Method for determining volatile organic compounds in surface water by purging and trapping and gas chromatography-mass spectrometry

Abstract

The invention discloses a method for determining volatile organic compounds in surface water by purging and trapping and gas chromatography-mass spectrometry, which comprises the following steps: acquiring mass spectrometry data of volatile organic compounds in a surface water sample, and separating and detecting a multi-component mixture in a complex matrix by gas chromatography-mass spectrometry, obtaining an original data set containing retention time and mass spectrum information; according to an output result of the classification model, carrying out qualitative analysis on unknown compounds in a to-be-detected sample, and if the classification probability is greater than a preset threshold value, determining the types of the compounds to obtain an accurate qualitative recognition result; and establishing a concentration standard curve by adopting a corresponding quantitative ion peak area according to a target compound confirmed in the qualitative identification result, and calculating a linear relation parameter between the peak area and the concentration to obtain a regression equation for quantitative analysis.

Description

Technical Field

[0001] This invention relates to the field of volatile organic compounds in surface water, and more particularly to a method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry. Background Technology

[0002] The detection of volatile organic compounds (VOCs) in surface water is a crucial issue in environmental monitoring, directly impacting water resource security and ecological protection. VOCs are diverse in type and origin, and their low concentration and high toxicity make accurate detection essential for ensuring water quality safety. However, current detection methods often suffer from insufficient sensitivity and selectivity in complex matrices, especially in multi-component mixtures. Traditional methods struggle to effectively distinguish target compounds from interfering substances, leading to decreased reliability of results. Furthermore, the lack of standardized criteria often affects the quantitative accuracy of existing technologies when processing VOCs with varying chemical properties.

[0003] In practical testing, the core challenge lies in achieving precise qualitative and quantitative analysis of target compounds in complex matrices. Surface water contains a diverse range of volatile organic compounds, such as benzene compounds and haloalkanes, with significantly different chemical properties, making comprehensive coverage difficult with a single detection method. Inaccurate selection of characteristic ions can lead to qualitative errors. For example, when detecting benzene compounds, interfering substances may produce similar ion signals to the target compound, resulting in misjudgment. This qualitative inaccuracy further affects the quantitative process, as quantification depends on the stability and linearity of the characteristic ion peak areas. If the characteristic ions are not selected appropriately, the correspondence between peak area ratios and concentrations may be distorted, leading to concentration measurement deviations. For instance, when analyzing a lake water sample, multiple volatile organic compounds may coexist, and interfering ion signals can cause the quantitative results to deviate from the actual values.

[0004] Therefore, how to achieve accurate qualitative and quantitative analysis of various volatile organic compounds in complex matrices by optimizing the selection of characteristic ions and peak area ratio analysis has become a key issue in the field of environmental monitoring. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for determining volatile organic compounds in surface water by purge and trap, gas chromatography-mass spectrometry, in order to overcome the shortcomings of the prior art.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: Methods for determining volatile organic compounds in surface water using purge-and-trap, gas chromatography-mass spectrometry, include: Mass spectrometry data of volatile organic compounds in surface water samples were obtained, and multi-component mixtures in complex matrices were separated and detected by gas chromatography-mass spectrometry to obtain a raw data set containing retention time and mass spectrum information. Based on the mass spectrum information in the original dataset, a spectral matching algorithm is used to analyze the mass spectrometric characteristics of each chromatographic peak. If the ion abundance in the mass spectrum is greater than a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, then the candidate characteristic ion corresponding to the peak is determined. By evaluating the stability of the candidate characteristic ions, the relative standard deviation of the peak area of ​​the characteristic ions at different concentration levels is calculated. If the relative standard deviation is less than the preset value, the ion is determined to be suitable as a quantitative ion, and the optimized combination of characteristic ions is obtained. The optimized characteristic ion combination is used to construct a peak area ratio matrix. The ion signal patterns of different volatile organic compounds are classified and trained using the support vector machine algorithm to obtain a classification model that can distinguish between target compounds and interfering substances. Based on the output of the classification model, a qualitative analysis is performed on the unknown compounds in the test sample. If the classification probability is greater than a preset threshold, the compound type is determined and an accurate qualitative identification result is obtained. Based on the target compounds identified in the qualitative identification results, a concentration standard curve is established using the corresponding quantitative ion peak area. The linear relationship parameters between peak area and concentration are calculated to obtain the regression equation for quantitative analysis. The peak area of ​​the target compound in the actual sample is converted into concentration based on the regression equation. At the same time, the peak recognition algorithm is used to separate the overlapping peaks in the chromatogram to obtain the corrected peak area value for the final concentration determination.

[0007] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a highly efficient method for the detection and quantitative analysis of volatile organic compounds (VOCs) in surface water samples, solving the challenges of separation, qualitative and quantitative analysis of multi-component mixtures in complex matrices. Raw data, including retention times and mass spectra, are obtained using gas chromatography-mass spectrometry (GC-MS). A spectral matching algorithm is employed to analyze mass spectrometry characteristics, screening candidate characteristic ions that match the target compound's features. Stability assessment is then used to optimize the quantitative ion combination. This invention utilizes a support vector machine (SVM) algorithm to construct a classification model, distinguishing the target compound from interfering substances based on ion signal patterns, achieving highly accurate qualitative identification. Furthermore, a concentration standard curve is established using quantitative ion peak areas, and overlapping peaks are separated using a peak identification algorithm. Peak areas are then corrected, and finally, the target compound concentration is calculated using a regression equation. This invention significantly improves the detection sensitivity and accuracy of VOCs in complex matrices, providing efficient and reliable technical support for environmental monitoring. Attached Figure Description

[0008] Figure 1 This is a flowchart of the method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to the present invention.

[0009] Figure 2 This is a schematic diagram of the method for determining volatile organic compounds in surface water using purge-and-trap and gas chromatography-mass spectrometry according to the present invention. Figure 1 .

[0010] Figure 3 This is a schematic diagram of the method for determining volatile organic compounds in surface water using purge-and-trap and gas chromatography-mass spectrometry according to the present invention. Figure 2 . Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0012] like Figure 1-3 The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry in this embodiment may specifically include: Step S101: Obtain mass spectrometry data of volatile organic compounds in surface water samples, and separate and detect multi-component mixtures in complex matrices using gas chromatography-mass spectrometry to obtain a raw data set containing retention time and mass spectrum information.

[0013] Surface water samples were pretreated using gas chromatography-mass spectrometry (GC-MS) to separate volatile organic compounds (VOCs) and obtain a raw dataset containing retention times and mass spectra. If the retention time signal intensity in the raw dataset was below a preset threshold, a signal enhancement algorithm was used to denoise the data, resulting in an enhanced dataset. Based on the enhanced dataset, principal component analysis (PCA) was used to extract the main features from the mass spectra, resulting in a feature vector set. If outliers were found in the feature vector set, an anomaly detection algorithm was used to identify and remove them, resulting in a cleaned feature vector set. Using the cleaned feature vector set, a support vector machine (SVM) algorithm was used to classify the VOCs and determine the category of each compound. Based on the classification results, a pre-established VOC mass spectrometry database was used to obtain the chemical structure information of the compounds. By combining the chemical structure information with the retention time, a dataset containing the compound category and structure was generated.

[0014] For example, the analysis of volatile organic compounds in surface water using gas chromatography-mass spectrometry involves multiple technical steps, each with a clear objective and implementation method. The following analysis will be conducted through specific examples, closely focusing on the business scenario of surface water sample analysis, highlighting the technical details and beneficial effects of each step.

[0015] For example, in the pretreatment stage of surface water samples, solid-phase microextraction (SPE) can be used to extract volatile organic compounds (VOCs). Take 1 liter of surface water sample, add an internal standard compound such as benzene-d6 to calibrate the recovery rate, place it in a 20 mL headspace vial, heat to 60°C, insert an extraction fiber to adsorb VOCs for 20 minutes. This method efficiently enriches low-concentration compounds, reduces matrix interference, and provides high-quality samples for subsequent separation. In the gas chromatography-mass spectrometry (GC-MS) separation stage, a non-polar capillary column such as DB-5 can be used, with the temperature program set from 40°C, increasing at 10°C per minute to 250°C, and the mass spectrometry scan range being 50-500 m / z.

[0016] For example, the retention time of benzene compounds such as toluene is approximately 8.5 minutes, and the mass spectrum shows characteristic ions at m / z 91 and 92. This step generates raw data based on retention time and mass spectrum, accurately distinguishing the chemical properties of different compounds. For raw data with signal intensities below a threshold, a wavelet transform denoising algorithm can be used. Assuming a signal intensity threshold of 1000 is set, signals below this value are considered noise. In one possible implementation, wavelet decomposition is performed on the mass spectrum signal of toluene, retaining the effective components in the high-frequency signal and removing low-frequency noise to obtain an enhanced dataset. Compared to the raw data, the signal-to-noise ratio of the enhanced data is improved by approximately 30%, significantly improving the accuracy of subsequent feature extraction. In the principal component analysis stage, mass spectrum features are extracted from the enhanced data.

[0017] For example, for samples containing toluene, ethylbenzene, and chlorobenzene, the characteristic ions at m / z 91, 106, and 112 of their mass spectra are analyzed. Principal component analysis reduces the data dimensionality to 2-3 principal components, retaining over 95% of the information variance. This method effectively condenses key information, reduces computational complexity, and facilitates subsequent classification. Outliers in the feature vector can be detected using the isolated forest algorithm.

[0018] For example, if the feature vector of a sample deviates abnormally due to contamination or instrument drift, anomaly scores are calculated using an isolation forest, and outliers with scores higher than 0.7 are removed. The cleaned feature vector set is more representative, avoiding interference from outlier data in the classification results. In the support vector machine classification stage, based on the cleaned feature vectors, the model can be trained to distinguish between benzene compounds and halogenated hydrocarbons.

[0019] For example, given feature vectors containing toluene and chlorobenzene as input, the model constructs classification boundaries using radial basis function kernels, achieving an accuracy of over 90%. This method efficiently identifies compound categories through machine learning, reducing manual interpretation time. Finally, it matches the data against a pre-built volatile organic compound mass spectrometry database to obtain chemical structure information.

[0020] For example, after matching the mass spectrum of toluene with a database, its molecular formula was confirmed as C7H8, and its structure as a monosubstituted benzene ring. Combined with a retention time of 8.5 minutes, a dataset containing compound categories and structures was generated. This dataset can be used for environmental monitoring, tracing pollution sources, assessing water quality risks, and providing a scientific basis for environmental protection.

[0021] It should be noted that the above process, through multi-stage collaboration, ensures efficiency and accuracy across the entire chain, from sample pretreatment to structure identification. Optimization of each step, such as noise reduction and outlier removal, significantly improves data quality, while automation of classification and matching greatly reduces labor costs, providing reliable technical support for surface water pollution monitoring.

[0022] Step S102: Based on the mass spectrum information in the original data set, the mass spectrometry characteristics of each chromatographic peak are analyzed using a spectral matching algorithm. If the ion abundance in the mass spectrum is greater than a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, then the candidate characteristic ion corresponding to the peak is determined.

[0023] Mass spectrum information is extracted from the original dataset. A spectral matching algorithm is used to analyze chromatographic peak characteristics. If the ion abundance exceeds a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, candidate characteristic ions are identified. Based on these candidate characteristic ions, a clustering analysis algorithm is used to group characteristic ions with similar mass-to-charge ratios in the mass spectrum information, resulting in a set of characteristic ion clusters. These clusters are then compared with a pre-established compound feature library. If the mass-to-charge ratio pattern of the characteristic ion cluster matches a record in the library, the preliminary identity of the corresponding compound is determined. Based on this preliminary identity, retention time information of the chromatographic peaks is obtained. Combined with the distribution characteristics of the characteristic ion clusters in the mass spectrum information, a temporary dataset containing retention time and compound identity is generated. Statistical analysis is used to verify the consistency of retention time and characteristic ion cluster matching using this temporary dataset. If the consistency exceeds a preset threshold, the final identity of the compound is determined. Based on the final identity, a structured dataset containing compound identity, retention time, and mass spectrometric characteristics is generated. Finally, data standardization methods are used to format the compound identity and mass spectrometric characteristics using this structured dataset, resulting in a unified dataset suitable for subsequent analysis.

[0024] For example, in the analysis of surface water samples, when extracting mass spectrum information from the raw dataset, a spectral matching algorithm can be used to analyze chromatographic peak characteristics. The core of the spectral matching algorithm lies in comparing the similarity between the ion abundance and mass-to-charge ratio patterns of the mass spectrum and standard spectra.

[0025] Specifically, an ion abundance threshold of 5000 can be set, with a mass-to-charge ratio range of 50-400 m / z. If the ion abundance of a chromatographic peak exceeds 5000, and its mass-to-charge ratio (m / z 78 and 104) matches the characteristics of benzene or ethylbenzene, it is marked as a candidate characteristic ion. This method rapidly identifies the ion signals related to the target compound through threshold screening, providing a precise starting point for subsequent analysis.

[0026] In one possible implementation, when performing cluster analysis based on candidate feature ions, the K-means clustering algorithm can be used to group ions with similar mass-to-charge ratios.

[0027] For example, for mass spectra containing m / z 78, 91, and 104, the algorithm divides the ions into two groups based on their mass-to-charge ratio and abundance distribution: one group is dominated by m / z 78, which may correspond to benzene; the other group is dominated by m / z 91 and 104, which may correspond to ethylbenzene. This process generates a set of characteristic ion clusters, clearly distinguishing the ion patterns of different compounds and improving the efficiency of subsequent alignment.

[0028] For example, when comparing a set of characteristic ion clusters with a pre-built compound feature library, a cosine similarity algorithm can be used to evaluate the mass spectrum matching degree. Suppose the feature library records that benzene's mass-to-charge ratio pattern is predominantly at m / z 78, while ethylbenzene's is predominantly at m / z 91 and 104. If the mass-to-charge ratio pattern of a certain characteristic ion cluster has a similarity of 0.95 with the recorded ethylbenzene, it is preliminarily identified as ethylbenzene. This step quickly infers the compound's identity through a high matching degree, reducing false positives.

[0029] In one possible implementation, retention time information of chromatographic peaks can be extracted when a temporary dataset is generated by combining retention times.

[0030] For example, ethylbenzene has a retention time of approximately 9.2 minutes, and its characteristic ion clusters are predominantly m / z 91 and 104. Provisional datasets record this information to ensure the correlation between compound identity and chromatographic behavior, providing a basis for subsequent validation.

[0031] For example, when verifying the consistency between retention time and characteristic ion clusters, the correlation coefficient method in statistical analysis can be used. Assuming a preset consistency threshold of 0.9, if the correlation coefficient between the retention time of ethylbenzene (9.2 minutes) and the characteristic ion cluster pattern reaches 0.93, its final identity is confirmed. This method ensures the reliability of identity determination by quantifying consistency.

[0032] In one possible implementation, when generating the structured dataset, the identity of ethylbenzene, its retention time of 9.2 minutes, and its mass spectrometry characteristics (m / z 91 and 104) can be integrated into a unified format. This dataset is easy to store and retrieve, supporting subsequent environmental monitoring and analysis.

[0033] For example, during data standardization, the ion abundance of mass spectrometry features can be normalized to the range of 0-1, and the retention time unit can be standardized to minutes.

[0034] For example, the m / z91 abundance of ethylbenzene is normalized to 0.85, generating a unified dataset. This formatting facilitates cross-platform analysis and data sharing, improving analytical efficiency.

[0035] Step S103: Through the stability evaluation of the candidate characteristic ions, calculate the relative standard deviation of the peak area of ​​the characteristic ions at different concentration levels. If the relative standard deviation is less than the preset value, then the ion is determined to be suitable as a quantitative ion, and the optimized combination of characteristic ions is obtained.

[0036] Peak area data is obtained from candidate characteristic ions. Data preprocessing methods are used to normalize the peak area data, resulting in a standardized peak area dataset. Based on the standardized peak area dataset, the mean and standard deviation of the peak area for each characteristic ion are calculated for different concentration levels, resulting in a peak area statistical data set. The relative standard deviation is extracted from the peak area statistical data set. If the relative standard deviation is lower than a preset threshold, the characteristic ion is determined to be a quantitative ion, resulting in a preliminary quantitative ion set. For the preliminary quantitative ion set, the K-means clustering algorithm is used to group the peak area data of the characteristic ions, resulting in a set of ion clusters with similar stability. Based on the ion cluster set, the mass-to-charge ratio information of the characteristic ions within each cluster is obtained and compared with a pre-established mass spectrometry feature library to determine the quantitative reliability of the ion clusters, resulting in an optimized quantitative ion set. Mass spectrometry data is extracted from the optimized quantitative ion set, and principal component analysis is used to reduce the dimensionality of the peak area and mass-to-charge ratio of the quantitative ions, resulting in a dimensionality-reduced feature vector set. Based on the dimensionality-reduced feature vector set, the stability score of each quantitative ion at different concentration levels is calculated. If the stability score is higher than a preset threshold, the final quantitative ion combination is determined.

[0037] For example, in the analysis of surface water samples, when obtaining peak area data from candidate characteristic ions, the peak area value of each characteristic ion can be extracted using the chromatographic peak integration method. Assuming the analysis of mass spectrometry data for benzene and ethylbenzene, the peak area of ​​the characteristic ion m / z 78 for benzene is 10000, while the peak areas of the characteristic ions m / z 91 and 104 for ethylbenzene are 12000 and 8000, respectively. In the data preprocessing stage, the maximum value normalization method can be used to map the peak area values ​​to the 0-1 range.

[0038] For example, dividing the peak area of ​​m / z 78 (10000) by the maximum peak area (12000) yields a normalized value of 0.833. The peak areas of m / z 91 and 104 are normalized to 1.000 and 0.667, respectively, generating a standardized peak area dataset. This process ensures the comparability of peak areas for different ions, facilitating subsequent statistical analysis.

[0039] In one possible implementation, the mean peak area and standard deviation of each characteristic ion are calculated for different concentration levels (e.g., 0.1 mg / L, 0.5 mg / L, 1.0 mg / L).

[0040] For example, the peak areas of m / z91 at three concentration levels are 1.000, 0.950, and 0.980, with a mean of 0.977 and a standard deviation of 0.025. The relative standard deviation (RSD) is calculated by dividing the standard deviation by the mean; the RSD of m / z91 is approximately 0.025 / 0.977 ≈ 0.026. If the preset RSD threshold is 0.05, the RSD of m / z91 is below the threshold, indicating it can be identified as a preliminarily quantitative ion. Similarly, the RSDs of m / z78 and 104 are 0.030 and 0.045, respectively, both meeting the requirements and forming a preliminary quantitative ion set. This step screens reliable ions through quantitative stability.

[0041] For example, when using the K-means clustering algorithm to group the peak area data of the preliminary quantitative ion set, the number of clusters can be set to 2. m / z 91 and 104, due to their similar peak area patterns, cluster together, possibly corresponding to ethylbenzene; m / z 78 forms its own cluster, corresponding to benzene. The clustering results generate an ion cluster set, reflecting the similarity between ions. Combined with mass spectrometry feature library comparison, assuming that the characteristics of ethylbenzene in the library are predominantly m / z 91 and 104, and that of benzene is predominantly m / z 78, the compound affiliation of the ion clusters is confirmed through comparison. This process improves the reliability of quantitative analysis through clustering and comparison.

[0042] In one possible implementation, principal component analysis (PCA) is used to reduce the dimensionality of peak area and mass-to-charge ratio data based on the optimized quantitative ion set.

[0043] For example, the peak area data of m / z 91 and 104 were processed by PCA to extract principal components, forming an eigenvector set while retaining 90% of the variance information. The stability scores of each quantitative ion at different concentration levels were calculated. Assuming the score of m / z 91 was based on peak area fluctuations and mass-to-charge ratio consistency, it yielded a score of 0.92, higher than the preset threshold of 0.85, thus confirming it as the final quantitative ion. The scores of m / z 104 and 78 were 0.88 and 0.90, respectively, together constituting the final quantitative ion combination. This method ensures the stability and representativeness of ion selection through dimensionality reduction and scoring, providing a reliable foundation for subsequent quantitative analysis.

[0044] It should be noted that the above process closely integrates with the characteristics of mass spectrometry data, forming a complete analytical chain from peak area extraction to final quantitative ion selection. Each step optimizes ion selection through data standardization, statistical analysis, clustering, and dimensionality reduction, making it suitable for quantitative analysis of compounds in surface water samples.

[0045] Step S104: Construct a peak area ratio matrix using the optimized feature ion combination, and use a support vector machine algorithm to classify and train the ion signal patterns of different volatile organic compounds to obtain a classification model that can distinguish between target compounds and interfering substances.

[0046] Peak area data is obtained from the optimized combination of characteristic ions, and the peak area ratio of each characteristic ion at different concentration levels is calculated to obtain a peak area ratio matrix. Based on the peak area ratio matrix, a support vector machine algorithm is used to classify and train the ion signal patterns to obtain a preliminary classification model. Classification boundary features are extracted from the preliminary classification model and compared with a pre-established volatile organic compound spectral library to determine whether the classification boundary features can distinguish between target compounds and interfering substances, thus obtaining an optimized classification model. If the classification accuracy of the optimized classification model is higher than a preset threshold, the ion signal patterns of the target compounds are extracted from the classification model to generate a feature vector set of the target compounds. Based on the feature vector set, a principal component analysis algorithm is used to reduce the dimensionality of the ion signal patterns of the target compounds to obtain a dimensionality-reduced signal feature set. Using the dimensionality-reduced signal feature set, the signal stability score of each target compound at different concentration levels is calculated, and it is determined whether the signal stability score is higher than a preset threshold to determine the final quantitative analysis model. The ion signal patterns of the target compounds are obtained from the final quantitative analysis model to generate an ion feature database for distinguishing target compounds from interfering substances.

[0047] For example, in the analysis of surface water samples, peak area data is extracted from optimized characteristic ion combinations to construct a peak area ratio matrix. Assuming the analysis of toluene and xylene, the peak area of ​​the characteristic ion m / z91 of toluene is 8000, 8200, and 8500 at concentrations of 0.1 mg / L, 0.5 mg / L, and 1.0 mg / L, respectively; while the peak area of ​​the characteristic ion m / z106 of xylene is 9000, 9200, and 9500, respectively. The peak area ratio is calculated; for example, the ratio of m / z91 to m / z106 at 0.1 mg / L is 8000 / 9000 = 0.889, generating a matrix containing the ratios at each concentration. This matrix reflects the relative changes in ion signals, providing a basis for classification.

[0048] In one possible implementation, a support vector machine (SVM) algorithm is used for classification training based on the peak area ratio matrix. Assuming the matrix is ​​input to the SVM, a linear kernel function is set, and the model is trained to distinguish the signal patterns of toluene and xylene, resulting in a preliminary classification model. The model separates the ion signals of different compounds by optimizing the hyperplane. Classification boundary features, such as the ratio distribution of m / z91 to m / z106, are extracted. Combined with comparison to a volatile organic compound (VOC) spectral library, it is confirmed that toluene is predominantly m / z91, while xylene is predominantly m / z106. Interfering substances such as benzene's m / z78 signal are excluded, and the classification model is optimized.

[0049] For example, if the optimized classification model achieves an accuracy of 95%, which is higher than the preset threshold of 90%, then the ion signal patterns of toluene and xylene are extracted to generate a feature vector set. Assume that the feature vector of toluene contains m / z91 ratios of 0.889, 0.891, and 0.895, while that of xylene is 0.947, 0.951, and 0.953. Principal component analysis is used to reduce the dimensionality of the feature vectors, retaining 80% of the variance, forming a dimensionality-reduced signal feature set, reducing data redundancy and improving analysis efficiency.

[0050] In one possible implementation, a signal stability score is calculated using the dimensionality-reduced signal feature set. Assuming toluene's m / z 91 ratio fluctuates little (score 0.93) and xylene's m / z 106 score is 0.90, both above the threshold of 0.85, the final quantitative analysis model is confirmed. Based on this model, ion signal patterns of toluene and xylene are extracted to generate an ion feature database containing m / z 91 and m / z 106 ratio patterns, used to distinguish target compounds from interfering substances, ensuring the specificity of the quantitative analysis.

[0051] Step S105: Based on the output of the classification model, perform qualitative analysis on the unknown compounds in the test sample. If the classification probability is greater than a preset threshold, determine the compound type and obtain an accurate qualitative identification result.

[0052] Ion signal data is acquired from the sample to be tested, and signal patterns are extracted using mass spectrometry to obtain the ion signal feature set of the sample. The ion signal feature set is matched against a pre-established ion feature database to obtain preliminary compound matching results. Classification feature vectors are extracted from the preliminary compound matching results and input into a classification model to obtain classification probability values. If the classification probability value is greater than a preset threshold, the type of the unknown compound is determined by the classification model, resulting in a qualitative identification result. Based on the qualitative identification result, a clustering analysis algorithm is used to group the ion signal feature set, resulting in signal pattern groups of compounds. The mean signal intensity of each group is calculated to obtain a signal stability score. If the signal stability score is higher than a preset threshold, the feature patterns of the target compound are extracted from the signal pattern groups to generate the final qualitative analysis result.

[0053] For example, in the field of mass spectrometry, the detection of volatile organic compounds (VOCs) in surface water samples can be achieved by acquiring ion signal data using mass spectrometry. After scanning the sample, the mass spectrometer generates a mass spectrum containing multiple ion signals, such as peaks at m / z 78, 91, and 106. These signals reflect the molecular fragmentation characteristics of the compounds in the sample. Assuming the sample may contain benzene, toluene, and xylene, the ion signal intensities extracted by mass spectrometry are 5000 at m / z 78, 8500 at m / z 91, and 9200 at m / z 106, forming a set of ion signal characteristics. This set includes the intensity and relative proportion of each ion peak, providing a data basis for subsequent matching.

[0054] In one possible implementation, a pre-established ion feature database contains ion signal patterns of known compounds, such as benzene predominantly at m / z 78, toluene predominantly at m / z 91, and xylene predominantly at m / z 106. A similarity score is calculated by comparing the sample's ion signal feature set with the database.

[0055] For example, the sample m / z 91 intensity of 8500 has a feature match of 0.92 with the toluene database, which is higher than the threshold of 0.85, and is preliminarily determined to possibly contain toluene. Similarly, the sample m / z 106 intensity of 9200 has a match of 0.90 with xylene, and the preliminary matching result indicates that it contains both toluene and xylene.

[0056] For example, a classification feature vector is extracted from the initial matching results. For instance, the intensity ratio of m / z91 to m / z106 is 0.924, combined with the intensity ratio of m / z78 (0.543), to form a feature vector. This vector is then input into a pre-trained classification model, such as a random forest model, which outputs classification probability values. Assuming the probability value for toluene is 0.93 and for xylene is 0.91, both exceeding the threshold of 0.80, the sample is confirmed to contain both toluene and xylene. This process quantifies signal patterns using feature vectors, improving the accuracy of qualitative identification.

[0057] In one possible implementation, based on the qualitative identification results, the K-means clustering algorithm is used to group the ion signal feature set. Assume that after clustering, two groups are formed: one group dominated by m / z 91, and the other dominated by m / z 106. The mean signal intensity of each group is calculated; for example, the mean of the m / z 91 group is 8300, and the mean of the m / z 106 group is 9100. Further calculation of the signal stability score shows that the m / z 91 group has a smaller fluctuation range and a score of 0.94; the m / z 106 group has a score of 0.91, both higher than the threshold of 0.85, indicating signal stability.

[0058] For example, characteristic patterns of target compounds are extracted from signal pattern groupings to generate final qualitative analysis results. Toluene's characteristic patterns are primarily based on its m / z 91 intensity and ratio, while xylene's are mainly based on m / z 106. These characteristic patterns form updated data in the ion characteristic database for subsequent sample analysis. This method, through multi-step validation, ensures the specificity and reliability of the qualitative results, supporting the rapid identification of compounds in environmental samples.

[0059] Step S106: Using the target compound identified in the qualitative identification results, establish a concentration standard curve using the corresponding quantitative ion peak area, calculate the linear relationship parameter between peak area and concentration, and obtain the regression equation for quantitative analysis.

[0060] Quantitative ion peak area data of the target compound is obtained from the qualitative identification results. Ion signal intensity is extracted using mass spectrometry to obtain a peak area dataset. Multiple samples of known concentrations are prepared using standard solutions, and the ion peak area of ​​each sample is measured using mass spectrometry to obtain a dataset corresponding to concentration and peak area. Linear regression analysis of the concentration-peak area dataset is performed using the least squares method to calculate the slope 'a' and intercept 'b', resulting in the regression equation Y = aX + b, where Y represents peak area, X represents concentration, 'a' represents slope, and 'b' represents intercept. If the linear correlation coefficient of the regression equation is greater than a preset threshold, a quantitative analysis model for the target compound is determined using the regression equation, resulting in a standardized concentration prediction formula. Ion signal intensity is obtained from the sample to be tested, and quantitative ion peak area is extracted using mass spectrometry. This is input into the regression equation to calculate the sample concentration, obtaining the quantitative analysis results. By repeatedly measuring the ion peak area of ​​the sample to be tested, the coefficient of variation of the peak area is calculated using statistical analysis to obtain an evaluation value of the analytical precision. If the evaluation value of the analytical precision is less than a preset threshold, the concentration value of the target compound is extracted from the quantitative analysis results to generate the final quantitative analysis data.

[0061] For example, in the field of mass spectrometry, the quantitative analysis of volatile organic compounds (VOCs) in surface water samples can be achieved by acquiring ion peak area data through mass spectrometry to determine the concentration of target compounds. Assuming the analytes are toluene and xylene, after the mass spectrometer scans the sample, it generates ion signal data, such as the peak areas at m / z 91 and m / z 106, which correspond to the characteristic ions of toluene and xylene, respectively. Mass spectrometry analysis obtains the peak area dataset by integrating the ion signal intensity, providing a foundation for subsequent quantitative analysis.

[0062] Specifically, the preparation of standard solutions is a crucial step in quantitative analysis. Standard solutions of toluene with concentrations of 0.1, 0.5, 1.0, 2.0, and 5.0 mg / L can be prepared and measured by mass spectrometry to obtain peak area data at m / z 91, such as 500, 2500, 5000, 10000, and 25000. Similarly, xylene standard solutions generate peak area data at m / z 106. These data form a dataset corresponding to concentration and peak area, reflecting the linear relationship between the two.

[0063] It should be noted that the preparation of standard solutions must ensure a reasonable concentration gradient that covers the expected concentration range of the test samples in order to improve the applicability of the regression model.

[0064] In one embodiment, the least squares method is used to perform linear regression analysis on the dataset. Taking toluene as an example, assuming the calculated regression equation is Y = 5000X + 200, where Y is the m / z91 peak area, X is the concentration, the slope is 5000, and the intercept is 200, the linear correlation coefficient is 0.98, which is higher than the preset threshold of 0.95, indicating that the regression equation is reliable and can be used as a quantitative analysis model for toluene. The regression equation for xylene is generated similarly, such as Y = 4800X + 150, with a correlation coefficient of 0.97. This model can be directly used for concentration prediction, ensuring accurate results.

[0065] For example, mass spectrometry analysis of the sample showed a peak area of ​​7500 at m / z 91. Substituting this into the toluene regression equation, the calculated concentration was approximately 1.46 mg / L. Similarly, the peak area at m / z 106 was 8000, and substituting this into the xylene equation, the concentration was approximately 1.60 mg / L. These results directly map concentration to peak area, simplifying the quantitative procedure.

[0066] It should be noted that measuring the peak area of ​​the sample multiple times can assess the accuracy of the analysis.

[0067] For example, the five measurements of the m / z91 peak area of ​​toluene were 7450, 7500, 7550, 7480, and 7520, respectively, with a calculated coefficient of variation of 0.008, which is less than the preset threshold of 0.05, indicating high analytical precision and reliable results.

[0068] In one embodiment, the quantitative results can further verify signal stability.

[0069] For example, by comparing the peak area fluctuations from multiple measurements, the signal consistency between toluene and xylene can be confirmed. Small fluctuations indicate stable mass spectrometry operation and high data reliability. This method, through standardized quantitative models and accuracy assessments, ensures the reliability of concentration results, supporting the rapid determination of pollutant concentrations in environmental monitoring.

[0070] Step S107: The peak area of ​​the target compound in the actual sample is converted into concentration based on the regression equation. At the same time, the peak recognition algorithm is used to separate the overlapping peaks in the chromatogram to obtain the corrected peak area value for the final concentration determination.

[0071] Ion signal data are obtained from the chromatograms of actual samples. Mass spectrometry is used to extract the peak areas of quantitative ions, resulting in an initial peak area dataset. If overlapping peaks exist in the initial peak area dataset, a Gaussian fitting algorithm is used to decompose the overlapping peaks in the chromatogram, obtaining separated peak area data. The separated peak area data is substituted into a pre-established regression equation Y=aX+b, where Y represents peak area, X represents concentration, a represents slope, and b represents intercept, to calculate the preliminary concentration value of the target compound. Outliers are extracted from the preliminary concentration values, and a box plot algorithm is used to detect outliers in the concentration data, resulting in a concentration dataset after outlier removal. For the concentration dataset after outlier removal, the concentration values ​​from multiple measurements are averaged using a mean calculation method to obtain standardized concentration values. If the standardized concentration value does not conform to the preset concentration threshold range, a second peak separation process is performed on the chromatogram. The updated peak area data is then substituted back into the regression equation to calculate the corrected concentration value. Using the corrected concentration value and the signal intensity of the quantitative ions, a linear interpolation method is used to smooth the concentration value, obtaining the final concentration determination result.

[0072] For example, in the field of mass spectrometry, acquiring ion signal data from chromatograms is a crucial step in the quantitative analysis of volatile organic compounds (VOCs) in surface water samples. Chromatograms are generated by scanning the sample with a mass spectrometer, recording the signal intensity of specific ions, such as m / z 91, to form an initial peak area dataset. Assuming the analyte is benzene, and the mass spectrometer detects an ion peak at m / z 78, the initial peak area data would be 6000, 6200, and 6100. If peak asymmetry is observed, it may be due to overlapping peaks caused by similar retention times of compounds in the chromatographic column, affecting quantitative accuracy.

[0073] In one possible implementation, a Gaussian fitting algorithm is used to decompose overlapping peaks. For the m / z 78 peak of benzene, the chromatogram shows two overlapping peaks, corresponding to benzene and another interfering compound, respectively. Gaussian fitting, through mathematical modeling, decomposes the overlapping peak into independent peaks, yielding a peak area of ​​5800 for benzene. This method effectively separates interfering signals, ensuring that the peak area data accurately reflects the concentration of the target compound.

[0074] For example, based on a pre-established regression equation Y=6000X+100, where Y is the peak area, X is the concentration, the slope is 6000, and the intercept is 100, substituting the separated peak area of ​​5800, the initial concentration of benzene is calculated to be 0.95 mg / L. Multiple initial concentration measurements, such as 0.95, 0.97, and 1.05 mg / L, require further outlier detection. Using a box plot algorithm, with a set interquartile range, 1.05 mg / L was identified as an outlier; after removal, the remaining concentrations were 0.95 and 0.97 mg / L.

[0075] In one possible implementation, the average of the concentration dataset after outlier removal is taken to obtain a standardized concentration value of 0.96 mg / L. If this value exceeds a preset threshold range of 0.5-0.9 mg / L, secondary peak separation is required. The chromatogram is reanalyzed, the Gaussian fitting parameters are adjusted to obtain an updated peak area of ​​5900, which is then substituted into the regression equation to correct the concentration to 0.98 mg / L. This step effectively corrects the deviation caused by insufficient initial peak separation.

[0076] For example, the corrected concentration value is smoothed using linear interpolation, taking into account the signal intensity at m / z 78. Assuming multiple concentration measurements are 0.98 and 0.99 mg / L, interpolation yields a final concentration of 0.985 mg / L. This method improves the stability of concentration measurements by smoothing fluctuations, ensuring results are closer to the true value. This multi-step analysis process, from peak separation to outlier removal and data smoothing, is interconnected, ensuring the reliability of quantitative analysis and providing accurate data support for environmental monitoring.

[0077] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit the scope of one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this specification should be included within the protection scope of one or more embodiments of this specification.

Claims

1. A method for determining volatile organic compounds in surface water by purge-and-trap, gas chromatography-mass spectrometry, characterized in that, The method includes: Mass spectrometry data of volatile organic compounds in surface water samples were obtained, and multi-component mixtures in complex matrices were separated and detected by gas chromatography-mass spectrometry to obtain a raw data set containing retention time and mass spectrum information. Based on the mass spectrum information in the original dataset, a spectral matching algorithm is used to analyze the mass spectrometric characteristics of each chromatographic peak. If the ion abundance in the mass spectrum is greater than a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, then the candidate characteristic ion corresponding to the peak is determined. By evaluating the stability of the candidate characteristic ions, the relative standard deviation of the peak area of ​​the characteristic ions at different concentration levels is calculated. If the relative standard deviation is less than the preset value, the ion is determined to be suitable as a quantitative ion, and the optimized combination of characteristic ions is obtained. The optimized characteristic ion combination is used to construct a peak area ratio matrix. The ion signal patterns of different volatile organic compounds are classified and trained using the support vector machine algorithm to obtain a classification model that can distinguish between target compounds and interfering substances. Based on the output of the classification model, a qualitative analysis is performed on the unknown compounds in the test sample. If the classification probability is greater than a preset threshold, the compound type is determined and an accurate qualitative identification result is obtained. Based on the target compounds identified in the qualitative identification results, a concentration standard curve is established using the corresponding quantitative ion peak area. The linear relationship parameters between peak area and concentration are calculated to obtain the regression equation for quantitative analysis. The peak area of ​​the target compound in the actual sample is converted into concentration based on the regression equation. At the same time, the peak recognition algorithm is used to separate the overlapping peaks in the chromatogram to obtain the corrected peak area value for the final concentration determination.

2. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The process involves acquiring mass spectrometry data of volatile organic compounds in surface water samples, separating and detecting multi-component mixtures in complex matrices using gas chromatography-mass spectrometry (GC-MS), and obtaining a raw data set containing retention time and mass spectrum information, including: Surface water samples were pretreated using gas chromatography-mass spectrometry to separate volatile organic compounds and obtain a raw data set including retention times and mass spectra. If the signal strength of the retained time in the original dataset is lower than the preset threshold, a signal enhancement algorithm is used to denoise the data to obtain an enhanced dataset. Based on the enhanced dataset, principal component analysis algorithm is used to extract the main features from the mass spectrum information to obtain a set of feature vectors. If there are outliers in the feature vector set, the outliers are identified and removed by an anomaly detection algorithm to obtain a cleaned feature vector set. Using the cleaned feature vector set, the support vector machine algorithm is used to classify volatile organic compounds and determine the category of each compound. Based on the classification results, the chemical structure information of the compounds is obtained by matching them with a pre-established volatile organic compound mass spectrometry database. By combining chemical structure information with retention time, a dataset containing compound categories and structures is generated.

3. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The step involves analyzing the mass spectrometric characteristics of each chromatographic peak using a spectral matching algorithm based on the mass spectrum information in the original dataset. If the ion abundance in the mass spectrum is greater than a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, then the candidate characteristic ions corresponding to that peak are determined, including: Mass spectrum information is extracted from the original dataset, and chromatographic peak characteristics are analyzed using a spectral matching algorithm. If the ion abundance exceeds a preset threshold and the mass-to-charge ratio matches the characteristics of the target compound, then candidate characteristic ions are determined. Based on the candidate characteristic ions, a clustering analysis algorithm is used to group characteristic ions with similar mass-to-charge ratios in the mass spectrum information to obtain a set of characteristic ion clusters. By comparing the set of characteristic ion clusters with a pre-established compound feature library, if the mass-to-charge ratio pattern of the characteristic ion clusters matches the records in the compound feature library, the preliminary identity of the corresponding compound is determined. Based on the initial identity, the retention time information of the chromatographic peaks is obtained. Combined with the distribution characteristics of characteristic ion clusters in the mass spectrum information, a temporary data set containing retention time and compound identity is generated. Using a temporary dataset, statistical analysis methods are employed to verify the consistency between retention time and characteristic ion clusters. If the consistency exceeds a preset threshold, the final identity of the compound is determined. Based on the final identity, a structured data set containing compound identity, retention time, and mass spectrometry features is generated; By using a structured dataset, the compound identities and mass spectrometry characteristics are formatted using data standardization methods, resulting in a unified dataset that can be used for subsequent analysis.

4. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The process involves evaluating the stability of the candidate characteristic ions, calculating the relative standard deviation of the peak area of ​​the characteristic ions at different concentration levels, and determining that the ion is suitable as a quantitative ion if the relative standard deviation is less than a preset value, thereby obtaining an optimized combination of characteristic ions. This includes: Peak area data were obtained from candidate feature ions, and the peak area data were normalized using data preprocessing methods to obtain a standardized peak area dataset. Based on the standardized peak area dataset, the mean and standard deviation of the peak area of ​​each characteristic ion are calculated for different concentration levels to obtain a peak area statistical data set. The relative standard deviation is extracted from the peak area statistical data set. If the relative standard deviation is lower than the preset threshold, the characteristic ion is determined to be a quantitative ion, and a preliminary quantitative ion set is obtained. For the preliminary quantitative ion set, the K-means clustering algorithm is used to group the peak area data of characteristic ions to obtain a set of ion clusters with similar stability; Based on the ion cluster set, the mass-to-charge ratio information of characteristic ions in each cluster is obtained, and compared with the pre-established mass spectrometry feature library to determine the quantitative reliability of the ion cluster and obtain the optimized quantitative ion set. Mass spectrometry data were extracted from the optimized quantitative ion set, and principal component analysis algorithm was used to reduce the dimensionality of the peak area and mass-to-charge ratio of the quantitative ions to obtain the dimensionality-reduced feature vector set. Based on the reduced feature vector set, the stability score of each quantitative ion at different concentration levels is calculated. If the stability score is higher than the preset threshold, the final quantitative ion combination is determined.

5. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The optimized characteristic ion combination is used to construct a peak area ratio matrix. A support vector machine algorithm is then used to classify and train the ion signal patterns of different volatile organic compounds, resulting in a classification model capable of distinguishing target compounds from interfering substances. This model includes: Peak area data is obtained from the optimized combination of characteristic ions, and the peak area ratio of each characteristic ion at different concentration levels is calculated to obtain the peak area ratio matrix. Based on the peak area ratio matrix, the support vector machine algorithm is used to classify and train the ion signal patterns to obtain a preliminary classification model; The classification boundary features are extracted from the preliminary classification model and compared with the pre-established volatile organic compound spectral library to determine whether the classification boundary features can distinguish the target compound from the interfering substances, thus obtaining the optimized classification model. If the classification accuracy of the optimized classification model is higher than the preset threshold, the ion signal pattern of the target compound is extracted from the classification model to generate a feature vector set of the target compound. Based on the feature vector set, the principal component analysis algorithm is used to reduce the dimensionality of the ion signal pattern of the target compound, and the reduced signal feature set is obtained. By using the reduced signal feature set, the signal stability score of each target compound at different concentration levels is calculated, and it is determined whether the signal stability score is higher than the preset threshold to determine the final quantitative analysis model. Ion signal patterns of the target compound are obtained from the final quantitative analysis model, and an ion feature database is generated to distinguish the target compound from interfering substances.

6. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The step involves performing qualitative analysis on unknown compounds in the test sample based on the output of the classification model. If the classification probability is greater than a preset threshold, the compound type is determined, and an accurate qualitative identification result is obtained. This includes: Ion signal data are obtained from the sample to be tested, and signal patterns are extracted using mass spectrometry analysis to obtain the set of ion signal features of the sample. By matching the set of ion signal features using a pre-established ion feature database, preliminary compound matching results are obtained. Extract classification feature vectors from the initial compound matching results, input them into the classification model, and obtain classification probability values; If the classification probability value is greater than the preset threshold, the type of the unknown compound is determined by the output of the classification model, and a qualitative identification result is obtained. Based on the qualitative identification results, a clustering analysis algorithm is used to group the ion signal feature set to obtain the signal pattern grouping of the compound; By grouping signals by pattern, the mean signal strength of each group is calculated to obtain a signal stability score. If the signal stability score is higher than the preset threshold, the characteristic patterns of the target compound are extracted from the signal pattern grouping to generate the final qualitative analysis results.

7. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The target compound identified through the qualitative identification results is used to establish a concentration standard curve using its corresponding quantitative ion peak area. The linear relationship parameters between peak area and concentration are calculated to obtain the regression equation for quantitative analysis, including: Quantitative ion peak area data of the target compound are obtained from the qualitative identification results, and ion signal intensity is extracted using mass spectrometry analysis to obtain the peak area dataset. Multiple samples of known concentrations were prepared using standard solutions, and the ion peak area of ​​each sample was measured using mass spectrometry to obtain a dataset corresponding to the concentration and peak area. The least squares method was used to perform linear regression analysis on the corresponding datasets of concentration and peak area, and the slope a and intercept b were calculated to obtain the regression equation Y=aX+b, where Y represents peak area, X represents concentration, a represents slope and b represents intercept. If the linear correlation coefficient of the regression equation is greater than the preset threshold, the quantitative analysis model of the target compound is determined by the regression equation, and a standardized concentration prediction formula is obtained. The ion signal intensity is obtained from the sample to be tested, the quantitative ion peak area is extracted by mass spectrometry, the sample concentration is calculated by inputting the regression equation, and the quantitative analysis results are obtained. By repeatedly measuring the ion peak area of ​​the sample to be tested, the coefficient of variation of the peak area is calculated using statistical analysis to obtain an evaluation value of the analytical accuracy. If the assessed value of analytical precision is less than the preset threshold, the concentration value of the target compound is extracted from the quantitative analysis results to generate the final quantitative analysis data.

8. The method for determining volatile organic compounds in surface water by purge-and-trap and gas chromatography-mass spectrometry according to claim 1, characterized in that, The process of converting the peak area of ​​the target compound in the actual sample to concentration based on the regression equation, and simultaneously using a peak identification algorithm to separate overlapping peaks in the chromatogram to obtain corrected peak area values ​​for final concentration determination, includes: Ion signal data were obtained from the chromatograms of actual samples, and the quantitative ion peak areas were extracted using mass spectrometry to obtain an initial peak area dataset. If there are overlapping peaks in the initial peak area data, the Gaussian fitting algorithm is used to decompose the overlapping peaks in the chromatogram to obtain the separated peak area data. By substituting the separated peak area data into the pre-established regression equation Y=aX+b, where Y represents peak area, X represents concentration, a represents slope, and b represents intercept, the preliminary concentration value of the target compound is calculated. Outliers are extracted from the initial concentration values, and outliers in the concentration data are detected using a box plot algorithm to obtain the concentration dataset after removing outliers. For the concentration dataset after removing outliers, the average of the concentration values ​​measured multiple times is calculated using the mean calculation method to obtain standardized concentration values. If the standardized concentration value does not match the preset concentration threshold range, the chromatogram is subjected to secondary peak separation processing, and the updated peak area data is resubmitted into the regression equation to calculate the corrected concentration value. By combining the corrected concentration value with the signal intensity of the quantitative ions, a linear interpolation method is used to smooth the concentration value, thus obtaining the final concentration measurement result.

Images