Method and System for Detecting Borate Solution Concentration Based on Raman Spectroscopy

Abstract

The present application discloses a method and system for detecting the concentration of boric acid solution based on Raman spectroscopy. The method includes: using a boric acid solution with a first preset concentration as a training sample, and interacting it with excitation light to form a first Raman spectrum; selecting a first interval in the first Raman spectrum that covers the characteristic peak of boric acid, and using the pixel points included in the first interval and the light intensity values corresponding to the pixel points as first spectral data; using the Lasso regression algorithm and establishing a regression model according to the first spectral data and the first preset concentration; using the regression model to detect a boric acid solution to be measured with an unknown concentration so as to determine the actual concentration of the boric acid solution to be measured. The above detection method establishes a regression model based on the Lasso regression algorithm, has higher detection accuracy and smaller error for the concentration of boric acid solution, and the regression model established based on the above method has good repeatability.

Description

Technical Field

[0001] This application relates to the technical field of pressurized water reactor analysis, and in particular to a method and system for detecting the concentration of boric acid solution based on Raman spectroscopy. Background Art

[0002] When light irradiates a substance, scattering occurs. In the scattered light, in addition to the elastic component (Rayleigh scattering) with the same frequency as the excitation light, there are also components with frequencies lower and higher than that of the excitation light. The latter phenomenon is collectively referred to as the Raman effect. The inelastic scattering generated by the interaction between molecular vibrations, optical phonons in solids and other elementary excitations and the excitation light is called Raman scattering. Generally, the spectrum formed by combining Rayleigh scattering and Raman scattering is called Raman spectrum. The Raman spectrum of the detected substance usually consists of a certain number of Raman peaks. Each Raman peak represents the wavelength position and intensity of the corresponding Raman scattered light. Each spectral peak corresponds to a specific molecular bond vibration, and the spectral peak spacing and position information are unique to a specific substance. Therefore, Raman spectra are usually used as fingerprint spectra for substance identification.

[0003] The Raman spectrum signal itself belongs to the detection of weak signals, and is affected by factors such as the test method, optical system, detection instrument, and detection environment, resulting in large errors and poor repeatability in the quantitative measurement of the solution concentration based on Raman spectroscopy for a long time, making it difficult to carry out quantitative detection in practice. Summary of the Invention

[0004] In order to solve the deficiencies of the prior art, the purpose of this application is to provide a method and system for detecting the concentration of boric acid solution based on Raman spectroscopy, which can improve the measurement accuracy of boric acid solution and have small measurement errors.

[0005] To achieve the above purpose, this application adopts the following technical solutions:

[0006] In a first aspect, this application provides a method for detecting the concentration of boric acid solution based on Raman spectroscopy, which is applied to a boric acid concentration detector without neutron radiation source in a pressurized water reactor. The method includes:

[0007] Taking a boric acid solution with a first preset concentration as a training sample, and interacting with the excitation light to form a first Raman spectrum;

[0008] Selecting a first interval covering the characteristic peaks of boric acid in the first Raman spectrum, and taking the pixel points included in the first interval and the light intensity values corresponding to the pixel points as the first spectral data;

[0009] Using the Lasso regression algorithm and establishing a regression model according to the first spectral data and the first preset concentration;

[0010] Using the regression model to detect the boric acid solution to be measured with an unknown concentration to determine the actual concentration of the boric acid solution to be measured.

[0011] Further, before using the regression model to detect the boric acid solution to be measured with an unknown concentration to determine the actual concentration of the boric acid solution to be measured, the method further includes:

[0012] Taking the boric acid solution with a second preset concentration as a training sample, and interacting with the excitation light to form a second Raman spectrum. Both the first Raman spectrum and the second Raman spectrum are Raman spectra that have not undergone background subtraction data processing;

[0013] Selecting a second interval covering the characteristic peak of boric acid in the second Raman spectrum, and taking the pixel points included in the second interval and the light intensity values corresponding to the pixel points as second spectral data;

[0014] Inputting the second preset concentration as a variable into the regression model, and comparing the calculation data generated by the regression model with the second spectral data to obtain the goodness-of-fit index of the regression model;

[0015] Determining the regression model according to the goodness-of-fit index.

[0016] Further, determining the regression model according to the goodness-of-fit index includes:

[0017] Judging whether the goodness-of-fit index is greater than the goodness-of-fit threshold;

[0018] If the judgment result is yes, outputting the regression model;

[0019] If the judgment result is no, adjusting the penalty coefficient in the Lasso regression algorithm and updating the regression model.

[0020] Further, using the Lasso regression algorithm and establishing a regression model according to the first spectral data and the first preset concentration includes:

[0021] Obtaining the first spectral data as the first input data;

[0022] Obtaining the first preset concentration as the second input data;

[0023] Selecting a penalty coefficient within a set range as the third input data;

[0024] Based on the Lasso regression algorithm to process the first input data, the second input data, and the third input data, and obtaining the regression coefficient corresponding to the minimum value of the objective function of the regression model through training.

[0025] Further, the penalty coefficient is an integer greater than 0 and less than 300.

[0026] Further, adjusting the penalty coefficient in the Lasso regression algorithm and updating the regression model includes:

[0027] Obtain the first spectral data and the second spectral data as the first input data, and extract the number of pixel points included in the first interval or the second interval from the first input data, where the number of pixel points included in the first interval and the second interval is the same;

[0028] Obtain the first preset concentration and the second preset concentration as the second input data, and extract the number of concentration types of training samples with different preset concentrations from the second input data;

[0029] Select a penalty coefficient within a set range as the third input data;

[0030] Use the number of concentration types and the number of pixel points as the fourth input data;

[0031] Obtain an updated regression model based on the first input data, the second input data, the third input data, and the fourth input data, and the regression coefficients corresponding to when the objective function of the updated regression model reaches the minimum value. The updated objective function satisfies the following relationship:

[0032] ;

[0033] where X represents the first input data, y represents the second input data, α represents the penalty coefficient, β represents the regression coefficient, N represents the number of concentration types of training samples with different concentrations in the second input data, and P represents the number of pixel points included in the characteristic peak interval;

[0034] Or, the updated objective function satisfies the following relationship:

[0035] .

[0036] where X represents the first input data, y represents the second input data, α represents the penalty coefficient, and β represents the regression coefficient.

[0037] Furthermore, the penalty coefficient selected for the training samples of the first preset concentration is different from the penalty coefficient selected for the training samples of the second preset concentration. When updating the regression model through training samples with different preset concentrations, the penalty coefficient is changed to obtain the regression coefficients corresponding to when the objective function reaches the minimum value.

[0038] Furthermore, the first Raman spectrum and the second Raman spectrum are spectra after data processing, where the data processing includes smoothing processing and / or noise reduction processing.

[0039] In a second aspect, the present application also provides a system for detecting the concentration of boric acid solution based on Raman spectroscopy, which is applied to a boric acid concentration detector without a neutron radiation source in a pressurized water reactor. The system includes:

[0040] A data acquisition unit for acquiring a first Raman spectrum, wherein the first Raman spectrum is formed by the interaction of a boric acid solution with a first preset concentration and an excitation light. The data acquisition unit selects a first interval covering the characteristic peak of boric acid in the first Raman spectrum, and takes the pixel points included in the first interval and the light intensity values corresponding to the pixel points as the first spectral data;

[0041] A model establishment unit for establishing a regression model according to the first spectral data and the first preset sodium concentration corresponding to the first spectral data. The regression model is established based on the Lasso regression algorithm, with the first spectral data as the independent variable and the first preset concentration as the dependent variable;

[0042] A concentration detection unit for detecting a boric acid solution to be measured with an unknown concentration according to the regression model to determine the actual concentration of the boric acid solution to be measured;

[0043] The data acquisition unit is further configured to acquire a second Raman spectrum, which is formed by the interaction of a boric acid solution with a second preset concentration and an excitation light. The data acquisition unit selects a second interval covering the characteristic peak of boric acid in the second Raman spectrum, and takes the pixel points included in the second interval and the light intensity values corresponding to the pixel points as the second spectral data; Both the first Raman spectrum and the second Raman spectrum are Raman spectra that have not undergone background subtraction data processing;

[0044] The model establishment unit inputs the second preset concentration as a variable into the regression model, compares the calculation data generated by the regression model with the second spectral data, obtains the goodness-of-fit index of the regression model, and determines the regression model according to the goodness-of-fit index.

[0045] The method for detecting the concentration of a boric acid solution based on Raman spectroscopy selects a boric acid solution with a preset concentration to generate a Raman spectrum through interaction with an excitation light, and the characteristic peak interval in the Raman spectrum. By taking the pixel points within the characteristic peak interval and the light intensity values corresponding to the pixel points as input data, a regression model is generated through the Lasso regression algorithm. Compared with the prior art, the regression model obtained by the technical solution of the present application has higher accuracy, smaller error, and higher reproducibility. Description of the Drawings

[0046] Figure 1 It is a flowchart of a method for detecting the concentration of a boric acid solution based on Raman spectroscopy provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of the Raman spectrum of the concentration of a boric acid solution provided by an embodiment of the present invention;

[0048] Figure 3 It is a flowchart of determining a regression model provided by an embodiment of the present invention;

[0049] Figure 4 It is an experimental data graph showing the relationship between the penalty coefficient and the goodness-of-fit index provided by an embodiment of the present invention;

[0050] Figure 5 It is a modeling flow chart provided by an embodiment of the present invention;

[0051] Figure 6 It is another modeling flow chart provided by an embodiment of the present invention;

[0052] Figure 7 It is an error measurement graph of a regression model established based on the Lasso regression algorithm provided by an embodiment of the present invention;

[0053] Figure 8 It is a schematic diagram of the model error obtained after training with spectral samples processed with and without baseline deduction provided by an embodiment of the present application;

[0054] Figure 9 It is a schematic diagram of a system for detecting the concentration of boric acid solution based on Raman spectroscopy provided by an embodiment of the present invention;

[0055] Figure 10 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0056] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the specific embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application.

[0057] An embodiment of the present invention provides a method and a system for detecting the concentration of boric acid solution based on Raman spectroscopy, so as to improve the detection accuracy of the concentration of boric acid solution, reduce the detection error, and help to carry out quantitative detection of the concentration of boric acid solution in practical applications.

[0058] As Figure 1 shown, the method for detecting the concentration of boric acid solution based on Raman spectroscopy provided by an embodiment of the present invention is applied to a boric acid concentration detector without a neutron radiation source in a pressurized water reactor. The method includes the following steps:

[0059] S101. Use a boric acid solution with a first preset concentration as a training sample, and interact with the excitation light to form a first Raman spectrum.

[0060] When the excitation light generated by the laser irradiates the boric acid solution, scattering occurs. In the scattered light, in addition to the elastic component (Rayleigh scattering) with the same frequency as the excitation light, there are also components with frequencies lower and higher than that of the excitation light. The latter phenomenon is collectively referred to as the Raman effect. The inelastic scattering generated by the interaction of elementary excitations such as molecular vibrations and optical phonons in solids with the excitation light is called Raman scattering. Generally, the spectrum formed by combining Rayleigh scattering and Raman scattering is called the Raman spectrum.

[0061] In the examples of the present application, boric acid solutions with concentrations of 0 mg / L, 35 mg / L, 60 mg / L, 113 mg / L, 282 mg / L, 517 mg / L, 800 mg / L, 1656 mg / L, 2660 mg / L, and 3990 mg / L are respectively used as training samples, which interact with the excitation light to generate Raman spectra with different concentrations. It should be noted that the boric acid solutions with the above concentrations are only one example and do not mean that only boric acid solutions within the above concentration range can be selected as training samples.

[0062] In the first modeling process, any one of the above concentrations of boric acid solution is selected as a training sample, and the Raman spectrum generated by the interaction of the training sample with the excitation light is used as the first Raman spectrum, and the first Raman spectrum is used for subsequent modeling processes.

[0063] Exemplarily, the embodiments of the present invention mainly detect the concentration of boric acid solution, which is specifically applied to the nuclear power field. Since boric acid controls the speed of uranium nuclear fission in nuclear power plants and affects the power generation of nuclear power plants. As a soluble poison for regulating the reactivity of nuclear power plants, boric acid is widely used in the coolant of pressurized water reactors to control the change of long-term reactivity by adjusting its concentration. It can be seen that for the above application scenarios, the concentration control requirements of boric acid solution are relatively high, especially for the detection accuracy of the concentration of boric acid solution. The existing concentration detection methods have relatively low accuracy, which has a greater impact on the production capacity of nuclear power plants.

[0064] S102. Select the first interval covering the characteristic peak of boric acid in the first Raman spectrum, and use the pixel points included in the first interval and the corresponding light intensity values as the first spectral data.

[0065] The first Raman spectrum includes a certain number of Raman peaks. Each Raman peak represents the wavelength position and intensity of the corresponding Raman scattered light, and each Raman peak corresponds to a specific molecular bond vibration.

[0066] Embodiments of the present invention select an interval representing the characteristic peak of boric acid, which contains at least one Raman peak. In the prior art, generally a Raman peak at a single wavelength is selected for subsequent modeling. Compared with the prior art, by selecting the interval representing the characteristic peak of boric acid for subsequent modeling, the influence of the power of the excitation light and other environmental factors on the generated model can be avoided.

[0067] Specifically, on the one hand, due to the influence of the fluctuation of the excitation light power, and on the other hand, there may be some impurities in the solution containing specific substances, resulting in large errors in the model established based on the Raman peak at a single wavelength when performing concentration detection. In addition, environmental interferences, such as factors like temperature, humidity, and electromagnetic field, may all lead to deviations in concentration detection, and due to the above uncertain factors, the reproducibility of the model established based on the Raman peak at a single wavelength is poor.

[0068] In the example of the present application, the interval representing the characteristic peak of boric acid contains several pixel points, and the number N of pixel points is greater than or equal to 450 and less than or equal to 650. This is to avoid the influence of the uncertainty of environmental factors on the model accuracy due to the too small range of the characteristic peak interval, and at the same time avoid too many invalid features due to the too large range of the characteristic peak interval.

[0069] As Figure 2 shown in the schematic diagram of the Raman spectrum, the abscissa represents pixel points, and the ordinate represents the light intensity value corresponding to the pixel points. The part between the two dashed lines is the selected characteristic peak interval, specifically the pixel points with the abscissa between 300 and 850.

[0070] It should be noted that taking the pixel points with the abscissa between 300 and 850 as the interval representing the characteristic peak of boric acid is only a preferred embodiment, and the interval selected in actual application can be appropriately offset horizontally.

[0071] S103. Use the Lasso regression algorithm and establish a regression model according to the first spectral data and the first preset concentration.

[0072] The Lasso regression algorithm obtains a relatively refined model by constructing a penalty function, which compresses some coefficients and sets some coefficients to zero.

[0073] In the specific calculation process, taking the first preset concentration as the independent variable and the first spectral data as the dependent variable, the regression coefficients are obtained by adjusting the penalty coefficient in the Lasso regression algorithm, and a regression model is established.

[0074] Since the concentration of the boric acid solution is a known value in each modeling process, the regression model calculated by the above method can be verified by the boric acid solution with a known concentration to obtain the final accurate regression model.

[0075] Specifically, the first spectral data includes N pixel points, and each pixel point corresponds to its respective light intensity value. Based on this, an N×2 data matrix is constructed. The data matrix includes one row of N pixel points and one row of light intensity values corresponding to the N pixel points, and this data matrix is used as the first spectral data.

[0076] Exemplarily, when the first preset concentration is 35 mg / L, the boric acid solution with a concentration of 35 mg / L forms a Raman spectrum after being irradiated by the excitation light. Pixel points between 300 and 800 are extracted from it to obtain the light intensity values corresponding to each pixel point, and a data matrix as shown in Table 1 is established as the first spectral data.

[0077] Table 1:

[0078]

[0079] S104. Detect the boric acid solution to be measured with an unknown concentration by using the regression model to determine the actual concentration of the boric acid solution to be measured.

[0080] Through the above settings, a regression model is established by using the Lasso regression algorithm for the functional relationship between the concentration of the boric acid solution and the light intensity values in the interval representing the characteristic peak of boric acid in the spectrogram, improving the detection accuracy of the concentration of the boric acid solution. Especially when applied to the nuclear power field, it can more precisely control the reaction process. In the embodiments of the present invention, the regression model established by the Lasso regression algorithm has a significant improvement in the detection accuracy of the concentration of the boric acid solution. In the boric acid solution with a concentration of 1000 mg / L, its detection error is greater than or equal to 12 ppm and less than or equal to 18 ppm, thereby improving the production capacity of nuclear power equipment and extending its service life.

[0081] As Figure 3 shown, for a method for detecting the concentration of a boric acid solution based on Raman spectroscopy provided in the embodiments of the present invention, before step S104 of the embodiment Figure 1 shown, the method may further include:

[0082] S301. Use the boric acid solution with the second preset concentration as a training sample and interact with the excitation light to form a second Raman spectrum.

[0083] It should be noted that the boric acid solution with the first preset concentration is used to establish a preliminary regression model, and the boric acid solution with the second preset concentration is used to verify the preliminary regression model, and the first preset concentration and the second preset concentration are different.

[0084] Since the principle of generating Raman spectra by training samples with different concentrations under the irradiation of excitation light is the same, it will not be elaborated here.

[0085] S302. Select the second interval in the second Raman spectrum that covers the characteristic peak representing boric acid, and use the pixel points included in the second interval and the corresponding light intensity values of the pixel points as the second spectral data.

[0086] Among them, the second interval selected from the second Raman spectrum has the same number of pixel points as the first interval selected from the first Raman spectrum, and the ranges of the characteristic peak intervals of the two are the same. From the above introduction of the characteristic peak interval selected in the first Raman spectrum, the pixel points with the abscissa between 300 and 850 are used as the interval of the second Raman spectrum. This controls the variables in the modeling process and avoids the poor reproducibility of the regression model caused by environmental factors or the influence of the excitation light power.

[0087] It can be understood that the second spectral data includes N pixel points, and each pixel point corresponds to its respective light intensity value. Based on this, an N*2 data matrix is constructed, and this data matrix is used as the second spectral data.

[0088] Exemplarily, when the first preset concentration is 60 mg / L, the boric acid solution with a concentration of 60 mg / L forms a Raman spectrum after being irradiated by the excitation light, and the pixel points between 300 and 800 are extracted from it. Based on this, the light intensity values corresponding to each pixel point are obtained, and a data matrix shown in Table 2 is established as the second spectral data.

[0089] Table 2:

[0090]

[0091] S303. Input the second preset concentration as a variable into the regression model, and compare the calculated data generated by the regression model with the second spectral data to obtain the goodness-of-fit index of the regression model.

[0092] The second preset concentration is a known quantity. Input the second preset concentration as an independent variable into the regression model. Through the regression model, the calculated data of the training samples of the second preset concentration can be calculated. Among them, the calculated data includes the pixel points included in the second interval and the corresponding light intensity values of the pixel points.

[0093] Furthermore, compare the calculated data with the second spectral data to obtain the goodness-of-fit index of the regression model. The goodness-of-fit index is used to measure the accuracy of the regression model.

[0094] S304. Determine whether the goodness-of-fit index is greater than the goodness-of-fit threshold. If the judgment result is yes, execute step S305; if the judgment result is no, execute step S306.

[0095] S305. Output the regression model.

[0096] S306. Adjust the penalty coefficient in the Lasso regression algorithm and update the regression model.

[0097] It should be noted that during the construction of the regression model, if the goodness-of-fit index of the regression model is less than the goodness-of-fit threshold, the penalty coefficient is adjusted to update the regression model so that the goodness-of-fit index of the updated regression model is closer to or exceeds the goodness-of-fit threshold.

[0098] In the example of this application, the goodness-of-fit threshold is greater than or equal to 0.99.

[0099] Through the above settings, the detection accuracy of the boric acid solution concentration can be improved, and the regression model established by the Lasso regression algorithm has a smaller error and higher reproducibility.

[0100] Based on the above updated regression model, it is also possible to select a boric acid solution with a third preset concentration as a training sample, interact with the excitation light to form a third Raman spectrum, select a third interval covering the characteristic peak of boric acid in the third Raman spectrum, use the pixel points included in the third interval and the corresponding light intensity values as the third spectral data, input the third preset concentration as a variable into the updated regression model, compare the calculated data generated by the updated regression model with the third spectral data to obtain the goodness-of-fit index of the updated regression model, and determine whether the goodness-of-fit index of the updated regression model is greater than the goodness-of-fit threshold. If the determination result is yes, the updated regression model is output as the final regression model. If the determination result is no, repeat the steps of S301 to S306 above. In the process of repeating the steps of S301 to S306 above, boric acid solutions with different concentrations can be selected as training samples to improve the accuracy of the final regression model and make the final regression model have high reproducibility.

[0101] As Figure 4 shown, this application also provides an experimental data graph of obtaining a higher goodness-of-fit index by adjusting the penalty coefficient during the establishment and update of the regression model. It can be seen that in the process of obtaining a more accurate model, since the relationship between the penalty coefficient and the goodness-of-fit index is non-linear, the accuracy of the regression model obtained by the above method is higher.

[0102] As Figure 5 shown, a method for detecting the concentration of a boric acid solution provided by an embodiment of the present invention uses the Lasso regression algorithm and establishes a regression model based on the first spectral data and the first preset concentration, including:

[0103] S501. Obtain the first spectral data as the first input data.

[0104] S502. Obtain the first preset concentration as the second input data.

[0105] S503. Select the penalty coefficient within the set range as the third input data.

[0106] S504. Process the first input data, the second input data, and the third input data based on the Lasso regression algorithm, and obtain the regression coefficients corresponding to when the objective function of the regression model reaches the minimum value through training.

[0107] As Figure 6 shown, a method for detecting the concentration of boric acid solution based on Raman spectroscopy provided by an embodiment of the present invention adjusts the penalty coefficient in the Lasso regression algorithm and updates the regression model, including:

[0108] S601. Obtain the first spectral data and the second spectral data as the first input data, and extract the number of pixel points included in the first interval or the second interval from the first input data, where the number of pixel points included in the first interval and the second interval is the same.

[0109] S602. Obtain the first preset concentration and the second preset concentration as the second input data, and extract the number of concentration types of training samples with different preset concentrations from the second input data.

[0110] S603. Select the penalty coefficient within the set range as the third input data.

[0111] S604. Use the number of concentration types and the number of pixel points as the fourth input data.

[0112] S605. Obtain the updated regression model according to the first input data, the second input data, the third input data, and the fourth input data, and the regression coefficients corresponding to when the objective function of the updated regression model reaches the minimum value. The updated objective function satisfies the following relationship:

[0113] ;

[0114] where X represents the first input data, y represents the second input data, α represents the penalty coefficient, β represents the regression coefficient, N represents the number of concentration types of training samples with different concentrations in the second input data, and P represents the number of pixel points included in the characteristic peak interval.

[0115] Or, the updated objective function satisfies the following relationship:

[0116] ;

[0117] where X represents the first input data, y represents the second input data, α represents the penalty coefficient, and β represents the regression coefficient.

[0118] In the examples of this application, the penalty coefficient is greater than or equal to 0 and less than or equal to 300. Select any value within the range of 0 to 300 as the penalty coefficient α of the Lasso regression algorithm. Through the Lasso regression algorithm, obtain the regression coefficient β of the regression model that meets the fitting degree threshold according to the training samples of different concentrations, and save the penalty coefficient α, where the penalty coefficient is an integer.

[0119] In some other examples, select any value within the range of 0 to 300 as the penalty coefficient α of the Lasso regression algorithm. Through the Lasso regression algorithm, obtain the regression coefficient β of the regression model that meets the fitting degree threshold according to different test conditions, and save the penalty coefficient α, where the different test conditions include at least one of the following test conditions:

[0120] The power of the excitation light, temperature, humidity, electromagnetic field, etc.

[0121] Optionally, different concentrations and different test conditions can also be used as a variable and incorporated into the calculation process of the Lasso regression algorithm to establish a more accurate regression model and retain the penalty coefficient α of the regression model.

[0122] In the examples of this application, the penalty coefficient selected for the training samples of the first preset concentration is different from the penalty coefficient selected for the training samples of the second preset concentration. When updating the regression model through the training samples of different preset concentrations, change the penalty coefficient to obtain the regression coefficient corresponding to the minimum value of the objective function.

[0123] As Figure 7 shown, the embodiments of the present invention also provide an error comparison chart of the regression models calculated based on the Lasso regression algorithm and the ElasticNet. Among them, the abscissa represents the boric acid solution with known concentration. For the regression models obtained through the above two regression algorithms, respectively input the boric acid solution with known concentration as the independent variable into their respective regression models, and calculate the respective calculation data through their respective regression models. Each calculation includes a number of pixels representing boric acid and the light intensity values corresponding to the number of pixels. Compare the light intensity values obtained by the above calculations with the light intensity values of the corresponding pixels in the spectrogram to obtain the prediction error represented by the ordinate.

[0124] When the value of the prediction error is closer to 0, it indicates that the accuracy of the regression model is higher. As Figure 7 can be seen, the change range of the prediction error of the regression model established based on the Lasso regression algorithm is smaller, and the prediction error is closer to 0 compared to the regression model established based on the ElasticNet.

[0125] In addition, compared with other regression algorithms, such as ridge regression algorithm, linear regression, polynomial regression, etc., the Lasso regression algorithm has higher detection accuracy when applied to the detection of boric acid solution concentration, and the reproducibility of its regression model is relatively high.

[0126] Since the Lasso regression algorithm is generally used for the establishment of classification models, in the field of concentration detection, it has better use effects compared with other regression algorithms. Especially in the special scenarios involving the detection of boric acid solution concentration in the nuclear power field, a regression algorithm with higher accuracy and better reproducibility is needed.

[0127] As an implementation manner, both the first Raman spectrum and the second Raman spectrum are Raman spectra that have not undergone background subtraction processing.

[0128] Optionally, the first Raman spectrum and the second Raman spectrum can be processed through data processing to eliminate the errors caused by environmental interference and noise. Among them, the data processing includes at least one of the following processing methods:

[0129] Smoothing processing, noise reduction processing.

[0130] It should be noted that since the Raman spectrum signal itself belongs to the detection of weak signals, it is affected by factors such as testing methods, optical systems, detection instruments, and detection environments. Since the information contained in the Raman spectrum is relatively complex, especially the information contained in the original Raman spectrum is the richest, simple mathematical processing is extremely likely to lose some physical information. Especially, even a slight change in the environment during the testing process will cause errors in the regression model. Therefore, using the Raman spectrum that has not undergone background subtraction processing as the training object and selecting the interval representing the boric acid characteristic peak from it can retain the original sample data in the machine learning process as much as possible. Thereby improving the detection accuracy of the obtained regression model.

[0131] Specifically, as Figure 8 shown, it shows a schematic diagram of the model error obtained after training with spectral samples before and after baseline subtraction. Among them, the abscissa represents the predicted sample concentration, and the ordinate represents the mean value of the sample concentration predicted by the training model and the error range. The central point on the line represents the predicted concentration mean value, the upper and lower error bars represent the error range of the predicted concentration, and the distance size represents the size of the error. According to Figure 8 it can be seen that under low concentration conditions (the concentration of the boric acid solution to be measured is lower than 700 mg / L), the predicted concentration error of the model obtained by training with the spectral samples after baseline subtraction is significantly greater than that without baseline subtraction. In the embodiments of the present application, baseline subtraction processing is not performed on the spectral data of the spectral training samples and the test samples, so as to reduce the predicted concentration error of the regression model under low concentration conditions and improve the detection accuracy of the regression model.

[0132] In summary, the embodiments of the present invention provide a regression model established based on a method for detecting the concentration of boric acid solution by Raman spectroscopy, which is used to detect the concentration of boric acid solution, improve the detection accuracy of the concentration of boric acid solution, reduce the detection error, and the regression model has high reproducibility when boric acid solutions with different concentrations are used as calculation conditions.

[0133] The embodiments of the present application also provide a system for detecting the concentration of boric acid solution by Raman spectroscopy, and this system can be a portable device. It should be noted that this system can also be other devices with processing functions; for example, it can be a certain control module in a portable device; for another example, it can be a central processing unit (CPU), a general-purpose processor, a network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller, a programmable logic device (PLD), or any combination thereof; for another example, it can be other circuits, devices, or software modules, etc.; as long as this system can implement the method for detecting the concentration of boric acid solution shown in the embodiments of the present application, other forms of systems are also within the protection scope of the present application, and the specific form of the system is not limited in the present application.

[0134] As Figure 9 shown, this system includes:

[0135] A data acquisition unit 11, configured to acquire a first Raman spectrum, where the first Raman spectrum is formed by the interaction between a boric acid solution with a first preset concentration and an excitation light. The data acquisition unit selects a first interval covering the characteristic peak of boric acid in the first Raman spectrum, and takes the pixel points included in the first interval and the light intensity values corresponding to the pixel points as the first spectral data;

[0136] A model establishment unit 12, configured to establish a regression model according to the first spectral data and the first preset sodium concentration corresponding to the first spectral data, where the regression model is a model established based on the Lasso regression algorithm with the first spectral data as the independent variable and the first preset concentration as the dependent variable;

[0137] A concentration detection unit 13, configured to detect the boric acid solution to be measured with an unknown concentration according to the regression model to determine the actual concentration of the boric acid solution to be measured.

[0138] It should be noted that the boric acid solution with the first preset concentration flows through the sample cell. By irradiating the boric acid solution with the first preset concentration in the sample cell with the excitation light, the first Raman spectrum representing the boric acid solution with the first preset concentration is generated. The sample cell can also be used to hold boric acid solutions with different concentrations. After a boric acid solution with a certain concentration forms a Raman spectrum under the irradiation of the excitation light, by increasing or decreasing the concentration of the boric acid solution in the sample cell and then irradiating it with the excitation light again, a Raman spectrum of another concentration is formed.

[0139] In the technical solution provided in the embodiment of the present application, boric acid solutions with different concentrations are used as training samples, and a regression model with concentration, light intensity value, and penalty coefficient as variables is established through the Lasso regression algorithm, which improves the detection accuracy of the concentration of the boric acid solution, and the reproducibility of the model is relatively high.

[0140] The embodiment of the present application also provides an electronic device, such as Figure 10 shown, including a processor 21 and a memory 22. The memory 22 stores machine-executable instructions that can be executed by the processor 21. The processor 21 is prompted by the machine-executable instructions to implement the steps of any of the above detection methods.

[0141] The memory 22 may include a random access memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk memory. Optionally, the memory 22 may also be at least one storage device located far from the aforementioned processor.

[0142] The above-mentioned processor 21 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; it may also be a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0143] In another embodiment provided by the present application, a computer-readable storage medium is also provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above detection methods are implemented.

[0144] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0145] It should be understood that those of ordinary skill in the art can make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the protection scope of the appended claims of this application.

Claims

1. A method for detecting the concentration of boric acid solution based on Raman spectroscopy, which is applied to a boric acid concentration detector without a neutron radiation source in a pressurized water reactor, and is characterized in that, The method includes: Taking a boric acid solution with a first preset concentration as a training sample, and interacting with excitation light to form a first Raman spectrum; Selecting a first interval covering the characteristic peaks of boric acid in the first Raman spectrum, and taking the pixel points included in the first interval and the light intensity values corresponding to the pixel points as first spectral data; Using the Lasso regression algorithm and establishing a regression model based on the first spectral data and the first preset concentration; Using the regression model to detect a boric acid solution to be measured with an unknown concentration to determine the actual concentration of the boric acid solution to be measured; Before using the regression model to detect a boric acid solution to be measured with an unknown concentration to determine the actual concentration of the boric acid solution to be measured, the method further includes: Taking a boric acid solution with a second preset concentration as the training sample, and interacting with excitation light to form a second Raman spectrum, where both the first Raman spectrum and the second Raman spectrum are Raman spectra that have not undergone background subtraction data processing; Selecting a second interval covering the characteristic peaks of boric acid in the second Raman spectrum, and taking the pixel points included in the second interval and the light intensity values corresponding to the pixel points as second spectral data; Taking the second preset concentration as a variable and inputting it into the regression model, and comparing the calculated data generated by the regression model with the second spectral data to obtain a goodness-of-fit index of the regression model; If the goodness-of-fit index is greater than the goodness-of-fit threshold, determining the regression model according to the goodness-of-fit index, where the goodness-of-fit threshold is greater than 0.

99.

2. The method according to claim 1, characterized in that, Determining the regression model according to the goodness-of-fit index includes: If the goodness-of-fit index is less than the goodness-of-fit threshold, adjusting the penalty coefficient in the Lasso regression algorithm and updating the regression model.

3. The method according to claim 1, characterized in that, The step of using the Lasso regression algorithm and establishing a regression model based on the first spectral data and the first preset concentration includes: Obtaining the first spectral data as the first input data; Obtaining the first preset concentration as the second input data; Selecting a penalty coefficient within a set range as the third input data; Based on the Lasso regression algorithm, processing the first input data, the second input data, and the third input data, and obtaining the regression coefficient corresponding to the minimum value of the objective function of the regression model through training.

4. The method according to claim 3, wherein The penalty coefficient is an integer greater than 0 and less than 300.

5. The method according to claim 2, wherein The step of adjusting the penalty coefficient in the Lasso regression algorithm and updating the regression model includes: Obtaining the first spectral data and the second spectral data as the first input data, and extracting the number of pixel points included in the first interval or the second interval from the first input data, where the number of pixel points included in the first interval and the second interval is the same; Obtaining the first preset concentration and the second preset concentration as the second input data, and extracting the number of concentration types of the training samples with different preset concentrations from the second input data; Selecting a penalty coefficient within a set range as the third input data; Take the number of concentration types and the number of pixel points as the fourth input data; Obtain the updated regression model according to the first input data, the second input data, the third input data, and the fourth input data, and the regression coefficients corresponding to when the objective function of the updated regression model reaches the minimum value. The updated objective function satisfies the following relational expression: ； where X represents the first input data, y represents the second input data, α represents the penalty coefficient, β represents the regression coefficient, N represents the number of concentration types of the training samples with different concentrations in the second input data, and P represents the number of pixel points included in the characteristic peak interval; Or, the updated objective function satisfies the following relational expression: ； where the X represents the first input data, y represents the second input data, α represents the penalty coefficient, and β represents the regression coefficient.

6. The method according to claim 5, wherein The penalty coefficient selected for the training samples of the first preset concentration is different from the penalty coefficient selected for the training samples of the second preset concentration. When updating the regression model with the training samples of different preset concentrations, the penalty coefficient is changed to obtain the regression coefficient corresponding to when the objective function reaches the minimum value.

7. The method according to claim 1, wherein The first Raman spectrum and the second Raman spectrum are spectra after data processing, wherein the data processing includes smoothing processing and / or noise reduction processing.

8. A system for detecting the concentration of boric acid solution based on Raman spectroscopy, which is applied to a boric acid concentration detector without a neutron radiation source in a pressurized water reactor, is characterized in that, The system includes: A data acquisition unit for acquiring a first Raman spectrum, wherein the first Raman spectrum is formed by the interaction of a boric acid solution of a first preset concentration with an excitation light. The data acquisition unit selects a first interval covering the characteristic peak of boric acid in the first Raman spectrum, and takes the pixel points included in the first interval and the light intensity values corresponding to the pixel points as the first spectral data; A model establishment unit for establishing a regression model according to the first spectral data and the first preset concentration corresponding to the first spectral data, wherein the regression model is a model based on the Lasso regression algorithm with the first spectral data as the independent variable and the first preset concentration as the dependent variable; A concentration detection unit for detecting a boric acid solution to be measured with an unknown concentration according to the regression model to determine the actual concentration of the boric acid solution to be measured; The data acquisition unit is further configured to acquire a second Raman spectrum, wherein the second Raman spectrum is formed by the interaction of a boric acid solution of a second preset concentration with an excitation light. The data acquisition unit selects a second interval covering the characteristic peak of boric acid in the second Raman spectrum, and takes the pixel points included in the second interval and the light intensity values corresponding to the pixel points as the second spectral data; both the first Raman spectrum and the second Raman spectrum are Raman spectra that have not undergone background subtraction data processing; The model establishment unit inputs the second preset concentration as a variable into the regression model, compares the calculation data generated by the regression model with the second spectral data to obtain a goodness-of-fit index of the regression model. If the goodness-of-fit index is greater than the goodness-of-fit threshold, the regression model is determined according to the goodness-of-fit index, and the goodness-of-fit threshold is greater than 0.99.

Images