Method and system for analyzing impurity content in oxide

Abstract

The invention relates to the field of oxide impurity analysis, in particular to a method and system for analyzing the content of impurities in oxide. The method comprises the following steps: setting oxide impurity samples, constructing a reference impurity database through the oxide impurity samples, and analyzing the reference impurity database to obtain a reference impurity data set; the method comprises the following steps: acquiring a to-be-detected spectrogram sequence of a to-be-detected oxide sample, analyzing the to-be-detected spectrogram sequence to obtain to-be-detected spectrogram subsequences, and constructing a to-be-detected spectrogram event chain; analyzing the spectrogram event chain to be detected to obtain a de-noised spectrogram event chain; analyzing the de-noised spectrogram event chain in combination with a reference impurity data set to obtain a matched impurity data set and an unmatched impurity data set of the to-be-detected oxide sample; and analyzing the matched impurity data set and the unmatched impurity data set to obtain an impurity analysis result of the to-be-detected oxide sample. According to the invention, the impurity identification precision can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of oxide impurity analysis, in particular to an analysis method and system for impurity content in oxide. BACKGROUND

[0002] With the development of material analysis and detection technology, especially in the fields of semiconductor materials, metal oxides, rare earth materials, etc., higher requirements are put forward for the qualitative identification and quantitative analysis of trace impurity components in materials. Small changes in impurity content often have a significant impact on material performance. In particular, in functional ceramics, electrocatalysts, optoelectronic devices and other applications, trace impurities may cause performance degradation such as conductivity, luminescence, catalytic activity, etc.

[0003] A Chinese patent with publication number CN115308347A discloses an analysis method for nitrogen oxide impurities in toptirum, comprising: preparing a standard solution of impurity F; using a liquid chromatograph-mass spectrometer to quantitatively detect impurity F, sample analysis, and record the chromatogram; linear regression is performed with the concentration (x) of impurity F solution as the abscissa and the peak area (y) of impurity F as the ordinate to obtain a linear regression equation and a linear graph; and the linear regression equation or linear graph obtained in step (2) is used to quantitatively determine the impurity F in the test sample by external standard method.

[0004] In the prior art, there are the following problems: the equipment is expensive, the operation is complex, the sample pretreatment is time-consuming, and the recognition ability of complex background interference is limited, and most of them are concentrated in single-point peak judgment or static spectrum analysis, lacking dynamic spectrum analysis ability, resulting in low recognition accuracy and concentration analysis ability in interference environment, which is the problem we need to solve. SUMMARY

[0005] The present application aims to solve the problems in the background art and provides an analysis method for impurity content in oxide.

[0006] The technical scheme of the present application: an analysis method for impurity content in oxide, comprising the following steps: S1, setting an oxide impurity sample, constructing a reference impurity database by the oxide impurity sample, analyzing the reference impurity database, and obtaining a reference impurity data set; S2, obtaining a test spectrum sequence of a test oxide sample, analyzing the test spectrum sequence, obtaining a test spectrum sub-sequence, and constructing a test spectrum event chain; S3, analyzing the test spectrum event chain to obtain a denoised spectrum event chain; combining the reference impurity data set to analyze the denoised spectrum event chain to obtain a matching impurity data set and a non-matching impurity data set of the test oxide sample; S4, analyzing the matching impurity data set and the non-matching impurity data set to obtain an impurity analysis result of the test oxide sample.

[0007] Preferably, the process of setting the oxide impurity samples, constructing the reference impurity database by the oxide impurity samples, and analyzing the reference impurity database to obtain the reference impurity dataset comprises: setting the quantum dot fluorescent probe and the fluorescent detection platform; setting the fixed excitation wavelength of the quantum dot fluorescent probe and the standard detection wavelength range; labeling the oxide impurity samples with the quantum dot fluorescent probe, placing multiple oxide impurity samples in the fluorescent detection platform respectively, exciting the oxide impurity samples under the action of the excitation light source through the quantum dot fluorescent technology, collecting the fluorescence emission spectrum of each oxide impurity sample after excitation, and correlating the fluorescence emission spectrum corresponding to each oxide impurity sample to construct the reference impurity database; performing peak normalization processing, background noise removal and filtering processing, and principal component analysis on the fluorescence emission spectrum of each group of oxide impurity samples with the same impurity and the same gradient concentration in the reference impurity database to obtain the reference emission spectrum; extracting the features of the reference emission spectrum to obtain the spectrum features, constructing the impurity feature vector and the impurity concentration feature relationship through the impurity type and concentration corresponding to the oxide impurity sample and the spectrum features, correlating the impurity feature vector with the corresponding reference emission spectrum through the impurity concentration feature relationship, and obtaining the reference impurity dataset.

[0008] Preferably, the process of obtaining the test spectrum sequence of the oxide sample to be tested and analyzing the test spectrum sequence to obtain the test spectrum sub-sequence comprises: setting the excitation wavelength interval through the fixed excitation wavelength, placing the oxide sample to be tested into the fluorescent detection platform, exciting the oxide sample to be tested in turn within the excitation wavelength interval through the quantum dot fluorescent technology, setting the acquisition time period, obtaining the test spectrum sequence of the oxide sample to be tested after excitation through the acquisition time period, and correlating the test spectrum sequence with the corresponding excitation wavelength interval; extracting the features of the test spectrum sequence to obtain the test spectrum sequence features, dividing the test spectrum sequence according to the change of the test spectrum sequence features, and obtaining the test spectrum sub-sequence.

[0009] Preferably, the process of constructing the test spectrum event chain comprises: setting the single feature threshold, if the difference between the test spectrum sequence features in the same time sequence test spectrum sub-sequence is less than or equal to the single feature threshold, then the corresponding test spectrum sub-sequence is correlated and recorded as a test spectrum sub-sequence set; If the difference between the sequence characteristics of the to-be-tested spectrum within the same time sequence of the to-be-tested spectrum subsequence is greater than the single characteristic threshold value, the corresponding to-be-tested spectrum subsequence and other irrelevant to-be-tested spectrum subsequences are analyzed until all to-be-tested spectrum subsequences have to-be-tested spectrum subsequences associated therewith; The to-be-tested spectrum subsequence set corresponding to the continuous time sequence is associated in the time sequence order to construct a to-be-tested spectrum event chain.

[0010] Preferably, the to-be-tested spectrum event chain is analyzed to obtain a denoised spectrum event chain; the process of analyzing the denoised spectrum event chain in combination with the reference impurity dataset comprises: The time sequence corresponding to the to-be-tested spectrum subsequence set within the to-be-tested spectrum event chain is processed by a dynamic time warping technology to obtain a dynamic time sequence, and the dynamic time sequence is associated with the corresponding to-be-tested spectrum subsequence set to obtain a dynamic spectrum event chain; the to-be-tested spectrum sequence characteristics of the dynamic spectrum event chain are processed and updated by a spectrum difference weighted distance technology to obtain a denoised spectrum event chain; The numerical value corresponding to each spectrum characteristic in the reference impurity dataset and the spectrum characteristic thereof are converted into a vector, denoted as a reference impurity characteristic vector; the numerical value corresponding to each characteristic in the denoised spectrum event chain and the characteristic thereof are converted into a vector, denoted as a to-be-tested spectrum characteristic vector.

[0011] Preferably, the process of further analyzing the denoised spectrum event chain in combination with the reference impurity dataset to obtain a matching impurity dataset and an unmatched impurity dataset of the to-be-tested oxide sample comprises: A cosine similarity threshold value is set; if the cosine similarity between each reference impurity characteristic vector in the reference impurity dataset and each to-be-tested spectrum characteristic vector in the denoised spectrum event chain is greater than or equal to the cosine similarity threshold value, the denoised spectrum event chain matches the reference impurity dataset, and the denoised spectrum event chain and the reference impurity dataset matching therewith are denoted as a matching impurity dataset; If the cosine similarity between each reference impurity characteristic vector in the reference impurity dataset and each to-be-tested spectrum characteristic vector in the denoised spectrum event chain is less than the cosine similarity threshold value, the denoised spectrum event chain does not match the reference impurity dataset, and the denoised spectrum event chain is denoted as an unmatched impurity dataset.

[0012] Preferably, the process of analyzing the matching impurity dataset and the unmatched impurity dataset to obtain an impurity analysis result of the to-be-tested oxide sample comprises: The impurity name and impurity concentration corresponding to the reference impurity dataset in the matching impurity dataset are obtained, denoted as the impurity name and impurity concentration of the denoised spectrum event chain in the matching impurity dataset, and the impurity name and impurity concentration are denoted as the impurity analysis result of the to-be-tested oxide sample; Obtain the impurity name and concentration corresponding to the reference impurity dataset within the unmatched impurity dataset, and denote them as the impurity name and unmatched impurity concentration of the denoised spectrum event chain within the unmatched impurity dataset. Input all reference impurity datasets and impurity concentrations related to the impurity name into the deep learning model for training. Input the unmatched impurity concentration and the denoised spectrum event chain into the trained deep learning model, and output the impurity concentration of the denoised spectrum event chain. Denote the output impurity concentration and the impurity name of the denoised spectrum event chain within the unmatched impurity dataset as the impurity analysis result of the oxide sample to be tested.

[0013] This invention also discloses an analysis system for the impurity content in oxides, including a management center, which is communicatively connected to a reference data module, a test data module, a data matching module, and a data analysis module. The reference data module is used to set oxide impurity samples, construct a reference impurity database using the oxide impurity samples, analyze the reference impurity database, and obtain a reference impurity dataset. The test data module is used to acquire the test spectrum sequence of the oxide sample, analyze the test spectrum sequence, obtain the test spectrum sub-sequence, and construct the test spectrum event chain; The data matching module is used to analyze the event chain of the spectrum under test to obtain the denoised spectrum event chain; and to analyze the denoised spectrum event chain in conjunction with the reference impurity dataset to obtain the matching impurity dataset and the unmatched impurity dataset of the oxide sample under test. The data analysis module is used to analyze the matched and unmatched impurity datasets to obtain the impurity analysis results of the oxide sample to be tested.

[0014] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: by constructing the event chain of the spectrum to be tested and denoising the spectrum sequence to be tested, the accuracy and robustness of feature extraction in spectrum recognition under interference conditions are improved; by analyzing the established reference impurity dataset and the denoised spectrum event chain, and combining the deep learning model to perform concentration analysis on unmatched impurities, integrated intelligent reasoning processing of impurity recognition and concentration analysis is realized, thereby improving the accuracy of impurity recognition. Attached Figure Description

[0015] Figure 1 This is a flowchart of one embodiment of the present invention. Detailed Implementation

[0016] Example 1, as Figure 1 As shown, the present invention proposes a method for analyzing the impurity content in oxides, comprising the following steps: S1, set an oxide impurity sample, construct a reference impurity database through the oxide impurity sample, analyze the reference impurity database, and obtain a reference impurity dataset; S2, obtain a to-be-tested spectrum sequence of a to-be-tested oxide sample, analyze the to-be-tested spectrum sequence, obtain a to-be-tested spectrum sub-sequence, and construct a to-be-tested spectrum event chain; S3, analyze the to-be-tested spectrum event chain to obtain a denoised spectrum event chain; analyze the denoised spectrum event chain in combination with the reference impurity dataset to obtain a matching impurity dataset and an unmatched impurity dataset of the to-be-tested oxide sample; S4, analyze the matching impurity dataset and the unmatched impurity dataset to obtain an impurity analysis result of the to-be-tested oxide sample.

[0017] It needs to be further explained that, in the specific implementation process, the process of setting an oxide impurity sample, constructing a reference impurity database through the oxide impurity sample, analyzing the reference impurity database, and obtaining a reference impurity dataset is as follows: The oxide impurity sample is prepared by selecting impurity elements to prepare multiple doped oxide samples, and multiple different concentration gradients are set for each impurity; Specifically, multiple copies of each group of oxide impurity samples of the same impurity and the same gradient concentration are repeatedly prepared to reduce experimental errors and improve data reliability; the oxide is zinc oxide; A quantum dot fluorescent probe and a fluorescent detection platform are set, the fluorescent detection platform is used to record the image of the oxide impurity sample after excitation; a fixed excitation wavelength and a standard detection wavelength range of the quantum dot fluorescent probe are set, and the excitation light power and the incident angle of the quantum dot fluorescent probe are controlled; The oxide impurity sample is labeled with a quantum dot fluorescent probe, multiple oxide impurity samples are placed in the fluorescent detection platform, the oxide impurity sample is excited under the action of an excitation light source through quantum dot fluorescence technology, and the fluorescence emission spectrum of each oxide impurity sample after excitation is collected; the fluorescence emission spectrum corresponding to each oxide impurity sample is associated to construct a reference impurity database; The fluorescence emission spectrum of each group of oxide impurity samples of the same impurity and the same gradient concentration in the reference impurity database is subjected to peak normalization processing, background noise removal and filtering processing, and principal component analysis to obtain a reference emission spectrum; The reference emission spectrum is subjected to feature extraction to obtain a spectrum feature, which includes but is not limited to a fluorescence peak shift amount, a half-peak width, a peak position, and a fluorescence intensity; The impurity feature vector and the impurity concentration feature relationship are constructed by the impurity type and concentration corresponding to the oxide impurity sample and the spectral feature; the impurity concentration feature relationship refers to the mapping relationship between the impurity concentration and the corresponding spectral feature; the reference impurity data set is obtained by associating the impurity feature vector with the corresponding reference emission spectrum through the impurity concentration feature relationship. Specifically, the spectral feature and its corresponding numerical value of the oxide impurity sample are stored in the impurity feature vector, and are stored in the form of a matrix in the impurity feature vector.

[0018] It needs to be further explained that in the specific implementation process, the measured spectral sequence of the measured oxide sample is obtained, the measured spectral sequence is analyzed to obtain the measured spectral sub-sequence, and the measured spectral event chain is constructed. The measured oxide sample refers to zinc oxide required for impurity content analysis; the excitation wavelength interval is set by fixing the excitation wavelength, the measured oxide sample is placed in the fluorescence detection platform, the measured oxide sample is excited in turn in the excitation wavelength interval by quantum dot fluorescence technology, the acquisition time period is set, the measured spectral sequence of the measured oxide sample after excitation is obtained through the acquisition time period, and the measured spectral sequence is associated with the corresponding excitation wavelength interval; the characteristics of the measured spectral sequence are extracted to obtain the measured spectral sequence characteristics, which include but are not limited to peak value change, half-peak width and peak position shift change; the measured spectral sequence is divided according to the change of the measured spectral sequence characteristics to obtain the measured spectral sub-sequence, and each measured spectral sequence characteristic in the measured spectral sub-sequence is unchanged; The single feature threshold is set, which includes but is not limited to the peak value deviation threshold and the peak position deviation threshold, if the difference between the measured spectral sequence characteristics in the measured spectral sub-sequence with the same time sequence is less than or equal to the single feature threshold, the corresponding measured spectral sub-sequence is associated and recorded as a measured spectral sub-sequence set; Specifically, the spectral variation of the same oxide impurity generally has consistency and similarity, so all single feature differences in the measured spectral sub-sequence are compared and analyzed with the single feature threshold at the same time. If the difference between the measured spectral sequence characteristics in the measured spectral sub-sequence with the same time sequence is greater than the single feature threshold, the corresponding measured spectral sub-sequence is analyzed with other unrelated measured spectral sub-sequences until all the measured spectral sub-sequences have associated measured spectral sub-sequences; Specifically, multiple quantum dot fluorescence technology experiments are performed on the measured oxide sample, and for the same impurity, there are multiple measured spectral sequences, so the analysis of the measured spectral sub-sequence has at least two and more measured spectral sub-sequences associated; corresponding to the time sequence, to construct a to-be-tested spectrum event chain; Specifically, for the process of constructing the to-be-tested spectrum event chain, since the spectrum change rule of the same oxide impurity generally has consistent similarity, at least two and more to-be-tested spectrum subsequence sets are associated.

[0019] It needs to be further explained that, in the specific implementation process, the to-be-tested spectrum event chain is analyzed to obtain a denoised spectrum event chain; the denoised spectrum event chain is analyzed in combination with the reference impurity dataset to obtain a matched impurity dataset of the to-be-tested oxide sample and an unmatched impurity dataset. The time sequence corresponding to the to-be-tested spectrum subsequence set in the to-be-tested spectrum event chain is processed by a dynamic time warping technology to obtain a dynamic time sequence, and the dynamic time sequence is associated with the corresponding to-be-tested spectrum subsequence set to obtain a dynamic spectrum event chain; the to-be-tested spectrum sequence features of the dynamic spectrum event chain are processed and updated by a spectrum difference weighted distance technology to obtain a denoised spectrum event chain. The values corresponding to the spectrum features in the reference impurity dataset and the spectrum features are converted into vectors, denoted as reference impurity feature vectors; the values corresponding to the features in the denoised spectrum event chain and the features are converted into vectors, denoted as to-be-tested spectrum feature vectors. A cosine similarity threshold is set, if the cosine similarity between each reference impurity feature vector in the reference impurity dataset and each to-be-tested spectrum feature vector in the denoised spectrum event chain is greater than or equal to the cosine similarity threshold, the denoised spectrum event chain matches the reference impurity dataset, and the denoised spectrum event chain and the reference impurity dataset matched therewith are denoted as a matched impurity dataset; If the cosine similarity between each reference impurity feature vector in the reference impurity dataset and each to-be-tested spectrum feature vector in the denoised spectrum event chain is less than the cosine similarity threshold, the denoised spectrum event chain does not match the reference impurity dataset, the denoised spectrum event chain is analyzed with other reference impurity datasets to obtain the maximum value of the number of reference impurity feature vectors whose cosine similarity with the corresponding to-be-tested spectrum feature vector is greater than or equal to the cosine similarity threshold, and the reference impurity dataset corresponding to the maximum value is obtained, which is denoted as an unmatched impurity dataset.

[0020] It needs to be further explained that, in the specific implementation process, the matched impurity dataset and the unmatched impurity dataset are analyzed to obtain the impurity analysis result of the to-be-tested oxide sample. Obtaining the impurity name and impurity concentration corresponding to the reference impurity data set in the matched impurity data set, denoted as the impurity name and impurity concentration of the denoised spectrum event chain in the matched impurity data set, and recording the impurity name and impurity concentration as the impurity analysis result of the oxide sample to be tested; Obtaining the impurity name and impurity concentration corresponding to the reference impurity data set in the matched impurity data set, denoted as the impurity name and impurity concentration of the denoised spectrum event chain in the matched impurity data set, and recording the impurity name and impurity concentration as the impurity analysis result of the oxide sample to be tested;

[0021] In the second embodiment, the oxide impurity content analysis system is applied to the oxide impurity content analysis method in the first embodiment, and specifically includes a management center, which is communicatively connected with a reference data module, a to-be-tested data module, a data matching module, and a data analysis module. The reference data module is used to set oxide impurity samples, construct a reference impurity database through the oxide impurity samples, analyze the reference impurity database, and obtain a reference impurity data set; The to-be-tested data module is used to obtain a to-be-tested spectrum sequence of the oxide sample to be tested, analyze the to-be-tested spectrum sequence, obtain a to-be-tested spectrum sub-sequence, and construct a to-be-tested spectrum event chain; The data matching module is used to analyze the to-be-tested spectrum event chain to obtain a denoised spectrum event chain, and analyze the denoised spectrum event chain in combination with the reference impurity data set to obtain a matched impurity data set and an unmatched impurity data set of the oxide sample to be tested; The data analysis module is used to analyze the matched impurity data set and the unmatched impurity data set to obtain an impurity analysis result of the oxide sample to be tested.

[0022] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.

Claims

1. A method of analyzing the content of impurities in an oxide, characterized in that, Includes the following steps: S1. Set up oxide impurity samples, construct a reference impurity database using the oxide impurity samples, analyze the reference impurity database, and obtain a reference impurity dataset; S2. Obtain the spectrum sequence of the oxide sample to be tested, analyze the spectrum sequence to be tested, obtain the sub-sequence of the spectrum to be tested, and construct the event chain of the spectrum to be tested. S3. Analyze the event chain of the spectrum to be tested to obtain the event chain of the denoised spectrum; By combining the reference impurity dataset with the analysis of the event chain of the denoised spectrum, the matching impurity dataset and the unmatched impurity dataset of the oxide sample to be tested are obtained; S4. Analyze the matched impurity dataset and the unmatched impurity dataset to obtain the impurity analysis results of the oxide sample to be tested.

2. The method for analyzing the content of impurities in an oxide according to claim 1, characterized by, The process of setting up oxide impurity samples, constructing a reference impurity database using these samples, and analyzing the reference impurity database to obtain a reference impurity dataset includes: Set up a quantum dot fluorescent probe and a fluorescence detection platform; set the fixed excitation wavelength and standard detection wavelength range for the quantum dot fluorescent probe; Oxide impurity samples were labeled using quantum dot fluorescent probes. Multiple oxide impurity samples were placed in a fluorescence detection platform. The oxide impurity samples were excited by the quantum dot fluorescence technology under the action of an excitation light source, and the fluorescence emission spectra of each oxide impurity sample after excitation were collected. The fluorescence emission spectra of each oxide impurity sample were correlated to construct a reference impurity database. The fluorescence emission spectra of multiple oxide impurity samples with the same impurity and the same gradient concentration in the reference impurity database were processed by peak normalization, background noise removal and filtering, and principal component analysis to obtain the reference emission spectrum. Feature extraction is performed on the reference emission spectrum to obtain spectral features; impurity feature vectors and impurity concentration feature relationships are constructed by using the impurity type and concentration corresponding to the oxide impurity samples and the spectral features; the impurity feature vectors are associated with the corresponding reference emission spectra through the impurity concentration feature relationships to obtain the reference impurity dataset.

3. The method for analyzing the content of impurities in an oxide according to claim 2, characterized by, The process of obtaining the spectral sequence of the oxide sample to be tested, and analyzing the spectral sequence to obtain the sub-sequence of the spectral sequence includes: By setting a fixed excitation wavelength range, the oxide sample to be tested is placed in the fluorescence detection platform. Using quantum dot fluorescence technology, the oxide sample to be tested is sequentially excited within the excitation wavelength range. A collection time period is set, and the spectrum sequence of the oxide sample after excitation is obtained through the collection time period. The spectrum sequence is then associated with the corresponding excitation wavelength range. Feature extraction is performed on the spectrum sequence to obtain its features. The spectrum sequence is then divided based on the changes in its features to obtain sub-sequences.

4. The method for analyzing the content of impurities in an oxide according to claim 3, characterized by, The process of constructing the event chain of the spectrum to be tested is as follows: If the differences between features of the test spectrum subsequences with the same time series are all less than or equal to the single feature threshold, then the corresponding test spectrum subsequences are associated and recorded as a set of test spectrum subsequences. If the difference between the sequence characteristics of the to-be-tested spectrum in the to-be-tested spectrum subsequence with the same time sequence is greater than the single feature threshold value, the corresponding to-be-tested spectrum subsequence and other irrelevant to-be-tested spectrum subsequences are analyzed until all to-be-tested spectrum subsequences have to-be-tested spectrum subsequences associated therewith; The to-be-tested spectrum subsequence sets corresponding to the continuous time sequence are associated in the time sequence order to construct a to-be-tested spectrum event chain.

5. The method for analyzing the content of impurities in an oxide according to claim 4, characterized by, The to-be-tested spectrum event chain is analyzed to obtain a denoised spectrum event chain; The process of analyzing the denoised spectrum event chain in combination with the reference impurity dataset includes: The time sequence corresponding to the to-be-tested spectrum subsequence set in the to-be-tested spectrum event chain is processed by a dynamic time warping technology to obtain a dynamic time sequence, and the dynamic time sequence is associated with the corresponding to-be-tested spectrum subsequence set to obtain a dynamic spectrum event chain; the to-be-tested spectrum sequence characteristics of the dynamic spectrum event chain are processed and updated by a spectral difference weighted distance technology to obtain a denoised spectrum event chain; The values corresponding to the spectrum characteristics in the reference impurity dataset and the spectrum characteristics thereof are converted into vectors, denoted as reference impurity feature vectors; the values corresponding to the characteristics in the denoised spectrum event chain and the characteristics thereof are converted into vectors, denoted as to-be-tested spectrum feature vectors.

6. The method for analyzing the content of impurities in an oxide according to claim 5, characterized by, The process of further analyzing the denoised spectrum event chain in combination with the reference impurity dataset to obtain a matching impurity dataset and an unmatched impurity dataset of the to-be-tested oxide sample includes: A cosine similarity threshold value is set, if the cosine similarity between each reference impurity feature vector in the reference impurity dataset and each to-be-tested spectrum feature vector in the denoised spectrum event chain is greater than or equal to the cosine similarity threshold value, the denoised spectrum event chain matches the reference impurity dataset, and the denoised spectrum event chain and the reference impurity dataset matched therewith are denoted as a matching impurity dataset; If the cosine similarity between each reference impurity feature vector in the reference impurity dataset and each to-be-tested spectrum feature vector in the denoised spectrum event chain is less than the cosine similarity threshold value, the denoised spectrum event chain does not match the reference impurity dataset, and the denoised spectrum event chain is denoted as an unmatched impurity dataset.

7. The method of claim 6, wherein the oxide is a titanium oxide. The process of analyzing the matching impurity dataset and the unmatched impurity dataset to obtain the impurity analysis result of the to-be-tested oxide sample includes: The impurity name and impurity concentration corresponding to the reference impurity dataset in the matching impurity dataset are obtained, denoted as the impurity name and impurity concentration of the denoised spectrum event chain in the matching impurity dataset, and the impurity name and impurity concentration are denoted as the impurity analysis result of the to-be-tested oxide sample; The reference impurity data set corresponding to the impurity name and the impurity concentration in the unmatched impurity data set is obtained, denoted as the impurity name of the denoising spectrum event chain in the unmatched impurity data set and the unmatched impurity concentration, all reference impurity data sets related to the impurity name and the impurity concentration are input into the deep learning model for training, the unmatched impurity concentration and the denoising spectrum event chain are input into the trained deep learning model, and the impurity concentration of the denoising spectrum event chain is output. The output impurity concentration and the impurity name of the denoising spectrum event chain in the unmatched impurity data set are denoted as the impurity analysis result of the to-be-tested oxide sample.

8. A system for analyzing the content of impurities in an oxide, in particular for use in the method for analyzing the content of impurities in an oxide according to any one of claims 1 to 7, comprising a management center, characterized in that, The management center is communicatively connected with a reference data module, a to-be-tested data module, a data matching module, and a data analysis module: The reference data module is used to set an oxide impurity sample, construct a reference impurity database through the oxide impurity sample, analyze the reference impurity database, and obtain a reference impurity data set; The to-be-tested data module is used to obtain a to-be-tested spectrum sequence of a to-be-tested oxide sample, analyze the to-be-tested spectrum sequence, obtain a to-be-tested spectrum sub-sequence, and construct a to-be-tested spectrum event chain; The data matching module is used to analyze the to-be-tested spectrum event chain and obtain a denoising spectrum event chain; The denoising spectrum event chain is analyzed in combination with the reference impurity data set to obtain a matching impurity data set and an unmatched impurity data set of the to-be-tested oxide sample; The data analysis module is used to analyze the matching impurity data set and the unmatched impurity data set to obtain an impurity analysis result of the to-be-tested oxide sample.

Images