Substance identification methods and devices, storage media and program products

By constructing a machine learning model, the medium-volatile and semi-volatile organic components in gas samples are identified using standard substance mass spectrometry data. This solves the problems of component co-eluting and insufficient identification in traditional methods, and achieves efficient component identification and air quality simulation.

CN119719763BActive Publication Date: 2026-11-17TSINGHUA UNIVERSITY
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411651561.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-11-18
Publication Date
2026-11-17
Estimated Expiration
2044-11-18

AI Technical Summary

Technical Problem

Traditional methods struggle to accurately identify moderately volatile and semi-volatile organic compounds (I/SVOCs) in complex gas samples. In particular, the co-eluting of components due to limitations in separation performance means that current technologies cannot fully utilize mass spectrometry features, resulting in a large number of components remaining unidentified and unclassified, thus affecting air quality simulation results.

Method used

A machine learning model is constructed by acquiring mass spectrometry data of standard substances, building feature vectors and labels, training the machine learning model, and using mass spectrometry data to identify elements and functional groups, thereby achieving non-targeted automatic identification of complex I/SVOC components.

Benefits of technology

It significantly improves the proportion of qualitatively classified components and the resolution of functional group composition of full-spectrum components, enabling rapid and accurate identification of all effluent components and improving the accuracy of air quality simulation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119719763B_ABST
    Figure CN119719763B_ABST
Patent Text Reader

Abstract

A substance identification method, apparatus, storage medium, and program product are applied to the identification of moderately volatile and semi-volatile organic components. The method includes: acquiring mass spectrometry data of the substance to be tested; constructing a feature vector based on the mass spectrometry data; inputting the feature vector into a machine learning model to obtain the probability of each element label and each structural label. Element labels include oxygen, nitrogen, sulfur, and silicon; structural labels include aliphatic functional groups, aromatic functional groups, oxygen-containing structural functional groups, nitrogen-containing structural functional groups, and sulfur-containing structural functional groups. Aromatic functional groups include single benzene rings, two benzene rings, multiple benzene rings, and heteroaromatic rings; aliphatic functional groups include alkyl groups, olefin bonds, aliphatic rings, and heteroaliphatic rings; determining the element and functional groups of the substance to be tested based on the probabilities of the element and structural labels; and determining the first, second, and third structures based on the element and functional groups of the substance to be tested, thereby determining the substance category.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to, but is not limited to, the fields of mass spectrometry and chemical analysis technology, and particularly to a substance identification method and apparatus, storage medium and program product. Background Technology

[0002] Gas samples, such as ambient air and vehicle exhaust, contain a wide variety of complex intermediate and semi-volatile organic compounds (I / SVOCs). These components have been proven to significantly contribute to the formation of secondary air pollution (especially secondary organic aerosols). Their separation and accurate identification and classification are crucial for subsequent air quality simulation and assessment of the contributions of each organic component and emission source to secondary air pollution. Traditional I / SVOC component identification methods involve separation by one-dimensional gas chromatography followed by mass spectrometry, comparing the eluted peak mass spectra with a standard spectral library to identify the most likely single species composition of the eluted component. Due to limitations in separation efficiency, co-elution of components is significant, and most peaks cannot be accurately matched. Two-dimensional gas chromatography-mass spectrometry (GC-MS) significantly improves chromatographic separation capabilities and alleviates co-elution problems, providing a data foundation for the accurate identification of complex I / SVOCs. To automatically process more elution peak data, some studies have developed non-targeted automatic identification methods for all components based on chromatographic retention time and the characteristics of major mass spectrometry peaks. However, these methods fail to fully utilize mass spectrometry features, and the classification criteria are all linear relationships or thresholds of the intensity of a single mass spectrometry peak. Therefore, only about 50% of the components can be classified relatively accurately, while the categories and compositions of the remaining species remain unknown. Summary of the Invention

[0003] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0004] This disclosure provides a substance identification method, including: applied to the identification of moderately volatile and semi-volatile organic components, the substance identification method comprising: Obtain mass spectrometry data of the substance to be tested, and construct a feature vector of the substance to be tested based on the mass spectrometry data of the substance to be tested; The feature vector of the substance to be tested is input into a machine learning model to obtain the probability of each element label and the probability of each structure label. The machine learning model is constructed as follows: Training samples are obtained, including multiple standard mass spectra; a feature vector of the training samples is constructed, comprising the relative peak intensities of multiple mass-to-charge ratio positions of the standard mass spectra; training sample labels are constructed, including element labels and structure labels. The element labels include oxygen, nitrogen, sulfur, and silicon; the structure labels include multiple aliphatic functional groups, multiple aromatic functional groups, multiple oxygen-containing functional groups, multiple nitrogen-containing functional groups, and multiple sulfur-containing functional groups. The aromatic functional groups include a single benzene ring, two benzene rings, multiple benzene rings, and heteroaromatic rings; the aliphatic functional groups include alkyl groups, olefin bonds, aliphatic rings, and heteroaliphatic rings; a machine learning model is constructed, and trained and tested based on the feature vector and training sample labels of the training samples. Based on the probability of each element tag and the probability of each structure tag, the elements and functional groups included in the substance to be tested are determined. Based on the elements and functional groups included in the substance to be tested, determine the first, second, and third structures of the substance to be tested, and determine the substance category of the substance to be tested according to the following table: .

[0005] This disclosure also provides a substance identification device, including a memory; and a processor connected to the memory, the memory being used to store instructions, the processor being configured to execute the steps of the substance identification method according to any embodiment of this disclosure based on the instructions stored in the memory.

[0006] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the substance identification method described in any embodiment of this disclosure.

[0007] This disclosure also provides a program product including instructions that, when executed by a computer, perform a substance identification method as described in any embodiment of this disclosure.

[0008] The substance identification method, apparatus, storage medium, and program product of this disclosure apply machine learning methods to mass spectrometry data. Compared with traditional I / SVOC identification technology, this significantly improves the proportion of qualitatively classified components and enhances the resolution of functional group composition of the full spectrum components. This disclosure enables rapid and accurate identification of large samples of functional group composition of all eluting components with machine learning assistance. The interpretability of emission characteristics of complex I / SVOC components is significantly improved compared to previous traditional identification methods, especially for components with low spectral library similarity, providing their functional group composition and determining whether the elution peak should be considered an emission.

[0009] Other features and advantages of this disclosure will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the disclosure. Other advantages of this disclosure may be realized and obtained by means of the methods described in the description and the accompanying drawings. Attached Figure Description

[0010] The accompanying drawings are used to provide an understanding of the technical solutions of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the technical solutions of this disclosure and do not constitute a limitation on the technical solutions of this disclosure.

[0011] Figure 1 A flowchart illustrating a model building method provided for an exemplary embodiment of this disclosure; Figure 2 A schematic flowchart of a substance identification method provided for an exemplary embodiment of this disclosure; Figure 3 A schematic diagram illustrating the workflow of a machine learning-assisted I / SVOC component screening and classification method provided for an exemplary embodiment of this disclosure; Figure 4 A schematic diagram illustrating a detailed explanation of screening and classifying components (taking methyl phenyl ketone as an example) provided for an exemplary embodiment of this disclosure; Figure 5 A schematic diagram of a model building apparatus provided for an exemplary embodiment of this disclosure; Figure 6 A schematic diagram of the structure of a substance identification device provided for an exemplary embodiment of this disclosure; Figure 7 A schematic diagram of another model building apparatus provided as an exemplary embodiment of this disclosure; Figure 8 A schematic diagram of another substance identification device provided as an exemplary embodiment of this disclosure. Detailed Implementation

[0012] This disclosure describes several embodiments, but these descriptions are exemplary and not limiting, and it will be apparent to those skilled in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with, or may replace, any feature or element of any other embodiment.

[0013] This disclosure includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features, and elements disclosed in this disclosure may also be combined with any conventional features or elements to form a unique inventive scheme as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive schemes to form another unique inventive scheme as defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this disclosure may be implemented individually or in any suitable combination. Therefore, the embodiments are not limited except by the limitations imposed by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.

[0014] Furthermore, in describing representative embodiments, the specification may have presented methods and / or processes as a specific sequence of steps. However, the method or process should not be limited to the specific order of steps described herein, to the extent that the method or process does not depend on the specific order of steps described herein. As will be understood by those skilled in the art, other sequences of steps are also possible. Therefore, the specific order of steps set forth in the specification should not be construed as a limitation of the claims. Moreover, the claims relating to the method and / or process should not be limited to the steps performed in the order written, and those skilled in the art will readily understand that these orders can be varied and still remain within the spirit and scope of the embodiments disclosed herein.

[0015] I / SVOC species with complex compositions contribute significantly to secondary atmospheric pollution, necessitating accurate assessment of their environmental impact through air quality models. Current I / SVOC separation and identification methods suffer from limitations such as insufficient utilization of spectral information and unclear understanding of the relationship between mass spectrometry and material structure, resulting in a large number of components remaining unidentified and unclassified. This may affect the understanding of the emission characteristics of I / SVOC components and consequently impact subsequent air quality simulation results.

[0016] Machine learning methods are well-suited for processing large-sample, multi-feature data, excel at uncovering nonlinear relationships between variable features and predicted labels, and allow for adjustments to input and output formats to meet specific application needs. They are particularly suitable for learning mass spectrometry features for substance identification. The main steps in applying machine learning methods to instrumental analysis data are: training a model on a standard database to learn the relationship between labels such as substance structure and physicochemical properties and the model input (mass spectrometry features); and then applying the model to unknown samples after validation to identify substances or predict their physicochemical properties. The application of machine learning in instrumental analysis data for substance identification is widespread across various research fields: in targeted identification, machine learning methods are used to learn the mass spectrometry features of target substances (such as proteins and genes) and predict their presence or absence in samples to identify species, diagnose diseases, etc. (medicine, metabolomics, botany, etc.); in non-targeted identification, studies using machine learning models to predict the toxicity of organic components in water samples have been reported and have demonstrated good predictive performance.

[0017] Machine learning methods have diverse applications in assisting instrumental analysis data processing and non-targeted substance identification, but their application in the atmospheric environment field, especially in assisting the non-targeted identification and classification of I / SVOC components, is relatively limited. Furthermore, previous machine learning methods have mostly focused on component identification based on single chromatographic peaks, primarily targeting mass spectrometry with high library similarity. This approach is unsuitable for identifying components with low library similarity (mainly due to poor mass spectrometry quality), and lacks automatic identification methods for full-component mass spectra. For I / SVOC components, their complex composition makes it difficult to cover all components during model training. Additionally, air quality simulations do not require resolution at the individual component level for I / SVOC inputs; I / SVOC components are integrated and classified according to indicators such as volatility and oxygen-to-carbon ratio before being input into the model. Therefore, the identification objective of this disclosure for air quality simulation results in a model output that differs from previous techniques.

[0018] like Figure 1 As shown, this disclosure provides a model building method, including: Step 101: Obtain training samples, which include multiple standard substance mass spectra; Step 102: Construct the feature vector of the training sample, which includes the relative peak intensities of multiple mass-to-charge ratio positions of the standard material mass spectrum. Step 103: Construct training sample labels, which include element labels and structure labels. Element labels include multiple element categories, and structure labels include multiple functional group structure categories. Step 104: Build a machine learning model, and train and test the machine learning model based on the feature vectors and labels of the training samples.

[0019] The model construction method of this disclosure involves acquiring training samples, including multiple standard substance mass spectra, constructing feature vectors for the training samples (comprising relative peak intensities at multiple mass-to-charge ratio positions in the standard substance mass spectra), constructing training sample labels (including elemental and structural labels; elemental labels include multiple elemental categories, and structural labels include multiple functional group structural categories), constructing a machine learning model, and training and validating the machine learning model based on the feature vectors and labels of the training samples. This applies machine learning methods to mass spectrometry data, significantly improving the proportion of qualitatively classified components and enhancing the resolution of functional group composition across the entire spectrum compared to traditional I / SVOC identification techniques. This method is highly versatile; the model training part can flexibly adjust the input feature dimensions according to the mass number range of the input spectrum. At the application level, it is applicable to all classification problems involving automatic non-targeted identification of substance composition based on mass spectrometry.

[0020] Traditional library comparison is only applicable to high-similarity mass spectrometry, failing to identify components absent from the library, co-eluting, or interfering with impurities. Automatic non-targeted identification methods do not fully exploit the full-spectrum information and its relationship with functional group composition, resulting in many unidentified components remaining in traditional I / SVOC analysis methods. The machine learning model constructed using the model building method of this disclosure is applicable to all mass spectrometry, including high and low library similarity, and aims to identify the functional group composition of the target mass spectrum rather than individual substances. Compared to other environmental methods that use machine learning to predict component properties, the machine learning model constructed using the model building method of this disclosure directly predicts the functional group composition of components, enabling flexible identification of physicochemical properties based on functional group composition.

[0021] In some exemplary embodiments, in step 101, multiple mass spectra of standard materials can be exported in batches from the National Institute of Standards and Technology (NIST) database.

[0022] For example, PyautoGUI can be used to automatically export standard mass spectra constructed by NIST in batches. Each spectrum includes the substance name, chemical formula, and a mass spectrum presented as a string of peak position-to-peak relative intensity pairs. More than 200,000 mass spectra can be selected for training samples.

[0023] In some exemplary embodiments, in step 102, the plurality of mass-to-charge ratio positions are all integer mass-to-charge ratio positions within a preset mass-to-charge ratio range.

[0024] For example, the preset mass-to-charge ratio range can be between 35 and 550; however, this disclosure does not limit it. The preset mass-to-charge ratio range can be set as needed.

[0025] In this embodiment of the disclosure, the multiple mass-to-charge ratio positions can be any integer mass-to-charge ratio positions within the range of 35 to 550, and the peak intensity of each mass-to-charge ratio position is the sum of the peak intensities of all mass-to-charge ratio positions within the range of ±0.5 of that mass-to-charge ratio. However, this disclosure does not impose any limitations on this.

[0026] In other examples, multiple mass-to-charge ratio (MTR) positions can be multiple MTR positions spaced apart by a preset step size within a preset MTR range. Assuming the preset step size is b, where b is a real number greater than 0, the peak intensity of each MTR position is the sum of the peak intensities of all MTRs within the range of ±b / 2 of that MTR. For example, the preset step size b can be 2 or 4, and the peak intensity of each MTR position is the sum of the peak intensities of all MTRs within the range of ±1 or ±2 of that MTR. When the initial MTR position is 35, the final MTR position is 550, and the preset step size is 1, the multiple MTR positions are all integer MTR positions within the range of 35 to 550.

[0027] In some exemplary embodiments, in step 102, the feature vector of the training sample further includes the positions of the first to M base peaks of the standard substance mass spectrometer, and the relative intensities of the second to M base peaks relative to the first base peak, where M is a natural number greater than 1.

[0028] For example, M can be 6; however, this disclosure does not limit this, and the value of M can be set as needed.

[0029] In this embodiment of the disclosure, when constructing the feature vector of the training sample, the string of the peak position-relative intensity pairing is transcribed into a multi-dimensional vector of the relative peak intensity corresponding to each integer mass-charge ratio position, and the positions of the first M base peaks (the peaks with the highest relative intensity to the Mth highest) and the relative base peak intensities (the ratio of the intensity of the second to sixth base peaks to the intensity of the first base peak) are identified as new features and together constitute the feature vector of the training sample.

[0030] In some exemplary embodiments, in step 103, constructing training sample labels includes: For each standard substance mass spectrometer, perform the following operations: Obtain the substance name and chemical formula corresponding to the mass spectrometry of the standard substance, determine the simplified molecular linear input canonical string corresponding to the substance name, determine the functional group structure category contained in the standard substance based on the simplified molecular linear input canonical string, determine the element category contained in the standard substance based on the chemical formula, and generate training sample labels corresponding to the mass spectrometry of the standard substance based on the determined functional group structure category and element category.

[0031] In this embodiment, when constructing training sample labels, the derived substance names and chemical formulas are searched for Simplified Molecular Linear Input Specification (SMILES) strings in the Pubchem chemical information database, which contain the structural composition information of the substance. Then, the open-source chemical information computing package RDKit is used to search for the existence of a specific structure among the substance structures given by SMILES. Finally, the functional group structure information of all standard substances matching SMILES is constructed (0 indicates no presence, 1 indicates presence). Furthermore, the presence of a certain element (e.g., C, O, N, S, Si, etc.) based on the chemical formula itself is also included in the label construction.

[0032] In this embodiment, the training sample labels include both element labels and functional group structure labels. When predicting the test sample, predictions of both element labels and functional group structure labels can be combined. Typically, if a specific functional group structure is predicted to exist, the corresponding element label will also be predicted as positive; therefore, functional group structure labels can be used as the primary basis for prediction. In the final species classification, either element labels or functional group structure labels can be selected as the criterion.

[0033] For example, the element categories include oxygen (O), nitrogen (N), sulfur (S), and silicon (Si). In this embodiment of the disclosure, the element category does not include a separate carbon element because the machine learning model of this disclosure is mainly for detecting organic matter, all of which contain carbon; carbon typically appears in the form of carbon chains, meaning that carbon prediction is characterized by whether or not a carbon chain is present.

[0034] For example, the structural tag includes multiple aliphatic functional groups, multiple aromatic functional groups, multiple oxygen-containing structural functional groups, multiple nitrogen-containing structural functional groups, and multiple sulfur-containing structural functional groups.

[0035] For example, aliphatic functional groups may include alkyl (saturated carbon chain), olefinic bonds, aliphatic rings, heteroaliphatic rings; however, this disclosure does not limit them.

[0036] For example, aromatic functional groups may include: aromatic rings, heteroaromatic rings, phenolic hydroxyl groups, benzyl groups, 0 benzene rings, 1 benzene ring, 2 benzene rings, or multiple benzene rings; however, this disclosure does not limit this.

[0037] For example, oxygen-containing functional groups may include hydroxyl, ether, aldehyde, ketone carbonyl, carboxyl, and ester groups; however, this disclosure does not limit them.

[0038] For example, nitrogen-containing functional groups may include amides, amines, imines, cyano groups, nitro groups, and nitroso groups; however, this disclosure does not limit them.

[0039] For example, sulfur-containing functional groups may include sulfides, sulfones, sulfoxides, and disulfide bonds; however, this disclosure does not limit them.

[0040] In some exemplary embodiments, the machine learning model in step 104 can be an XGBoost machine learning model; however, this disclosure is not limiting. In other examples, the machine learning model can also be a random forest machine learning model, a convolutional neural network (CNN) machine learning model, etc.

[0041] For example, the machine learning model disclosed herein employs the XGBoost machine learning model, which is trained based on the training sample features and labels constructed above (multi-classification model). The training samples can be classified into training and test sets in a 4:1 ratio, and K-fold cross-validation is used (for example, K can be 5); the classification performance is evaluated using accuracy (P), recall (R), and F1 score.

[0042] In this embodiment, the prediction result directly output by the model is the probability of each label (0-1 continuous distribution) presented as a floating-point number. When verifying the model's prediction performance, a threshold of 0.5 is used to convert the above probability into a label prediction result with a two-point distribution of 0-1. However, in practical applications, the model is instructed to directly output the classification prediction probability (i.e., the probability of containing the structure corresponding to the label) so that the probability threshold can be adjusted based on the recognition result when identifying functional group composition.

[0043] like Figure 2 As shown in the embodiments of this disclosure, a method for identifying a substance is also provided, including: Step 201: Obtain the mass spectrometry data of the substance to be tested, and construct the feature vector of the substance to be tested based on the mass spectrometry data. Step 202: Input the feature vector of the substance to be tested into the machine learning model to obtain the probability of each element label and the probability of each structure label. The machine learning model is constructed using the model construction method described in the previous step. Step 203: Determine the elements and functional groups included in the substance to be tested based on the probability of each element tag and the probability of each structure tag. Step 204: Determine the substance category of the substance to be tested based on the elements and functional groups included in the substance.

[0044] The substance identification method of this disclosure can achieve rapid and accurate identification of the functional group composition of all effluent components with machine learning assistance. The interpretability of the emission characteristics of complex I / SVOC components is significantly improved compared with the traditional identification method. In particular, it can give the functional group composition of components with low spectral library similarity and use the functional group composition to determine whether the effluent peak should be regarded as emission.

[0045] The substance identification method of this disclosure can provide more accurate categorized emission inputs for air quality simulation, thereby updating the air quality impact and source apportionment results of I / SVOCs. The functional group composition and species classification provided by this method can guide the detailed development of functional group sub-functional groups in the organic component simulation module of the air quality model.

[0046] The substance identification method of this disclosure can be applied to automatic non-targeted identification of emission sources of various components based on chromatography-mass spectrometry analysis for air quality simulation, including mobile sources, combustion sources, atmospheric environment, etc. It is also applicable to other automatic identification needs of functional groups of a large number of components based on mass spectrometry.

[0047] like Figure 3 As shown, the substance identification method of this disclosure first collects the NIST standard substance mass spectrometry library, uses the positions and corresponding intensities of all mass spectrometry peaks as the model input feature vector, and trains an XGBoost model using the substance structure (elements and functional groups) as the model output label. The model output contains 32 label probabilities. Based on these probabilities and the candidate substance names and structures obtained from the spectral library matching, the elemental and functional group composition is comprehensively determined. Then, based on the predicted label composition, component screening and classification are performed, achieving non-targeted automatic screening and classification of all components based on functional group structure.

[0048] In some exemplary embodiments, in step 202, the element labels include: oxygen (O), nitrogen (N), sulfur (S), and silicon (Si); the structure labels include multiple aliphatic functional groups, multiple aromatic functional groups, multiple oxygen-containing structural functional groups, multiple nitrogen-containing structural functional groups, and multiple sulfur-containing structural functional groups.

[0049] For example, aliphatic functional groups may include alkyl (saturated carbon chain), olefinic bonds, aliphatic rings, heteroaliphatic rings; however, this disclosure does not limit them.

[0050] For example, aromatic functional groups may include: aromatic rings, heteroaromatic rings, phenolic hydroxyl groups, benzyl groups, 0 benzene rings, 1 benzene ring, 2 benzene rings, or multiple benzene rings; however, this disclosure does not limit this.

[0051] In this embodiment of the disclosure, the aromatic ring is used as the criterion for determining whether a substance has one benzene ring, two benzene rings, or multiple benzene rings. "One benzene ring" corresponds to a monocyclic aromatic substance, "two benzene rings" corresponds to a substance containing two benzene rings (including biphenyls and fused rings, i.e., naphthalene series), and "multiple benzene rings" corresponds to a substance containing three or more benzene rings (without distinguishing the type of benzene ring combination).

[0052] In this embodiment of the disclosure, both heteroaromatic rings and heteroalicylic rings can be classified into the category of heterocyclic substances.

[0053] For example, oxygen-containing functional groups may include hydroxyl, ether, aldehyde, ketone carbonyl, carboxyl, and ester groups; however, this disclosure does not limit them.

[0054] For example, nitrogen-containing functional groups may include amides, amines, imines, cyano groups, nitro groups, and nitroso groups; however, this disclosure does not limit them.

[0055] For example, sulfur-containing functional groups may include sulfides, sulfones, sulfoxides, and disulfide bonds; however, this disclosure does not limit them.

[0056] In some exemplary embodiments, in step 203, determining the elements and functional groups included in the analyte based on the probability of each element tag and the probability of each structure tag includes: Determine whether the analyte includes column loss based on the probability of silicon. Determine whether the analyte contains nitrogen-containing components based on the probability of multiple nitrogen-containing functional groups. Determine whether the analyte contains sulfur-containing components based on the probability of multiple sulfur-containing functional groups. The probability of heterocyclic functional groups is used to determine whether the analyte contains heterocyclic components, where heterocyclic functional groups include heteroaromatic rings and heteroalicylic rings. The probability of a single benzene ring, two benzene rings, or multiple benzene rings is used to determine whether the analyte includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, or polycyclic aromatic hydrocarbons. Whether the analyte contains aliphatic and oxygen-containing functional groups is determined by the probability of multiple aliphatic functional groups and multiple oxygen-containing functional groups.

[0057] In some exemplary embodiments, determining whether the analyte includes column bleed based on the probability of silicon includes: Compare the probability of silicon with a preset probability threshold. When the probability of silicon is greater than a preset probability threshold, the analyte is determined to include column loss. When the probability of silicon is less than or equal to a preset probability threshold, the analyte is determined to exclude column loss.

[0058] In some exemplary embodiments, determining whether the analyte includes a nitrogen-containing component based on the probability of multiple nitrogen-containing functional groups includes: Compare the probabilities of functional groups in multiple nitrogen-containing structures with a preset probability threshold. When the maximum probability among multiple nitrogen-containing functional groups is greater than a preset probability threshold, the analyte is determined to include a nitrogen-containing component. When the maximum probability among multiple nitrogen-containing functional groups is less than or equal to a preset probability threshold, the substance to be tested is determined to not contain nitrogen-containing components.

[0059] In some exemplary embodiments, determining whether the analyte includes a sulfur-containing component based on the probability of multiple sulfur-containing functional groups includes: Compare the probabilities of functional groups in multiple sulfur-containing structures with a preset probability threshold. When the maximum probability among multiple sulfur-containing functional groups is greater than a preset probability threshold, the substance to be tested is determined to include sulfur-containing components. When the maximum probability among multiple sulfur-containing functional groups is less than or equal to a preset probability threshold, the substance to be tested is determined to not contain sulfur-containing components.

[0060] Because element-based tags can produce false positives (N element, S element), when determining whether a structure belongs to nitrogen-containing or sulfur-containing structures, the prediction should be based primarily on the functional group structure, rather than on the element-based tags.

[0061] In some exemplary embodiments, determining whether the analyte includes a heterocyclic component based on the probability of the heterocyclic functional group includes: Compare the probabilities of heteroaromatic rings and heteroalicylic rings with the preset probability thresholds; When the probability of heterocyclic aromatic rings and / or heteroalicylic rings is greater than a preset probability threshold, the analyte is determined to include heterocyclic components. When the probability of heterocyclic aromatic rings and heterocyclic aliphatic rings are both less than or equal to the preset probability threshold, the analyte is determined to not contain heterocyclic components.

[0062] In some exemplary embodiments, determining whether the analyte includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons based on the probability of a single benzene ring, two benzene rings, or multiple benzene rings includes: The probability of detecting a single benzene ring, two benzene rings, or multiple benzene rings is greater than a preset probability threshold. When one or more probabilities of the structural tags of a single benzene ring, two benzene rings, and multiple benzene rings are greater than a preset probability threshold, the first structure of the analyte is determined based on the structural tag corresponding to the one or more probabilities greater than the preset probability threshold. The first structure includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons.

[0063] In some exemplary embodiments, determining whether the analyte includes both aliphatic and oxygen-containing functional groups based on the probabilities of multiple aliphatic functional groups and multiple oxygen-containing functional groups includes: Compare the probabilities of multiple adipose functional groups and multiple oxygen-containing functional groups with a preset probability threshold. When the probability of any aliphatic functional group and / or any oxygen-containing functional group is greater than a preset probability threshold, the substance to be tested is determined to include the aliphatic functional group and / or oxygen-containing functional group. When the probability of any aliphatic functional group and / or any oxygen-containing functional group is less than or equal to a preset probability threshold, the analyte is determined to not contain that aliphatic functional group and / or oxygen-containing functional group.

[0064] For example, based on the probabilities of each label given by the aforementioned machine learning model, the presence of structures of interest in substances is identified and presented in a list format. The main criterion for judgment is whether the predicted probability of the corresponding label is higher than a preset probability threshold. In one example, except for the aldehyde group, the preset probability threshold for each element label and structure label is 0.5. Since the mass spectrometry characteristics of the aldehyde group are not obvious, in this embodiment of the disclosure, the preset probability threshold for the aldehyde group is set to 0.1 to improve the resolution of the aldehyde group.

[0065] Considering the significant environmental impact of polycyclic aromatic hydrocarbons (PAHs), this embodiment classifies aromatic components into monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons. The discrimination criterion is to select the one with the highest probability among the three corresponding labels ("single benzene ring (aromatic_1)", "two benzene rings (aromatic_2)", and "multiple benzene rings (PAHs)") that is higher than a preset probability threshold of 0.5.

[0066] Considering the potential impact of impurities on element identification, and assuming that impurities do not contain significant functional group level features, the identification criterion for whether or not N or S elements is present is changed in this embodiment: if the predicted probability of the label containing N (or S) functional group is greater than 0.5, then the component is considered to contain N (or S).

[0067] In actual sample applications, the spectral library matching effect of silicon-containing components (column bleed) and internal standards was generally good. In order to fully identify column bleed and deuterated internal standard components and exclude them, the embodiments of this disclosure directly identify whether the component contains Si or deuterium by using the chemical formula given by the spectral library matching, so as to determine whether it is column bleed or internal standard.

[0068] Step 203 provides information on the elements and functional groups present in the target mass spectrometer, thus providing a data foundation for the next step of substance identification. The output data from this step can also be directly used as a criterion for component classification, because most species contain multiple structures of interest, especially for co-eluting or interfering elution peaks that are not suitable for single-substance classification.

[0069] In some exemplary embodiments, in step 203, when determining the elements and functional groups included in the substance to be tested, the elements included in the substance to be tested can be determined comprehensively based on information such as the probability of element labels output by the model and the matching similarity obtained by spectral library matching.

[0070] In some exemplary embodiments, in step 203, the elements included in the analyte are determined comprehensively based on information such as the probability of the element labels output by the model and the matching similarity obtained from spectral library matching, including: Compare the similarity of matches obtained from the spectrum library matching with the preset similarity threshold; If the matching similarity is above the preset matching similarity threshold (i.e., greater than or equal to the preset matching similarity threshold), then when the predicted probability of the "nitrogen element" label is greater than or equal to the preset probability threshold, it is determined that the substance to be tested includes nitrogen-containing components. If the matching similarity is above the preset matching similarity threshold, then when the predicted probability of the "nitrogen element" label is less than the preset probability threshold, it is determined that the substance to be tested does not contain nitrogen-containing components.

[0071] If the matching similarity is above the preset matching similarity threshold, then when the predicted probability of the "sulfur element" label is greater than or equal to the preset probability threshold, it is determined that the substance to be tested includes sulfur-containing components. If the matching similarity is above the preset matching similarity threshold, then when the predicted probability of the "sulfur element" label is less than the preset probability threshold, it is determined that the substance to be tested does not contain sulfur-containing components.

[0072] For example, the preset matching similarity threshold is a real number greater than or equal to 800.

[0073] In some exemplary embodiments, in step 204, determining the substance category of the analyte based on the elements and functional groups it comprises includes: When the first structure of the substance to be tested is a monocyclic aromatic hydrocarbon or a bicyclic aromatic hydrocarbon, the second structure of the substance to be tested is determined. The second structure is an aliphatic functional group or an oxygen-containing functional group. The substance category is determined based on the first structure and the second structure. When the first structure of the analyte is not a monocyclic aromatic hydrocarbon, bicyclic aromatic hydrocarbon, or polycyclic aromatic hydrocarbon, the test is performed to determine whether the analyte includes an alkyl group. If the analyte includes an alkyl functional group, the first structure of the analyte is determined to be an alkane, and the second structure of the analyte is determined. The second structure is an aliphatic ring, an alkene bond, or an oxygen-containing functional group. If the second structure is an aliphatic ring, the test is performed to determine whether the third structure of the analyte includes an alkene bond. If the second structure is an alkene bond, the test is performed to determine whether the third structure of the analyte includes an oxygen-containing functional group. The substance category is determined based on the first and second structures, or the first, second, and third structures.

[0074] In some exemplary embodiments, the method further includes: Determine whether the elements and functional groups included in the analyte are non-interest components. Non-interest components include: deuterium-containing components, silicon-containing components, carbon-free components, components with a library similarity lower than a preset library similarity threshold and whose predicted elements and structures contain at least X elements and / or functional groups, components with predicted oxygen-containing functional groups including more than Y oxygen-containing functional groups, and components with a library similarity higher than a preset library similarity threshold and whose predicted matching components correspond to chemical formulas with more than Z oxygen atoms. Here, X is a natural number greater than or equal to 3, Y is a natural number between 2 and 4, and Z is a natural number greater than or equal to 4.

[0075] In this embodiment of the disclosure, the elution peak components can be screened and classified based on the elemental and functional group composition given in step 203. In this embodiment of the disclosure, before determining the components of interest, non-components can be screened and eliminated. This step mainly involves screening and eliminating mass spectrometers with column loss, excessive internal standards, and excessive impurities, as mentioned above. The following description uses motor vehicle emission components as an example (it can also be used for atmospheric environmental monitoring, stationary source emission characteristic analysis, etc.). The screening criteria for non-components include: (1) Components containing deuterium atoms are not emission components and are classified as "internal standards".

[0076] (2) Components containing silicon or not containing carbon are not (organic) emission components (e.g., N2O, SO2, etc.).

[0077] (3) Components with a library similarity lower than a preset library similarity threshold and whose predicted functional group list contains X elements or more functional groups are not considered emission components, where X is a natural number greater than or equal to 3. Library similarity refers to the degree of matching between the mass spectrum of the elution peak and the standard material library, with a value ranging from 0 to 1000. The higher the library similarity, the higher the degree of matching. In some exemplary embodiments, the preset library similarity threshold can be between 800 and 850; for example, the preset library similarity threshold can be 800. For example, X can be 5.

[0078] (4) Since motor vehicle emission components are generally considered to have a low degree of oxidation, resulting in fewer primary emission components containing multiple oxygen-containing functional groups, components with more than Y types (excluding) oxygen-containing functional groups in the prediction list are not considered emission components, where Y is a natural number between 2 and 4. For example, Y can be 2; however, this disclosure does not limit this.

[0079] (5) Components with a similarity to the spectral library higher than the preset similarity threshold and more than Z oxygen atoms in their chemical formulas are not considered emission components, for the same reason as above, where Z is a natural number greater than or equal to 4. For example, Z can be 4; however, this disclosure does not limit this.

[0080] This screening method was used to identify blank analysis samples from adsorption tubes. These blank samples were obtained directly from the instrument after adsorption tube desorption and activation without sampling, and mainly contained components such as column runoff, impurity carrier gas, and background from the adsorption tube. The identification results showed that this method can identify non-emission components accounting for more than 60% of the peak area, demonstrating its ability to effectively identify non-interested components such as column runoff and inorganic gases that interfere with component identification. This can improve the reliability of subsequent identification of components of interest to a certain extent.

[0081] Next, the components of interest are classified. Based on the predicted component elements and functional group composition, the most likely substance category to which the component belongs is identified, so as to facilitate the application of calibrated quantitative component mass and subsequent calculation of emission factors and input into the air quality model. Since most substance categories consist of two or more predicted labels, it is necessary to integrate the structures listed in the above label prediction list to determine which category the component belongs to. The specific classification criteria are shown in Tables 1 and 2. Table 1 Table 2 The machine learning model constructed in this disclosure can simultaneously recognize multiple structures, such as Figure 4 As shown, taking methyl phenyl ketone as an example, it is correctly predicted that it contains benzene ring, alkyl, and ketone carbonyl groups. In the final classification stage, the most important functional group is selected as the category of the organic component based on the functional group composition.

[0082] Furthermore, in this embodiment, 88 gas chromatographic grade pure standard substances (including alkanes, alkenes, cycloalkanes, benzene and its derivatives, polycyclic aromatic hydrocarbons, alcohols, aldehydes, and ketones) were selected and pretreated by Tenax adsorption tube adsorption-thermal desorption before being introduced into a two-dimensional gas chromatography-time-of-flight mass spectrometry (GCGC-ToF-MS) instrument to analyze the structural composition of the eluted components using this method. As shown in Table 3, among the identified standard substance eluted peaks, the predicted functional group composition of 84 components was completely correct (e.g., benzaldehyde required identification of alkyl groups, benzene rings, and aldehyde groups). Among them, alkanes, cycloalkanes, and aldehydes were all correctly identified. For the remaining components, one component was partially misidentified, but all were misidentified as similar structures (e.g., alkene bonds were identified as aliphatic rings, and alcohol hydroxyl groups were identified as ether bonds, etc. We believe that although there were identification errors, most of the misidentifications were similar structures and would not have a significant impact on the physicochemical properties of the species and subsequent simulations), detailed below: (1) Octene, no alkene bond was identified; (2) Ethylhexanol, with an additional ether bond identified; (3) Hexanone, an additional ether bond was identified; (4) The olefin bond of styrene is identified as a ring. Table 3 The substance identification method of this disclosure, by applying machine learning methods to the analytical data of a full two-dimensional gas chromatography-time-of-flight mass spectrometry (GC-TOF-MS), can significantly improve the proportion of qualitatively classified components and enhance the resolution of functional group composition of full-spectrum components compared to traditional I / SVOC identification techniques. It should be noted that, theoretically, all techniques capable of deriving mass spectra (mass number - peak intensity) and related analytical samples can use this method for non-targeted species classification and identification.

[0083] like Figure 5 As shown, this disclosure also provides a model building apparatus, including a sample acquisition module 510, a feature building module 520, a label building module 530, and a model training module 540, wherein: The sample acquisition module 510 is configured to acquire training samples, the training samples including multiple standard material mass spectra; The feature construction module 520 is configured to construct a feature vector of the training sample, the feature vector of the training sample including the relative peak intensities of multiple mass-to-charge ratio positions of the mass spectrum of the standard substance; The label building module 530 is configured to build training sample labels, the training sample labels including element labels and structure labels, the element labels including multiple element categories, and the structure labels including multiple functional group structure categories; The model training module 540 is configured to build a machine learning model and train and test the machine learning model based on the feature vectors of the training samples and the labels of the training samples.

[0084] In some exemplary embodiments, the plurality of mass-to-charge ratio positions are all integer mass-to-charge ratio positions within a preset mass-to-charge ratio range.

[0085] In some exemplary embodiments, the feature vector of the training sample further includes the positions of the first to M base peaks of the standard material mass spectrometer, and the relative intensities of the second to M base peaks relative to the first base peak, where M is a natural number greater than 1.

[0086] In some exemplary embodiments, the label building module 530 constructs training sample labels, including: For each standard substance mass spectrometer, perform the following operations: Obtain the substance name and chemical formula corresponding to the mass spectrometry of the standard substance, determine the simplified molecular linear input standard string corresponding to the substance name, determine the functional group structure category contained in the standard substance based on the simplified molecular linear input standard string, determine the element category contained in the standard substance based on the chemical formula, and generate training sample labels corresponding to the mass spectrometry of the standard substance based on the determined functional group structure category and element category.

[0087] In some exemplary embodiments, the machine learning model is the XGBoost machine learning model.

[0088] like Figure 6 As shown, this disclosure also provides a model building apparatus, including a feature construction module 610, a model input module 620, a component determination module 630, and a category determination module 640, wherein: The feature construction module 610 is configured to acquire mass spectrometry data of the substance to be tested and construct a feature vector of the substance to be tested based on the mass spectrometry data of the substance to be tested. The model input module 620 is configured to input the feature vector of the substance to be tested into a machine learning model to obtain the probability of each element label and the probability of each structure label. The machine learning model is constructed using the model construction method as described in any embodiment of this disclosure. The component determination module 630 is configured to determine the elements and functional groups included in the analyte based on the probability of each element tag and the probability of each structure tag. The category determination module 640 is configured to determine the substance category of the substance to be tested based on the elements and functional groups included in the substance to be tested.

[0089] In some exemplary embodiments, the elemental label includes oxygen, nitrogen, sulfur, and silicon; the structural label includes multiple aliphatic functional groups, multiple aromatic functional groups, multiple oxygen-containing structural functional groups, multiple nitrogen-containing structural functional groups, and multiple sulfur-containing structural functional groups.

[0090] In some exemplary embodiments, the aromatic functional group includes: a single benzene ring, two benzene rings, multiple benzene rings, and heteroaromatic rings; the aliphatic functional group includes: alkyl groups, olefinic bonds, aliphatic rings, and heteroaliphatic rings; The component determination module 630 determines the elements and functional groups included in the analyte based on the probability of each element tag and the probability of each structural tag, including: The probability of silicon element is used to determine whether the analyte includes column loss. Whether the analyte includes a nitrogen-containing component is determined based on the probability of multiple nitrogen-containing functional groups. Whether the analyte includes sulfur-containing components is determined based on the probability of multiple sulfur-containing functional groups. The probability of heterocyclic functional groups is used to determine whether the analyte includes heterocyclic components, wherein the heterocyclic functional groups include heteroaromatic rings and heteroalicylic rings; The probability of a single benzene ring, two benzene rings, or multiple benzene rings is used to determine whether the analyte includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons. Whether the analyte includes aliphatic and oxygen-containing functional groups is determined based on the probabilities of multiple aliphatic functional groups and multiple oxygen-containing functional groups.

[0091] In some exemplary embodiments, the component determination module 630 determines whether the analyte includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons based on the probability of a single benzene ring, two benzene rings, or multiple benzene rings, including: The probability of detecting a single benzene ring, two benzene rings, or multiple benzene rings is greater than a preset probability threshold. When one or more probabilities of the structural tags of a single benzene ring, two benzene rings, and multiple benzene rings are greater than a preset probability threshold, the first structure of the substance to be tested is determined according to the structural tag corresponding to the one or more probabilities greater than the preset probability threshold. The first structure includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons.

[0092] In some exemplary embodiments, the category determination module 660 determines the substance category of the analyte based on the elements and functional groups included in the analyte, including: When the first structure of the substance to be tested is a monocyclic aromatic hydrocarbon or a bicyclic aromatic hydrocarbon, the second structure of the substance to be tested is determined, wherein the second structure is the aliphatic functional group or the oxygen-containing functional group, and the substance category is determined based on the first structure and the second structure. When the first structure of the analyte is not a monocyclic aromatic hydrocarbon, bicyclic aromatic hydrocarbon, or polycyclic aromatic hydrocarbon, the analyte is tested to determine whether it includes an alkyl group. When the analyte includes an alkyl functional group, the first structure of the analyte is determined to be an alkane, and the second structure of the analyte is determined, wherein the second structure is the aliphatic ring, the alkene bond, or the oxygen-containing functional group. When the second structure is the aliphatic ring, the third structure of the analyte is determined to include an alkene bond. When the second structure is the alkene bond, the third structure of the analyte is determined to include an oxygen-containing functional group. The substance category is determined based on the first structure and the second structure, or the first structure, the second structure, and the third structure.

[0093] In some exemplary embodiments, the category determination module 660 is further configured to: determine whether the elements and functional groups included in the analyte are non-interest components, wherein the non-interest components include: deuterium-containing components, silicon-containing components, carbon-free components, components with a library similarity lower than a preset library similarity threshold and predicted elements and / or functional groups containing at least X elements and / or functional groups, components with predicted oxygen-containing functional groups containing more than Y oxygen-containing functional groups, and components with a library similarity higher than a preset library similarity threshold and predicted matching components corresponding to chemical formulas containing at least Z oxygen atoms, wherein X is a natural number greater than or equal to 3, Y is a natural number between 2 and 4, and Z is a natural number greater than or equal to 4.

[0094] This disclosure also provides a model building apparatus, including a memory; and a processor connected to the memory, the memory being used to store instructions, the processor being configured to execute the steps of the model building method as described in any embodiment of this disclosure based on the instructions stored in the memory.

[0095] like Figure 7As shown, in one example, the model building apparatus may include: a first processor 710, a first memory 720, a first bus system 730, and a first transceiver 740. The first processor 710, first memory 720, and first transceiver 740 are connected via the first bus system 730. The first memory 720 stores instructions, and the first processor 710 executes the instructions stored in the first memory 720 to control the first transceiver 740 to transmit and receive signals. Specifically, the first transceiver 740, under the control of the first processor 710, can acquire training samples, which include multiple standard material mass spectra. The first processor 710 constructs feature vectors for the training samples, which include the relative peak intensities of multiple mass-to-charge ratio positions of the standard material mass spectra; constructs training sample labels, which include element labels and structure labels. The element labels include multiple element categories, and the structure labels include multiple functional group structure categories. A machine learning model is constructed, and the machine learning model is trained and tested based on the feature vectors and training sample labels of the training samples.

[0096] It should be understood that the first processor 710 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0097] The first memory 720 may include read-only memory and random access memory, and provides instructions and data to the first processor 710. A portion of the first memory 720 may also include non-volatile random access memory. For example, the first memory 720 may also store device type information.

[0098] In addition to a data bus, the first bus system 730 may also include a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 7 The general designated all buses as the first bus system 730.

[0099] In implementation, the processing performed by the processing device can be accomplished through integrated logic circuits in the hardware of the first processor 710 or through software instructions. That is, the method steps of this embodiment can be executed by a hardware processor, or by a combination of hardware and software modules within the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other storage media. This storage medium is located in the first memory 720. The first processor 710 reads information from the first memory 720 and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, further details are omitted here.

[0100] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the model building method as described in any embodiment of this disclosure. The model building method driven by executing executable instructions is essentially the same as the model building method provided in the above embodiments of this disclosure, and will not be described in detail here.

[0101] In some possible implementations, various aspects of the model building method provided in this disclosure may also be implemented as a program product comprising program code that, when run on a computer device, causes the computer device to perform the steps in the model building method according to various exemplary embodiments of this disclosure as described above, for example, the computer device may execute the model building method described in the embodiments of this disclosure.

[0102] This disclosure also provides a substance identification device, including a memory; and a processor connected to the memory, the memory being used to store instructions, the processor being configured to perform steps of the substance identification method as described in any embodiment of this disclosure based on the instructions stored in the memory.

[0103] like Figure 8As shown, in one example, the substance identification device may include: a second processor 810, a second memory 820, a second bus system 830, and a second transceiver 840. The second processor 810, the second memory 820, and the second transceiver 840 are connected via the second bus system 830. The second memory 820 stores instructions, and the second processor 810 executes the instructions stored in the second memory 820 to control the second transceiver 840 to transmit and receive signals. Specifically, the second transceiver 840, under the control of the second processor 810, acquires mass spectrometry data of the substance to be tested. The second processor 810 constructs a feature vector of the substance to be tested based on the mass spectrometry data, inputs the feature vector of the substance to be tested into a machine learning model, and obtains the probability of each element label and the probability of each structural label. The machine learning model is constructed using the model construction method described in any embodiment of this disclosure. Based on the probability of each element label and the probability of each structural label, the elements and functional groups included in the substance to be tested are determined. Based on the elements and functional groups included in the substance to be tested, the substance category of the substance to be tested is determined.

[0104] It should be understood that the second processor 810 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0105] The second memory 820 may include read-only memory and random access memory, and provides instructions and data to the second processor 810. A portion of the second memory 820 may also include non-volatile random access memory. For example, the second memory 820 may also store device type information.

[0106] In addition to the data bus, the second bus system 830 may also include a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 8 The general designated all buses as the second bus system 830.

[0107] In implementation, the processing performed by the processing device can be accomplished through integrated logic circuits in the hardware of the second processor 810 or through software instructions. That is, the method steps of this embodiment can be executed by a hardware processor, or by a combination of hardware and software modules within the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other storage media. This storage medium is located in the second memory 820. The second processor 810 reads information from the second memory 820 and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, further details are omitted here.

[0108] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the substance identification method as described in any embodiment of this disclosure. The substance identification method driven by executing executable instructions is essentially the same as the substance identification method provided in the above embodiments of this disclosure, and will not be described in detail here.

[0109] In some possible implementations, various aspects of the substance identification method provided in this disclosure may also be implemented as a program product comprising program code that, when run on a computer device, causes the computer device to perform the steps of the substance identification method according to various exemplary embodiments of this disclosure as described above. For example, the computer device may execute the substance identification method described in the embodiments of this disclosure.

[0110] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0111] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0112] It should be noted that the above embodiments or implementation methods are merely exemplary and not restrictive. Therefore, this disclosure is not limited to the content specifically shown and described herein. Various modifications, substitutions, or omissions can be made to the form and details of the implementations without departing from the scope of this disclosure.

Claims

1. A method for identifying substances, characterized in that, The method for identifying moderately volatile and semi-volatile organic components includes: Obtain mass spectrometry data of the substance to be tested, and construct a feature vector of the substance to be tested based on the mass spectrometry data of the substance to be tested; The feature vector of the substance to be tested is input into a machine learning model to obtain the probability of each element label and the probability of each structure label. The machine learning model is constructed as follows: Training samples are obtained, including multiple standard mass spectra; a feature vector of the training samples is constructed, comprising the relative peak intensities of multiple mass-to-charge ratio positions of the standard mass spectra; training sample labels are constructed, including element labels and structure labels. The element labels include oxygen, nitrogen, sulfur, and silicon; the structure labels include multiple aliphatic functional groups, multiple aromatic functional groups, multiple oxygen-containing functional groups, multiple nitrogen-containing functional groups, and multiple sulfur-containing functional groups. The aromatic functional groups include a single benzene ring, two benzene rings, multiple benzene rings, and heteroaromatic rings; the aliphatic functional groups include alkyl groups, olefin bonds, aliphatic rings, and heteroaliphatic rings; a machine learning model is constructed, and trained and tested based on the feature vector and training sample labels of the training samples. Based on the probability of each element tag and the probability of each structure tag, the elements and functional groups included in the substance to be tested are determined. Based on the elements and functional groups included in the substance to be tested, determine the first, second, and third structures of the substance to be tested, and determine the substance category of the substance to be tested according to the following table: 。 2. The method according to claim 1, characterized in that, The step of determining the elements and functional groups included in the analyte based on the probability of each element tag and the probability of each structural tag includes: The probability of silicon element is used to determine whether the analyte includes column loss. Whether the analyte includes a nitrogen-containing component is determined based on the probability of multiple nitrogen-containing functional groups. Whether the analyte includes sulfur-containing components is determined based on the probability of multiple sulfur-containing functional groups. The probability of heterocyclic functional groups is used to determine whether the analyte includes heterocyclic components, wherein the heterocyclic functional groups include heteroaromatic rings and heteroalicylic rings; The probability of a single benzene ring, two benzene rings, or multiple benzene rings is used to determine whether the analyte includes monocyclic aromatic hydrocarbons, bicyclic aromatic hydrocarbons, and polycyclic aromatic hydrocarbons. Whether the analyte includes aliphatic and oxygen-containing functional groups is determined based on the probabilities of multiple aliphatic functional groups and multiple oxygen-containing functional groups.

3. The method according to claim 1, characterized in that, The method further includes: Determine whether the elements and functional groups included in the analyte are non-interest components, wherein the non-interest components include: deuterium-containing components, silicon-containing components, carbon-free components, components with a library similarity lower than a preset library similarity threshold and predicted elements and / or functional groups containing at least X elements and / or functional groups, components with predicted oxygen-containing functional groups containing more than Y oxygen-containing functional groups, and components with a library similarity higher than a preset library similarity threshold and predicted matching components corresponding to chemical formulas containing at least Z oxygen atoms, wherein X is a natural number greater than or equal to 3, Y is a natural number between 2 and 4, and Z is a natural number greater than or equal to 4.

4. A substance identification device, characterized in that, The method includes a memory; and a processor connected to the memory, the memory being used to store instructions, the processor being configured to perform the steps of the substance identification method as described in any one of claims 1 to 3 based on the instructions stored in the memory.

5. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the substance identification method as described in any one of claims 1 to 3.

6. A computer program product, characterized in that, The instructions include, when the computer program product is executed by a computer, performing the substance identification method as described in any one of claims 1 to 3.

Citation Information

Patent Citations

  • Mass spectrometric method for determining the presence or absence of a chemical element in an analyte

    US20200243315A1