A material testing method based on an energy dispersive X-ray spectrometer, and an electronic device

Abstract

The application relates to a material testing method based on an energy dispersion X-ray spectrometer, and an electronic device, the testing method sets different fitting calculation peak strengths and peak positions for single-peak, overlapping-peak and light-element-dominant peaks, and cooperates with a quantitative calculation method of a peak area-sensitivity factor method, so that the success rate of material analysis and the analysis result accuracy can be effectively improved; by extracting EDX energy spectrum diagrams to perform characteristic analysis on peak fitting degrees, signal-to-noise ratios, energy intervals and total peak numbers, a random forest classifier is used to perform classification analysis on input characteristics to output energy spectrum type labels, and then corresponding algorithms in an algorithm library are automatically matched, the reproducibility of material analysis results is improved, batch processing is supported, 1000+ samples per day can be analyzed, and the testing results are reliable, accurate, and have good reproducibility.

Description

Technical Field

[0001] This invention relates to the field of materials composition analysis technology, and in particular to a materials testing method and electronic equipment based on an energy-dispersive X-ray spectrometer. Background Technology

[0002] Energy-dispersive X-ray spectroscopy (EDX) is a core tool for qualitative identification and quantitative analysis of material elements and is widely used in materials science, metallurgy, electronics, geology and other fields.

[0003] Current energy-dispersive X-ray spectroscopy analysis techniques mainly rely on the semi-automated software built into the energy-dispersive X-ray spectrometer, which has the following drawbacks:

[0004] 1) The accuracy of quantitative analysis is affected by standard substances, standard establishment, and standard sample preparation; 2) The detection limits for complex matrices or low-content elements are too high, making it difficult to meet the requirements of high-precision analysis; 3) Lack of adaptive calibration mechanism, instrument drift leads to insufficient long-term stability.

[0005] 4) Treating the content and detection limit of each element as independent analytical items without establishing the correlation between elements and the mapping relationship between energy spectrum characteristics and analytical results leads to low efficiency in subsequent data tracing and in-depth analysis. Summary of the Invention

[0006] Therefore, the purpose of this invention is to provide a material testing method based on energy-dispersive X-ray spectrometer.

[0007] A material testing method based on energy-dispersive X-ray spectroscopy includes the following steps: S1: Extract the peak fit, signal-to-noise ratio, and energy range features of each characteristic peak in the EDX spectrum, and extract the total number of peaks and the features of the analysis type in the EDX spectrum to form the feature vector of each characteristic peak; S2: Input the feature vectors of each feature peak into the random forest classifier to obtain the energy spectrum type label with the highest confidence [analysis type + peak type + element type]; S3: Match the corresponding analysis rule from the algorithm library based on the energy spectrum type label; S4: Analyze the EDX spectrum according to the analysis rules to obtain qualitative and / or quantitative analysis results.

[0008] Furthermore, the analysis types include: qualitative analysis and quantitative analysis; The peak shapes are classified according to the peak fit: If the fitting residual rate If the percentage is ≤5%, then it is a single peak; If the fitting residual rate is less than 5% If the percentage is ≤15%, it is considered an overlapping peak. If the fitting residual rate A peak value greater than 15% indicates a weak signal. The method for determining the element type is as follows: A peak matching algorithm is used to compare the characteristic peaks with the energy ranges in the standard element characteristic peak energy database; The percentage of standard element characteristic peaks with peak values ​​<2keV and >10keV corresponding to all characteristic peaks is calculated as follows: If the proportion of characteristic peaks with a peak position < 2keV > 60%, then light elements dominate. If the proportion of characteristic peaks with a peak position >10 keV >60%, then heavy elements dominate. The rest are mixed elements.

[0009] Furthermore, the analysis rules include first calculating the intensity and position of different characteristic peaks based on peak shape, light elements, and heavy elements; then removing false peaks based on the intensity of each characteristic peak; and finally matching qualitative and quantitative algorithms based on the intensity and position of the effective characteristic peaks and the qualitative and / or quantitative analysis of the energy spectrum labels.

[0010] Furthermore, the algorithm library includes characteristic peak intensity and peak position calculation, which includes single peak intensity and peak position algorithms, overlapping peak intensity and peak position algorithms, and light element intensity and peak position algorithms: The algorithm for determining single-peak intensity and peak position is as follows: Lorentz fitting or Gaussian fitting is performed on the characteristic peak of the single peak to obtain the actual peak intensity. Compared with the actual peak ; The algorithm for determining the intensity and position of overlapping peaks is as follows: Determine the peak positions of the two peaks in the overlapping peaks; Determine a unified fitting interval that covers the bimodal energy range; Construct and initialize the bimodal Lorentz-Gaussian mixture function; Within the fitting interval, peak signal splitting is achieved using the Levenberg-Marquardt iterative optimization algorithm; For the split bimodal peaks, Lorentz fitting and Gaussian fitting were used to obtain the actual peak intensities of each peak, respectively. and actual peak ,Bit ; The light element intensity and peak position algorithm, after calculating the intensity and peak position of single or overlapping peaks, uses an ultra-thin window detector correction method to correct the intensity, specifically as follows: The actual peak intensity of the characteristic peak is adjusted using the correction formula. The correction is performed using the following formula:

[0011] in, This represents the actual peak intensity of the characteristic peak. Indicates the corrected peak intensity of the characteristic peak, ( , The coefficients are for standard sample calibration, specifically for pure C (graphite) and pure Al2O3 standard samples. This is the peak position value of the characteristic peak.

[0012] Furthermore, in addition to calculating the characteristic peak intensity and position as described above, the method also includes compensating for the peak intensity of the characteristic peak based on dead time. Specifically: Correcting the actual peak intensity results in a corrected actual peak intensity. satisfy:

[0013] in, For the death time rate.

[0014] Furthermore, after completing the above calculations of characteristic peak intensity and peak position, the process also includes peak position deviation compensation and correction for the characteristic peaks. Specifically: Calibration is performed based on the energy spectrum of standard samples, and a peak position deviation compensation model is established to correct the actual peak position. satisfy:

[0015] in, This is the theoretical peak value.

[0016] Furthermore, the algorithm library includes qualitative algorithms, specifically: Obtain the standard peak position of the corresponding characteristic peak from the standard element characteristic peak energy database. ; Calculate the standard peak position Compared with the actual peak absolute value of deviation ; If the absolute value of the deviation <0.1keV, identify it as the element, and complete the elemental qualitative analysis of this characteristic peak; If the absolute value of the deviation If the peak value is ≥0.1 keV, continue comparing it with other elements or determine it as an unknown peak.

[0017] Furthermore, the algorithm library includes quantitative algorithms, which, based on qualitative algorithms, add the peak area-sensitivity factor method to calculate the content of the j-th element.

[0018]

[0019] in, Let be the intensity of the characteristic peak of the j-th element. Let be the sensitivity factor for the j-th element. For high-purity samples with a major element content > 95%, the sensitivity factor is 1.

[0020] Furthermore, it also includes step S51: S51: Perform ZAF correction on the quantitative analysis results to obtain the corrected quantitative results. ZAF correction includes atomic number correction, absorption correction and fluorescence correction.

[0021]

[0022] Among them, atomic number correction is ; Absorption correction ; Fluorescence correction is ; in, Indicates the intensity of the secondary fluorescence peak. This indicates the intensity of the primary X-ray characteristic peak.

[0023] Compared to existing technologies, this invention effectively improves the success rate and accuracy of material analysis by setting different fitting calculations for peak intensity and position for single peaks, overlapping peaks, and light element-dominated peaks, combined with a quantitative calculation method using the peak area-sensitivity factor method. By extracting features from EDX energy spectra, such as peak fitting degree, signal-to-noise ratio, energy range, and total number of peaks, and using a random forest classifier to classify the input features and output energy spectrum type labels, the corresponding algorithms are automatically matched in the algorithm library, improving the reproducibility of material analysis results. Even with different operators using this algorithm, the analysis deviation is less than 1.5%, far lower than the 8% deviation of manual analysis. Through this complete set of analysis methods, the efficiency of material analysis and testing is significantly improved, reducing the analysis time for a single energy spectrum from 3-8 minutes manually to less than 5 seconds. It supports batch processing, enabling the analysis of 1000+ samples per day, with reliable, accurate, and reproducible test results. Attached Figure Description

[0024] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings.

[0025] Figure 1 This is a flowchart of the material testing method based on energy-dispersive X-ray spectrometer of the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings of the embodiments of the present invention. The described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0027] The material testing method based on energy-dispersive X-ray spectrometer described in this invention includes the following steps: S0: Obtain EDX energy spectrum data and preprocess the EDX energy spectrum data to obtain the EDX energy spectrum.

[0028] Preprocessing of EDX energy spectrum data includes, but is not limited to, noise filtering, baseline correction, and energy calibration.

[0029] In practice, the Savitzky-Golay smoothing algorithm is used to smooth the energy spectrum curve, reducing noise while preserving peak characteristics.

[0030] A baseline correction algorithm is used to zero out the baseline of the energy spectrum curve, thus eliminating background drift.

[0031] An energy scale conversion algorithm is used to convert the channel number of the energy spectrum data into energy value.

[0032] The preprocessed energy spectrum data has clear peaks, stable baselines, and accurate energy axis calibration.

[0033] S1: Extract the peak fit, signal-to-noise ratio, and energy range characteristics of each characteristic peak in the EDX spectrum, and extract the total number of peaks and the characteristics of the analysis type in the EDX spectrum to form the feature vector of each characteristic peak.

[0034] The wavelet transform algorithm is used to identify all possible peak positions in the EDX energy spectrum as characteristic peaks, and to preliminarily determine the peak position (i.e., the energy value corresponding to the maximum intensity of the characteristic peak), energy range (i.e., an energy range containing the characteristic peak), and the total number of peaks in the EDX energy spectrum. Using the fitting residual rate The peak fit of each characteristic peak is represented by the residual rate of the fit. And satisfy:

[0035] In the formula: Indicates the measured peak intensity. This indicates the intensity of the fitted peak.

[0036] The signal-to-noise ratio is calculated using the local standard deviation algorithm, including the following steps: The EDX spectrum was divided into different regions, including peakless regions and peak regions; Select a flat region within the peakless region and calculate the standard deviation of the counts in this flat region as the noise level. ; Calculate the intensity of characteristic peaks ; Based on noise level Characteristic peak intensity Calculate the signal-to-noise ratio of this characteristic peak. .

[0037] The analysis types include qualitative analysis and quantitative analysis.

[0038] The analysis type can be obtained based on the user's measurement mode; or the analysis type can be determined based on whether the energy spectrum contains standard sample calibration information. If it does, it is a quantitative analysis; if it does not, it is a qualitative analysis.

[0039] S2: Input the feature vectors of each feature peak into the random forest classifier to obtain the energy spectrum type label with the highest confidence [analysis type + peak type + element type].

[0040] In practice, the random forest classifier is configured as follows: 1) Setting the number of decision trees to 100 can effectively balance classification accuracy and computational efficiency, because with more than 100 trees, the accuracy improvement is less than 0.5% but the time consumption increases by 30%; 2) The feature subset is 3. For example, 3 features are randomly selected for each tree, such as "peak fit + signal-to-noise ratio + energy range", "quantitative analysis + peak fit + signal-to-noise ratio", etc., to avoid a single feature dominating the classification. 3) The node splitting threshold Gini coefficient is ≤0.1 to ensure the consistency of leaf node features and reduce misclassification.

[0041] Compared to SVM (Support Vector Machine) and CNN (Convolutional Neural Network), the Random Forest classifier is more resistant to overfitting in multi-dimensional discrete feature classification; it does not require a large amount of labeled data, and only 200 sets of energy spectrum samples are needed to achieve a classification accuracy of over 92%; the training time is short, and the model training can be completed within 10 minutes; it is more suitable for rapid deployment in industrial scenarios.

[0042] The peak types include single peaks, overlapping peaks, and weak signal peaks. They are categorized based on peak fit: If the fitting residual rate If the percentage is ≤5%, then it is a single peak; If the fitting residual rate is less than 5% If the percentage is ≤15%, it is considered an overlapping peak. If the fitting residual rate A value greater than 15% indicates a weak signal peak.

[0043] The element types include light element-dominated, heavy element-dominated, and mixed elements.

[0044] A peak matching algorithm is used to compare the characteristic peaks with the energy ranges in the standard element characteristic peak energy database; The percentage of standard element characteristic peaks with peak values ​​<2keV and >10keV corresponding to all characteristic peaks is calculated as follows: If the proportion of characteristic peaks with a peak position < 2keV > 60%, then light elements dominate. If the proportion of characteristic peaks with a peak position >10 keV >60%, then heavy elements dominate. The rest are mixed elements.

[0045] S3: Match the corresponding analysis rule from the algorithm library based on the energy spectrum type label.

[0046] The algorithm library includes characteristic peak intensity and peak position calculation, false peak removal algorithm, qualitative algorithm, and quantitative algorithm.

[0047] The analysis rules include first calculating the intensity and position of different characteristic peaks based on peak shape, light elements, and heavy elements; then removing false peaks based on the intensity of each characteristic peak; and finally matching qualitative and quantitative algorithms based on the intensity and position of the effective characteristic peaks and the qualitative and / or quantitative analysis of the energy spectrum labels.

[0048] The calculation of characteristic peak intensity and peak position includes algorithms for single peak intensity and peak position, algorithms for overlapping peak intensity and peak position, and algorithms for light element intensity and peak position.

[0049] The algorithm for determining single-peak intensity and peak position is as follows: Lorentz fitting or Gaussian fitting is performed on the characteristic peak of the single peak to obtain the actual peak intensity. Compared with the actual peak .

[0050] The algorithm for determining the intensity and position of overlapping peaks is as follows: Determine the peak positions of the two peaks in the overlapping peaks; Determine a unified fitting interval that covers the bimodal energy range; Construct and initialize the bimodal Lorentz-Gaussian mixture function; Within the fitting interval, peak signal decomposition is achieved using the Levenberg-Marquardt iterative optimization algorithm, with the number of iterations ≤ 50 and the residual convergence threshold < 1%. For the split bimodal peaks, Lorentz fitting and Gaussian fitting were used to obtain the actual peak intensities of each peak, respectively. and actual peak ,Bit .

[0051] The light element intensity and peak position algorithm, after calculating the intensity and peak position of single or overlapping peaks, uses an ultra-thin window detector correction method to correct the intensity, specifically as follows: The actual peak intensity of the characteristic peak is adjusted using the correction formula. The correction is performed using the following formula:

[0052] in, This represents the actual peak intensity of the characteristic peak. Indicates the corrected peak intensity of the characteristic peak, ( , The coefficients are for standard sample calibration, specifically for pure C (graphite) and pure Al2O3 standard samples. This is the peak position value of the characteristic peak.

[0053] Through the above correction, the error in light element content was reduced from 15% to less than 3%.

[0054] After calculating peak intensity and peak position, a false peak removal calculation is performed. The specific false peak removal algorithm is as follows: Determine the actual peak intensity of each characteristic peak Is it less than 3 times the noise level? : If so, the characteristic peak is determined to be a false peak and is removed. If not, the characteristic peak is determined to be a valid peak.

[0055] When the characteristic peak is a light element peak, the peak intensity is corrected. Compare them.

[0056] Furthermore, after completing the above calculations of characteristic peak intensity and peak position, the method also includes compensation and correction of the characteristic peak intensity based on dead time, and peak position deviation compensation and correction of the characteristic peak.

[0057] Specifically, peak intensity correction adjusts the actual peak intensity based on detector dead time data, thus correcting the actual peak intensity. satisfy:

[0058] in, For the death time rate.

[0059] Peak position correction is performed based on the energy spectrum of a standard sample. A peak position deviation compensation model is established to correct the actual peak position. satisfy:

[0060] in, This is the theoretical peak value.

[0061] The actual peak intensity and position can be corrected before false peak removal, qualitative analysis, and quantitative calculation are performed.

[0062] The qualitative algorithm is as follows: Obtain the standard peak position of the corresponding characteristic peak from the standard element characteristic peak energy database. ; Calculate the standard peak position Compared with the actual peak absolute value of deviation ; If the absolute value of the deviation <0.1keV, identify it as the element, and complete the elemental qualitative analysis of this characteristic peak; If the absolute value of the deviation If the peak value is ≥0.1 keV, continue comparing it with other elements or determine it as an unknown peak.

[0063] The quantitative algorithm, based on the qualitative algorithm, adds the peak area-sensitivity factor method to calculate the content of the j-th element.

[0064]

[0065] in, Let be the intensity of the characteristic peak of the j-th element. Let be the sensitivity factor for the j-th element. For high-purity samples with a major element content > 95%, the sensitivity factor is 1.

[0066] S4: Analyze the EDX spectrum according to the analysis rules to obtain qualitative and / or quantitative analysis results.

[0067] Furthermore, to ensure the accuracy and reliability of the results, corrections for systematic errors in EDX detection, such as matrix effects, are also included for the quantitative analysis results.

[0068] S5: The quantitative analysis results are corrected using an error correction module.

[0069] Specifically, the following steps are included: S51: Perform ZAF correction on the quantitative analysis results to obtain the corrected quantitative results. ZAF correction includes atomic number correction, absorption correction and fluorescence correction.

[0070]

[0071] Among them, atomic number correction is ,in, Let j be the atomic weight of the j-th element; Absorption correction ,in, Let be the mass absorption coefficient of the j-th element. The mass density of the micro-region of the sample to be analyzed. This refers to the effective sampling thickness for energy dispersive spectroscopy (EDS). Specifically, A standard parameter database can be accessed to retrieve the target element and its corresponding characteristic peak energy based on the qualitative analysis. If the standard sample is known, Take the standard density of the standard sample; if it is an unknown sample, The current estimated density is obtained through quantitative analysis during the iterative process. The calculations are based on the instrument's operating parameters, such as accelerating voltage and beam current, combined with the sample type, and are performed using the instrument's built-in model. Fluorescence correction is This is applicable when a high-energy element excites a low-energy element, such as W exciting Kα rays in Fe.

[0072] To address the issues of high dispersion and poor stability in single-point quantitative results caused by uneven composition of micro-regions on the sample surface and statistical fluctuations in electron beam scanning in energy dispersive spectroscopy analysis, this method also includes a regional averaging correction operation for the calibrated quantitative results. By averaging the neighboring regions in the spatial domain, random errors are reduced, and more statistically significant content values ​​are output.

[0073] S52: The regional averaging correction method is used to correct the fluctuation values ​​of the content of each element in the calibration quantitative results.

[0074] The regional averaging correction method is based on the spatial distribution characteristics of multi-point scanning. It uses the regional averaging method to smooth and normalize the element content fluctuation values ​​of a single calibration quantitative result, and specifically includes the following steps.

[0075] (1) Constructing the data matrix: Suppose that M×N pixels (or analysis points) have been scanned in the current field of view. For the j-th element, construct a two-dimensional matrix of its original quantitative content:

[0076] Where cj(x,y) is the content value of the j-th element at point (x,y) after S51 absorption correction, x∈[1,M], y∈[1,N].

[0077] (2) Set a square sliding window with an odd number of side lengths, and the window size is W×W, for example, 3×3 or 5×5, where W is a configurable parameter; the window traverses the matrix with a step size of 1. Each pixel in the image.

[0078] (3) For the center pixel (x0, y0), its corrected final content value is The calculation formula is:

[0079] in: Ω represents the set of all valid pixels within a W×W window centered at (x0, y0); w(x,y) is the weighting coefficient; to balance accuracy and edge preservation, an inverse distance weighting is used, that is, the closer to the center, the greater the weight; for a simplified scheme, an equal-weighted average (all w=1) can be used. S is the normalization coefficient. This ensures that the total content after correction remains 100%.

[0080] Boundary processing rules: For pixels at the edge of the field of view, i.e. when the window exceeds the matrix range, the "edge mirror fill" or "discard the edge (only output the valid inner core area)" method is used to avoid introducing invalid zero values ​​that would lead to lower results.

[0081] Furthermore, it also includes updating the random forest classifier, specifically by supplementing the training sample library or feature weight configuration of the random forest classifier with peak position bias compensation parameters and feature data after multi-region averaging of random errors.

[0082] The following are examples of using the material testing method provided by this invention: Example 1: Analysis of Cu-Zn alloy samples Take a standard Cu-Zn alloy sample and analyze it according to steps S1-S5: After preprocessing, feature extraction was performed, and the random forest classifier determined it to be "quantitative analysis spectrum + overlapping peaks + mixed elements"; After matching the algorithm library, the Lorentz-Gaussian mixture fitting algorithm is called to separate the overlapping peaks of Cu Kα and Zn Kα. The content was calculated using the peak area-sensitivity factor method. After ZAF correction, the deviation between the measured Cu content and the theoretical value was 1.8%, and the deviation between the Zn content and the theoretical value was 2.1%. The analysis took 4.2 seconds, which meets the requirements for automated detection.

[0083] Example 2: Analysis of Al2O3 ceramic samples By taking Al2O3 ceramic standard samples and correcting the characteristic peak of light element O with an ultrathin window detector signal correction model, the content determination error was reduced from 15% to less than 2.5%, verifying the effectiveness of the light element quantitative algorithm.

[0084] Example 3: Batch Sample Analysis Batch processing of 100 sets of EDX spectra took a total of 4 minutes, which is 150 times faster than manual processing (about 600 minutes). The deviation between the elemental content analysis values ​​and theoretical values ​​of the standard samples was ≤2.5%.

[0085] Compared with the prior art, the present invention has the following beneficial effects: 1) By setting different fitting calculations for single peaks, overlapping peaks, and light element-dominated peaks, and combining this with the quantitative calculation method of peak area-sensitivity factor, the success rate of material analysis and the accuracy of analysis results can be effectively improved. 2) By extracting features such as peak fitting degree, signal-to-noise ratio, energy range, and total number of peaks from the EDX energy spectrum, a random forest classifier is used to classify and analyze the input features and output energy spectrum type labels. Then, the corresponding algorithm is automatically matched in the algorithm library, which improves the reproducibility of material analysis results. Even if different operators use this algorithm, the analysis deviation is <1.5%, which is far lower than the 8% deviation of manual analysis. 3) Through a complete set of analytical methods, the efficiency of material analysis and testing is greatly improved, reducing the time for a single energy dispersive spectroscopy analysis from 3-8 minutes manually to less than 5 seconds. It supports batch processing and can analyze more than 1,000 samples per day. The test results are reliable, accurate, and reproducible.

[0086] The aforementioned material testing method based on energy-dispersive X-ray spectroscopy is stored in an electronic device and executed by this device to achieve variable slope speed regulation of the heat pump fan. The electronic device includes, but is not limited to, memory, processor, and network interface that can communicate with each other via a system bus.

[0087] The memory includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. The memory can be an internal storage unit of the electronic device, such as the hard disk or RAM of the electronic device. The memory can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. The memory may also include both internal storage units and external storage devices of the electronic device.

[0088] The processor can be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. This processor is typically used to control the overall operation of the electronic device, such as performing control and processing related to data interaction or communication with the electronic device. The processor is used to run program code stored in the memory or process data, for example, to run the material testing method based on energy-dispersive X-ray spectroscopy.

[0089] The network interface may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the electronic device and other electronic devices. For example, the network interface is used to connect the electronic device to an external data platform via a network, establishing a data transmission channel and communication connection between the electronic device and the external data platform. The network may be an intranet, the Internet, Global System for Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0090] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0091] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0092] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A material testing method based on energy-dispersive X-ray spectroscopy, characterized in that, Includes the following steps: S1: Extract the peak fit, signal-to-noise ratio, and energy range features of each characteristic peak in the EDX spectrum, and extract the total number of peaks and the features of the analysis type in the EDX spectrum to form the feature vector of each characteristic peak; S2: Input the feature vectors of each feature peak into the random forest classifier to obtain the energy spectrum type label with the highest confidence [analysis type + peak type + element type]; S3: Match the corresponding analysis rule from the algorithm library based on the energy spectrum type label; S4: Analyze the EDX spectrum according to the analysis rules to obtain qualitative and / or quantitative analysis results.

2. The material testing method based on energy-dispersive X-ray spectrometer according to claim 1, characterized in that: The types of analysis include: qualitative analysis and quantitative analysis; The peak shapes are classified according to the peak fit: If the fitting residual rate If the percentage is ≤5%, then it is a single peak; If the fitting residual rate is less than 5% If the percentage is ≤15%, it is considered an overlapping peak. If the fitting residual rate A peak value greater than 15% indicates a weak signal. The method for determining the element type is as follows: A peak matching algorithm is used to compare the characteristic peaks with the energy ranges in the standard element characteristic peak energy database; The percentage of standard element characteristic peaks with peak values ​​<2keV and >10keV corresponding to all characteristic peaks is calculated as follows: If the proportion of characteristic peaks with a peak position < 2keV > 60%, then light elements dominate. If the proportion of characteristic peaks with a peak position >10 keV >60%, then heavy elements dominate. The rest are mixed elements.

3. The material testing method based on energy-dispersive X-ray spectrometer according to claim 2, characterized in that, The analysis rules include first calculating the intensity and position of different characteristic peaks based on peak shape, light elements, and heavy elements; then removing false peaks based on the intensity of each characteristic peak; and finally matching qualitative and quantitative algorithms based on the intensity and position of the effective characteristic peaks and the qualitative and / or quantitative analysis of the energy spectrum labels.

4. The material testing method based on energy-dispersive X-ray spectrometer according to claim 2, characterized in that, The algorithm library includes characteristic peak intensity and peak position calculation, which includes single peak intensity and peak position algorithms, overlapping peak intensity and peak position algorithms, and light element intensity and peak position algorithms. The algorithm for determining single-peak intensity and peak position is as follows: Lorentz fitting or Gaussian fitting is performed on the characteristic peak of the single peak to obtain the actual peak intensity. Compared with the actual peak ; The algorithm for determining the intensity and position of overlapping peaks is as follows: Determine the peak positions of the two peaks in the overlapping peaks; Determine a unified fitting interval that covers the bimodal energy range; Construct and initialize the bimodal Lorentz-Gaussian mixture function; Within the fitting interval, peak signal splitting is achieved using the Levenberg-Marquardt iterative optimization algorithm; For the split bimodal peaks, Lorentz fitting and Gaussian fitting were used to obtain the actual peak intensities of each peak, respectively. and actual peak ,Bit ; The light element intensity and peak position algorithm, after calculating the intensity and peak position of single or overlapping peaks, uses an ultra-thin window detector correction method to correct the intensity, specifically as follows: The actual peak intensity of the characteristic peak is adjusted using the correction formula. The correction is performed using the following formula: in, This represents the actual peak intensity of the characteristic peak. Indicates the corrected peak intensity of the characteristic peak, ( , The coefficients are for standard sample calibration, specifically for pure C (graphite) and pure Al2O3 standard samples. This is the peak position value of the characteristic peak.

5. The material testing method based on energy-dispersive X-ray spectrometer according to claim 4, characterized in that, After completing the above calculations of characteristic peak intensity and peak position, the calculation also includes compensation and correction of the characteristic peak intensity based on dead time, specifically: Correcting the actual peak intensity results in a corrected actual peak intensity. satisfy: in, For the death time rate.

6. The material testing method based on energy-dispersive X-ray spectroscopy according to claim 4, characterized in that, After completing the above calculations of characteristic peak intensity and peak position, the process also includes peak position deviation compensation and correction for the characteristic peaks. Specifically: Calibration is performed based on the energy spectrum of standard samples, and a peak position deviation compensation model is established to correct the actual peak position. satisfy: in, This is the theoretical peak value.

7. The material testing method based on energy-dispersive X-ray spectrometer according to claim 2, characterized in that, The algorithm library includes qualitative algorithms, specifically: Obtain the standard peak position of the corresponding characteristic peak from the standard element characteristic peak energy database. ; Calculate the standard peak position Compared with the actual peak absolute value of deviation ; If the absolute value of the deviation <0.1keV, identify it as the element, and complete the elemental qualitative analysis of this characteristic peak; If the absolute value of the deviation If the peak value is ≥0.1 keV, continue comparing it with other elements or determine it as an unknown peak.

8. The material testing method based on energy-dispersive X-ray spectrometer according to claim 2, characterized in that, The algorithm library includes quantitative algorithms, which, based on qualitative algorithms, add the peak area-sensitivity factor method to calculate the content of the j-th element. in, Let be the intensity of the characteristic peak of the j-th element. Let be the sensitivity factor for the j-th element. For high-purity samples with a major element content > 95%, the sensitivity factor is 1.

9. The material testing method based on energy-dispersive X-ray spectroscopy according to any one of claims 1 to 8, characterized in that, It also includes step S51: S51: Perform ZAF correction on the quantitative analysis results to obtain the corrected quantitative results. ZAF correction includes atomic number correction, absorption correction and fluorescence correction. Among them, atomic number correction is ; Absorption correction ; Fluorescence correction is ; in, Indicates the intensity of the secondary fluorescence peak. This indicates the intensity of the primary X-ray characteristic peak.

10. An electronic device, characterized in that, include: A chip, a processor, and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the chip executes the computer instructions, the electronic device performs a material testing method based on an energy-dispersive X-ray spectrometer as described in any one of claims 1 to 9.

Images