An accurate and uncalibrated quantitative method based on automatic recognition and matching of different substrates

Abstract

The present application relates to the technical field of elemental analysis detection, and particularly relates to an accurate non-calibration quantitative method based on automatic recognition and matching of different substrates, comprising: detection; acquiring actual fluorescence intensity and an actual spectrum diagram, and determining the type of main dark matter; determining the corresponding diagram of the current detection sample and selecting a corresponding non-calibration quantitative model; acquiring a simulation diagram; calculating the similarity of the simulation diagram and the actual spectrum diagram, if the similarity is greater than a similarity threshold value, outputting each content in the sample, if the similarity is less than or equal to the similarity threshold value, outputting new each element content, and re-performing full spectrum matching of the simulation diagram and the actual spectrum diagram until the matching result converges. By constructing a diagram library, the change trend of the actual measurement diagram and the simulation diagram is compared, so that the condition that the same matter is determined as different matters due to the pure comparison of the similarity is avoided, the error caused by the wrong model is avoided, and the accuracy of the non-calibration quantification is increased.

Description

Technical Field

[0001] This invention relates to the field of elemental analysis and detection technology, and in particular to an accurate, label-free quantitative method based on automatic identification and matching of different matrices. Background Technology

[0002] X-ray fluorescence spectroscopy (XRF) analysis technology is widely used in materials science, geological exploration, environmental monitoring, and industrial quality control due to its advantages such as speed, non-destructive testing, and multi-element detection. With the increasing complexity of industrial testing scenarios, higher demands are being placed on the real-time performance and standard-free detection capabilities of XRF technology. Traditional quantitative methods rely on standard samples to establish calibration curves, but in practical applications, they often face problems such as difficulties in standard sample preparation and low matrix matching. Standard-free quantitative techniques, due to their advantage of not relying on standard samples, are gradually becoming an important tool in industrial testing, environmental monitoring, and geological exploration.

[0003] In related technologies, although some studies have attempted to improve them by expanding the database or introducing machine learning, there are still limitations such as low matrix matching accuracy, coarse dark matter compensation, and inaccurate background subtraction, which make it impossible to guarantee the accuracy of the detection results. Summary of the Invention

[0004] The purpose of this invention is to provide an accurate label-free quantitative method based on automatic identification and matching of different matrices, in order to solve the problem that, although some studies have attempted to improve the method by expanding the database or introducing machine learning, there are still limitations such as low matrix matching accuracy, coarse dark matter compensation, and inaccurate background subtraction, which make it impossible to guarantee the accuracy of the detection results.

[0005] This invention provides an accurate label-free quantification method based on automatic identification and matching of different matrices, comprising:

[0006] The samples are tested, and the test data is preprocessed.

[0007] Obtain the actual fluorescence intensity and actual spectrum of the substance being tested, and determine the main types of dark matter based on the actual spectrum;

[0008] Obtain preset spectra from the spectral library, calculate the similarity of the change trend between each preset spectra and the actual spectra, select the preset spectra with the highest similarity of change trend as the spectrum corresponding to the current sample, and select the corresponding label-free quantitative model.

[0009] The main dark matter types and experimental parameters are input into a standard-free quantitative model to obtain the theoretical fluorescence spectra of each element, and the theoretical fluorescence spectra are fitted to obtain the simulated spectra.

[0010] The simulated spectrum is matched with the actual spectrum and the similarity is calculated. If the similarity is greater than the similarity threshold, the content of each element in the sample is output. If the similarity is less than or equal to the similarity threshold, the new content of each element is output. The simulated spectrum and the actual spectrum are matched again until the matching results converge.

[0011] The experimental parameters include the voltage and current of the X-ray tube, as well as the measurement time.

[0012] As a preferred technical solution for an accurate, label-free quantitative method based on automatic identification and matching of different matrices, the determination of the main types of dark matter includes:

[0013] Obtain scattering peaks within a specific energy range in the actual spectrum;

[0014] The characteristic scattering peaks are compared with a pre-established database of dark matter characteristic peaks to determine the types of the main dark matter.

[0015] As a preferred technical solution for an accurate, label-free quantitative method based on automatic identification and matching of different matrices, the dark matter characteristic peak database includes energy and intensity information of characteristic scattering peaks of various dark matter under different conditions.

[0016] As a preferred technical solution for an accurate, label-free quantitative method based on automatic identification and matching of different matrices, the step of performing full-spectrum matching between simulated spectra and actual spectra includes:

[0017] The least squares method is used to calculate the similarity between the simulated spectrum and the actual spectrum.

[0018] Obtain the correlation degree between each element and the similarity, generate a priority sequence for the elements according to the correlation degree from high to low, and adjust the element content in turn according to the priority sequence;

[0019] Obtain the theoretical fluorescence intensity of each substance, generate a simulated spectrum, and repeat the above steps;

[0020] In response to the gradual increase in similarity and the decreasing rate of increase, the change rate is compared with the magnitude threshold. If the change rate is less than the magnitude threshold, the detection is considered complete.

[0021] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the element content is adjusted sequentially, and the gradient descent algorithm is used to adjust the content of each element. The adjustment step size is adaptively adjusted according to the difference in similarity.

[0022] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the spectral library stores the spectra of various samples, as well as the main components of dark matter, scattering peak characteristic parameters, and corresponding label-free quantitative model optimization parameters of the samples corresponding to the spectra. The optimization parameters include normalization factor, background correction coefficient, and spectral peak fitting offset.

[0023] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the calculation of the similarity of the changing trends of the preset spectral image and the actual spectral image further includes the following steps:

[0024] The preset spectrum and the actual spectrum are segmented according to energy range to form several corresponding sub-spectrums, which are then recorded as comparison groups.

[0025] Symbolic trend coding is used to analyze the changing trends within each sub-spectrum to obtain the corresponding symbol sequences;

[0026] Calculate the intra-group similarity of symbol sequences in the comparison group, and determine the weight coefficient of the comparison group based on the length of the energy interval;

[0027] The trend similarity is calculated based on the intra-group similarity and the weighting coefficient.

[0028] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the label-free quantitative model is configured as a relationship model between experimental parameters, dark matter types and theoretical fluorescence intensities of different elements. When the experimental parameters and dark matter types are input, the theoretical fluorescence intensity of each substance is output.

[0029] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the label-free quantitative model includes a piecewise multi-parameter background compensation algorithm to eliminate background noise generated by incident X-ray scattering, detector Compton escape effect and incomplete charge collection effect.

[0030] The compensation algorithm includes one or more combinations of polynomial fitting, cubic spline interpolation, B-spline interpolation, and Snip algorithm.

[0031] As a preferred technical solution for an accurate label-free quantitative method based on automatic identification and matching of different matrices, the segmented multi-parameter background compensation is divided into multiple compensation segments according to the energy range or the position of the element characteristic peak. Each compensation segment independently selects a background compensation algorithm and sets parameters. During the compensation process, the contribution of background noise in each segment to the element peak fitting is calculated in real time and the compensation weight is dynamically adjusted.

[0032] Compared with the prior art, the beneficial effects of the present invention are that by constructing a spectral library, the present invention compares the measured spectra with the spectra in the spectral library and compares the changing trends of the measured spectra with the simulated spectra. This avoids the situation where substances with the same content but different contents are judged as different substances due to simple comparison of similarity. Because the most suitable model is automatically matched (such as the carbon model for coal and the oxide model for ore), it can avoid errors caused by using the wrong model, thereby increasing the accuracy of standard-free quantification. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the steps of an accurate, label-free quantitative method based on automatic identification and matching of different matrices according to an embodiment of the present invention. Detailed Implementation

[0034] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0036] Please see Figure 1 The diagram shows a flowchart of the steps of an accurate label-free quantification method based on automatic identification and matching of different matrices according to an embodiment of the present invention, including:

[0037] Step S1: Detect the sample and preprocess the detection data;

[0038] Step S2: Obtain the actual fluorescence intensity and actual spectrum of the substance being tested, and determine the main types of dark matter based on the actual spectrum.

[0039] Step S3: Obtain the preset spectra in the spectrum library, calculate the similarity of the change trend between each preset spectra and the actual spectra, select the preset spectra with the highest similarity of change trend as the spectrum corresponding to the current sample, and select the corresponding label-free quantitative model.

[0040] Step S4: Input the main dark matter types and experimental parameters into the label-free quantitative model to obtain the theoretical fluorescence spectra of each element, and fit the theoretical fluorescence spectra to obtain the simulated spectra.

[0041] Step S5: Perform full-spectrum matching between the simulated spectrum and the actual spectrum and calculate the similarity. If the similarity is greater than the similarity threshold, output the content of each element in the sample. If the similarity is less than or equal to the similarity threshold, output the new content of each element. Perform full-spectrum matching between the simulated spectrum and the actual spectrum again until the matching results converge.

[0042] The experimental parameters include: the voltage and current of the X-ray tube, and the measurement time.

[0043] Specifically, this invention constructs a spectral library, thereby comparing the measured spectra with those in the library and comparing the changing trends of the measured and simulated spectra. This avoids situations where substances with the same content but different concentrations are classified as different substances due to simple comparison of similarity. Because it automatically matches the most suitable model (e.g., a carbon model for coal and an oxide model for ore), it avoids errors caused by using the wrong model, thereby increasing the accuracy of standard-free quantification.

[0044] Furthermore, determining the type of primary dark matter involves the following steps:

[0045] Obtain scattering peaks within a specific energy range in the actual spectrum;

[0046] Different types of dark matter produce characteristic scattering peaks at specific energies. By comparing these characteristic scattering peaks with a pre-established database of dark matter characteristic peaks, the main types of dark matter can be determined.

[0047] Furthermore, when the element content reaches a certain level, the ratio of analytical line intensity to scattered radiation intensity is essentially independent of the matrix. Based on this characteristic, it can be used to participate in the quantification of dark matter scattering peaks, thereby achieving accurate quantification of target elements and avoiding the problem of incorrect normalization calculation caused by dark matter, thus further increasing the accuracy of standard-free quantitative detection results.

[0048] In detail, the dark matter characteristic peak database was established by acquiring and analyzing XRF spectra of a large number of standard samples with known dark matter composition. The database includes the energy and intensity information of characteristic scattering peaks of various dark matter under different conditions.

[0049] Specifically, the dark matter characteristic peak database of this invention includes the spectra of 200 standard samples with known components (including coal, minerals, and polymers). The spectra are labeled with their corresponding dark matter types (such as carbon and oxygen). The data can be obtained directly through large-scale experiments or a limited number of experiments.

[0050] Further, in step S5, the simulated spectrum is fully matched with the actual spectrum, including:

[0051] Step S51: Use the least squares method to calculate the similarity between the simulated spectrum and the actual spectrum;

[0052] Step S52: Obtain the correlation between each element and the similarity, generate a priority sequence for the elements according to the correlation from high to low, and adjust the element content according to the priority sequence in turn;

[0053] Step S53: Obtain the theoretical fluorescence intensity of each substance, generate a simulated spectrum, and repeat the above steps;

[0054] Step S54: In response to the gradual increase in similarity and the decreasing magnitude of the increase, the magnitude of the change is compared with the magnitude threshold. If the magnitude of the change is less than the magnitude threshold, the detection is considered complete.

[0055] In detail, the degree of correlation can be determined by the degree of fluctuation in similarity when different elements are adjusted by the same amount. For example, if the iron content increases by 1%, the intensity of the simulated spectrum at a certain key energy point is closer to the actual value, and the increase in similarity is large, indicating that the correlation between iron and similarity is greater. Conversely, if the silicon content increases by 1%, the intensity of the simulated spectrum at a certain key energy point changes little or shows a negative growth, indicating that the correlation between silicon and similarity is weaker.

[0056] Furthermore, by comparing the correlation between different elements and similarity, a priority sequence can be generated for elements based on their correlation from high to low, prioritizing the adjustment of elements with greater influence, thereby accelerating the convergence speed and increasing the detection speed.

[0057] In detail, the content of each element is adjusted sequentially, and the gradient descent algorithm is used to adjust the content of each element. The adjustment step size is adaptively adjusted according to the difference in similarity. Taking iron as an example in this embodiment of the invention, when the similarity difference is greater than 20%, the step size of iron content adjustment is 5%, and when the similarity is less than or equal to 20%, the step size of iron content adjustment is 1%. The adjustment step size for trace elements should be even smaller.

[0058] Specifically, this invention increases the fitting speed by using large-step rapid adjustment and small-step fine adjustment, while avoiding over-adjustment or insufficient adjustment accuracy caused by using the same step size, thereby further increasing the accuracy of label-free quantitative detection results.

[0059] Furthermore, the spectral library stores the spectra of various samples, as well as the main components of dark matter, scattering peak characteristic parameters, and corresponding scale-free quantitative model optimization parameters of the samples corresponding to the spectra. The optimization parameters include normalization factor, background correction coefficient, and peak fitting offset. The determination and use of the values ​​of normalization factor, background correction coefficient, and peak fitting offset are all existing technologies and will not be elaborated here.

[0060] In detail, the construction of the spectral library includes XRF spectral acquisition of standard samples, dark matter composition calibration, and label-free quantitative model training. The standard samples include typical matrix types in the field of target detection, and each standard sample is associated with spectral data of at least three different concentration gradients.

[0061] Specifically, in step S3, calculating the similarity of the changing trends between the preset spectrum and the actual spectrum also includes the following steps:

[0062] Step S31: Divide the preset spectrum and the actual spectrum into segments according to energy range to form several corresponding sub-spectral maps and record them as comparison groups;

[0063] Step S32: Use symbolic trend coding to analyze the changing trends within each sub-spectrum and obtain the corresponding symbol sequence;

[0064] Step S33: Calculate the intra-group similarity of symbol sequences in the comparison group, and determine the weight coefficient of the comparison group based on the length of the energy interval;

[0065] Step S34: Calculate the trend similarity based on the intra-group similarity and weighting coefficients.

[0066] Furthermore, this invention divides the spectrum into several segments (e.g., 0-5keV, 5-15keV) using symbolic trend encoding. Each segment is represented by an arrow indicating the trend. If two segments have the same arrow order, they are considered to be of the same type of sample, even if the concentrations are different. This avoids focusing on specific numerical values ​​and only looking at the shape trend, thus preventing misjudgments caused by different concentrations. It also increases the accuracy of selecting preset spectra and provides an accurate basis for subsequent similarity determination, thereby further increasing the accuracy of label-free quantitative detection results.

[0067] In detail, the standard-free quantitative model is configured as a model relating experimental parameters, dark matter types, and the theoretical fluorescence intensity of different elements. When the experimental parameters and dark matter types are input, the theoretical fluorescence intensity of each substance is output.

[0068] Furthermore, the scale-free quantitative model includes a piecewise multi-parameter background compensation algorithm to eliminate background noise caused by incident X-ray scattering, detector Compton escape effect and incomplete charge collection effect.

[0069] The compensation algorithms include one or more combinations of polynomial fitting, cubic spline interpolation, B-spline interpolation, and Snip algorithm.

[0070] Furthermore, in actual detection processes, background can significantly impact the results, particularly for trace elements. This background effect stems from the scattering of incident X-rays (Compton and Rayleigh scattering), the Compton scattering escape effect of the detector, and incomplete charge collection, photoelectron, and Auger electron escape effects. Therefore, in quantitative calculations using the fundamental parameter method, background subtraction is necessary to eliminate the influence of this background on elemental peak fitting, thereby further increasing the accuracy of standard-free quantitative detection results.

[0071] Specifically, the segmented background compensation is divided into multiple compensation segments according to the energy range or the position of the element characteristic peak. Each compensation segment independently selects a background compensation algorithm and sets parameters. During the compensation process, the contribution of background noise in each segment to the element peak fitting is calculated in real time and the compensation weight is dynamically adjusted. The segmented compensation process is an existing technology and will not be described in detail here.

[0072] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. An accurate, label-free quantitative method based on automatic identification and matching of different matrices, characterized in that, include: The samples are tested, and the test data is preprocessed. Obtain the actual fluorescence intensity and actual spectrum of the substance being tested, and determine the main types of dark matter based on the actual spectrum; Obtain preset spectra from the spectral library, calculate the similarity of the change trend between each preset spectra and the actual spectra, select the preset spectra with the highest similarity of change trend as the spectrum corresponding to the current sample, and select the corresponding label-free quantitative model. The main dark matter types and experimental parameters are input into a standard-free quantitative model to obtain the theoretical fluorescence spectra of each element, and the theoretical fluorescence spectra are fitted to obtain the simulated spectra. The simulated spectrum is matched with the actual spectrum and the similarity is calculated. If the similarity is greater than the similarity threshold, the content of each element in the sample is output. If the similarity is less than or equal to the similarity threshold, the new content of each element is output. The simulated spectrum is matched with the actual spectrum again until the matching result converges. The experimental parameters include the voltage and current of the X-ray tube and the measurement time; The calculation of the similarity of the changing trends between the preset spectrum and the actual spectrum also includes the following steps: The preset spectrum and the actual spectrum are segmented according to energy range to form several corresponding sub-spectrums, which are then recorded as comparison groups. Symbolic trend coding is used to analyze the changing trends within each sub-spectrum to obtain the corresponding symbol sequences; Calculate the intra-group similarity of symbol sequences in the comparison group, and determine the weight coefficient of the comparison group based on the length of the energy interval; The trend similarity is calculated based on the intra-group similarity and the weighting coefficient.

2. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 1, characterized in that, The determination of the types of the main dark matter includes: Obtain scattering peaks within a specific energy range in the actual spectrum; The characteristic scattering peaks are compared with a pre-established database of dark matter characteristic peaks to determine the types of the main dark matter.

3. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 2, characterized in that, The dark matter characteristic peak database includes energy and intensity information of characteristic scattering peaks of various dark matter under different conditions.

4. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 1, characterized in that, Performing full-spectrum matching between the simulated spectrum and the actual spectrum includes: The least squares method is used to calculate the similarity between the simulated spectrum and the actual spectrum. Obtain the correlation degree between each element and the similarity, generate a priority sequence for the elements according to the correlation degree from high to low, and adjust the element content in turn according to the priority sequence; Obtain the theoretical fluorescence intensity of each substance, generate a simulated spectrum, and repeat the above steps; In response to the gradual increase in similarity and the decreasing rate of increase, the change rate is compared with the magnitude threshold. If the change rate is less than the magnitude threshold, the detection is considered complete.

5. The accurate label-free quantification method based on automatic identification and matching of different matrices according to claim 4, characterized in that, The element content is adjusted sequentially, and the gradient descent algorithm is used to adjust the content of each element. The adjustment step size is adaptively adjusted according to the difference in similarity.

6. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 1, characterized in that, The spectral library stores spectra of various samples, as well as the main dark matter components, scattering peak characteristic parameters, and corresponding scale-free quantitative model optimization parameters of the samples corresponding to the spectra. The optimization parameters include normalization factors, background correction coefficients, and spectral peak fitting offsets.

7. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 1, characterized in that, The standard-free quantitative model is configured as a model relating experimental parameters, types of dark matter, and the theoretical fluorescence intensity of different elements. When the experimental parameters and types of dark matter are input, the theoretical fluorescence intensity of each substance is output.

8. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 7, characterized in that, The scale-free quantitative model includes a piecewise multi-parameter background compensation algorithm to eliminate background noise generated by incident X-ray scattering, detector Compton escape effect and incomplete charge collection effect. The compensation algorithm includes one or more combinations of polynomial fitting, cubic spline interpolation, B-spline interpolation, and Snip algorithm.

9. The accurate label-free quantitative method based on automatic identification and matching of different matrices according to claim 8, characterized in that, The segmented multi-parameter background compensation is divided into multiple compensation segments according to the energy range or the position of the element characteristic peak. Each compensation segment independently selects a background compensation algorithm and sets parameters. During the compensation process, the contribution of background noise in each segment to the element peak fitting is calculated in real time and the compensation weight is dynamically adjusted.

Images