in the high temperature range, realizing continuous and high-precision temperature measurement from deep low temperature (such as -196°C) to high temperature (such as 600°C).

[0062] In the high-temperature radiation temperature measurement mode, the infrared radiation signal generated by the fluorescent material in the fluorescent probe due to high temperature is directly detected, and the signal is directly transmitted to the processing unit through optical fiber. This "contact probe, optical fiber guided signal" method fundamentally avoids the loss, environmental interference and background radiation effects caused by the transmission of radiation signal in the air in traditional non-contact radiation temperature measurement. At the same time, it reduces the dependence on the emissivity of the surface of the measured object and significantly improves the measurement stability and reliability in high-temperature complex environments.

[0063] By employing a unified signal preprocessing and temperature prediction model to process signals under both modes, the system architecture is simplified and the data processing flow is unified. The temperature measurement mode is automatically selected by judging the temperature prediction value corresponding to the intensity of the upconversion fluorescence signal, without the need for manual intervention, thus realizing intelligent and fully automatic measurement.

[0064] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A high-precision upconversion fluorescence wide-temperature-range temperature measurement method, characterized in that, Includes the following steps: S1. Obtain fluorescence afterglow curve data at multiple calibration temperature points to form a training dataset; S2. Normalize and preprocess each fluorescence afterglow curve in the training dataset to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve. S3. Establish a temperature prediction model, and use multiple first fluorescence feature vectors and their corresponding calibration temperature points as training data to train the temperature prediction model to obtain a mapping model. S4. Obtain an upconversion fluorescence signal by exciting the fluorescent material, and determine whether the temperature value at which the intensity of the upconversion fluorescence signal is located is higher than a preset switching temperature threshold. S5. If the temperature value at which the intensity of the upconversion fluorescence signal is located is lower than or equal to the switching temperature threshold, then the fluorescence afterglow thermometry mode is executed: The upconversion fluorescence signal is converted into a first electrical signal, and the first electrical signal is then converted from analog to digital to obtain a first light intensity array; If the temperature at which the intensity of the upconversion fluorescence signal is located is higher than the switching temperature threshold, then the fluorescence radiation thermometry mode is executed: The infrared radiation signal generated by the fluorescent material under thermal radiation is acquired, converted into a second electrical signal, and the second electrical signal is converted from analog to digital to obtain a second light intensity array. S6. Perform normalization preprocessing on the first light intensity array or the second light intensity array obtained in step S5 to generate the second fluorescence feature vector; S7. Input the second fluorescence feature vector into the mapping model obtained in step S3 to obtain the temperature value at the measured location.

2. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, The normalization preprocessing in step S2 specifically includes the following steps: Map the light intensity value of each fluorescence afterglow curve to the interval [0, 1]; The mathematical expression for normalization is: ； in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The minimum light intensity value among all sampling points of the fluorescence afterglow curve. For the first The maximum light intensity value among all sampling points of the fluorescence afterglow curve. After normalization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

3. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, Step S3 specifically includes: S31. Divide the dataset consisting of the multiple first fluorescence feature vectors and their corresponding calibration temperature points into a training set, a validation set, and a test set. S32. Using the first fluorescence feature vector in the training set as input data and the corresponding calibration temperature point as the training target, the model parameters of the temperature prediction model are updated through the optimization algorithm to obtain the preliminary trained temperature prediction model. S33. Using the first fluorescence feature vector in the validation set as input data and the corresponding calibration temperature point as validation target, evaluate the performance of the pre-trained temperature prediction model under different hyperparameters, and determine the optimal hyperparameter combination based on the evaluation results. S34. Using the first fluorescence feature vector in the test set as input data and the corresponding calibration temperature point as the test target, the performance of the temperature prediction model using the optimal hyperparameter combination is tested, and the model that passes the test is used as the completed mapping model.

4. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 3, characterized in that, In step S31, the division ratio of the dataset is as follows: The training set accounts for 70%, the validation set accounts for 20%, and the test set accounts for 10%.

5. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, If there are extreme outliers in the fluorescence afterglow curve in step S2, the normalization preprocessing method will be replaced with the standardization method, and the mean of the fluorescence afterglow curve will be 0, and the standard deviation of the fluorescence afterglow curve will be 1. The mathematical expression for standardization is: ； in, For the first The first fluorescence afterglow curve The original light intensity value of each sampling point For the first The average light intensity of the fluorescence afterglow curves. Standard deviation After standardization, the first The fluorescence afterglow curve is repeated. The standardized light intensity value of each sampling point.

6. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, The normalization preprocessing operation in step S2 is the same as the normalization preprocessing operation in step S6.

7. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, In step S2, before performing normalization preprocessing on each fluorescence afterglow curve in the training dataset, the following steps are also included: A moving average filter operation is performed on each fluorescence afterglow curve in the training dataset to obtain a smooth curve; The early attenuation segment of the smooth curve is extracted as valid data; The valid data is normalized to generate a first fluorescence feature vector corresponding to each fluorescence afterglow curve.

8. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, In step S6, before performing normalization preprocessing on the first or second light intensity array obtained in step S5, the following steps are also included: The first or second light intensity array obtained in step S5 is subjected to a moving average filtering operation to obtain a smoothed array. The early decay segment of the smoothed array is extracted as valid data; The valid data is normalized to generate a second fluorescence feature vector.

9. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, In step S1, the temperature range of the calibration temperature point is -196℃ to 600℃.

10. The upconversion fluorescence wide-temperature-range high-precision temperature measurement method according to claim 1, characterized in that, The temperature prediction model in step S3 is either a ridge regression model or a support vector regression model.

Images