A real-time mass spectrometry analysis system for detecting volatile gas components of grain

Abstract

The present application relates to the technical field of gas detection, in particular to a real-time mass spectrum analysis system for detecting volatile gas components of grain, which comprises a sample spectrum acquisition unit for obtaining original sample spectra at different depths, then weighting each depth original spectrum according to preset weight, and outputting characteristic fingerprint spectrum; a differential background suppression unit for acquiring real-time background spectrum, and using a closed-loop adaptive multi-scale background suppression method to remove background ion signals of the original sample spectrum based on the real-time background spectrum and the characteristic fingerprint spectrum, and generating a net characteristic spectrum without background; and a target VOC determination unit for determining target VOC in the grain pile based on the net characteristic spectrum through a triple verification mechanism. The system improves the recognition sensitivity and quantitative accuracy of the system for low-concentration target volatile compounds, enhances the adaptability and robustness of the system in a complex grain storage environment, and effectively avoids the misjudgment and missed detection problems existing in traditional methods.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas detection, in particular to a real-time mass spectrum analysis system for volatile gas component detection of grain. BACKGROUND

[0002] In the process of grain storage and transportation, complex physical, chemical and microbial changes often occur inside the grain pile, which releases a variety of volatile organic compounds (VOCs) with low concentration. These VOCs not only reflect the evolution trend of grain quality, but also can be used as key markers for early identification of mold, insect damage or heat problems. Therefore, real-time detection of grain VOCs has become an important research direction to ensure the safety of grain storage and quality management.

[0003] Mass spectrometry technology has a wide application prospect in volatile gas monitoring due to its high sensitivity, wide dynamic range and ability to analyze multiple components simultaneously. In particular, in the application scenario of headspace analysis combined with online monitoring, with the help of automated sampling and spectrum processing technology, the rapid detection and dynamic response of gas components at different depths of the grain pile can be realized, providing important technical support for accurate grain condition evaluation.

[0004] The traditional method assumes that each m / z channel has the same weight, without considering the importance difference of different ion channels in feature expression, resulting in the dominant role of high background area on the overall similarity, and the dilution or even neglect of low-abundance but representative target feature peaks in the calculation. In the complex environment of the grain depot, the background gas composition fluctuates greatly, and there are a large number of short-time interference sources. The traditional method is difficult to adapt to the situation, and there are a large number of interference substances (ethanol, CO2, etc.) in the grain depot, so the mass spectrometry identification of grain VOCs is often misjudged. Therefore, a real-time mass spectrum analysis system for volatile gas component detection of grain is designed. SUMMARY

[0005] The purpose of the present application is to provide a real-time mass spectrum analysis system for volatile gas component detection of grain to solve the problem of assuming that each m / z channel has the same weight without considering the importance difference of different ion channels in feature expression in the background art.

[0006] To achieve the above purpose, the present application provides a real-time mass spectrum analysis system for volatile gas component detection of grain, which comprises:

[0007] A sample spectrum acquisition unit is used to obtain headspace sample gas at different depths of the grain pile, and the headspace sample gas is scanned in parallel within the m / z range, thereby obtaining original sample spectra at different depths. Then, the original spectra at each depth are weighted according to the preset weight, and a characteristic fingerprint spectrum is output.

[0008] A differential background suppression unit is configured to collect a real-time background spectrum and remove background ion signals of an original sample spectrum based on the real-time background spectrum and a characteristic fingerprint spectrum by using a closed-loop adaptive multi-scale background suppression method to generate a net characteristic spectrum with background removed;

[0009] A target VOC determination unit is configured to determine a target VOC in the grain pile based on the net characteristic spectrum by using a triple verification mechanism, output a target VOC name and a characteristic m / z, and perform secondary review on the target VOC.

[0010] As a further improvement of the technical solution, the differential background suppression unit comprises a background spectrum collection module and a background deduction module.

[0011] The background spectrum collection module is provided with a multi-way valve capable of automatically switching between background gas and sample gas, and the background gas is automatically collected by the multi-way valve after the headspace sample gas is obtained. The background gas is scanned in parallel in the m / z range to obtain a real-time background spectrum.

[0012] The background deduction module uses a closed-loop adaptive multi-scale background suppression method to fuse and process the input characteristic fingerprint spectrum and real-time background spectrum to output a net characteristic spectrum with background removed.

[0013] As a further improvement of the technical solution, the background deduction module outputs a net characteristic spectrum with background removed by the following specific steps:

[0014] S21, the input characteristic fingerprint spectrum and real-time background spectrum are respectively subjected to peak alignment and normalization processing to obtain normalized sample spectrum intensity and normalized background spectrum intensity ; and a dynamic time warping algorithm is used to fine-tune and align the main peak positions in the characteristic fingerprint spectrum and real-time background spectrum.

[0015] S22, the normalized sample spectrum intensity and the normalized background spectrum intensity are used to calculate the initial differential spectrum of each m / z channel and the correlation coefficient of the sample spectrum and the background spectrum , the background removal coefficient is determined according to the correlation coefficient , the adaptive differential spectrum is obtained based on the background removal coefficient ;

[0016] ;

[0017] ;

[0018] ;

[0019] wherein, is the average of the sample spectrum; is the average of the background spectrum; is the reference similarity threshold; is the adaptive background removal sensitivity factor;

[0020] obtaining an adaptive difference spectrum :

[0021] ;

[0022] S23, in the determination of the background removal coefficient process, the feature peak specificity weight is introduced, and the correlation coefficient is optimized as a weighted Pearson correlation coefficient , and then an optimized adaptive difference spectrum is generated;

[0023] S24, the optimized adaptive difference spectrum is subjected to discrete wavelet transform, decomposed into a plurality of scale coefficients, then soft threshold denoising is applied on each scale, and then inverse wavelet transform is used to reconstruct the denoised spectrum, to obtain a multi-scale smoothed reconstructed spectrum ;

[0024] S25, the reconstructed spectrum is used to calculate a residual spectrum with the optimized adaptive difference spectrum , and if the root mean square of the residual spectrum is higher than a preset threshold, the background coefficient is adjusted;

[0025] S26, steps S21 to S25 are repeated until the root mean square of the residual spectrum is less than or equal to a preset root mean square threshold, and then a background-removed spectrum after multiple iterations is obtained ;

[0026] S27, the background-removed spectrum after multiple iterations is inversely normalized according to the original intensity to generate a background-removed net feature spectrum .

[0027] As a further improvement of the technical solution, in S23, the specific steps of generating the optimized adaptive difference spectrum are as follows:

[0028] S231, based on a preset feature channel weight factor library, a weight value is assigned to each m / z channel ;

[0029] S232, generating an optimized correlation coefficient based on the correlation coefficient and the weight value optimizing the correlation coefficient into a weighted Pearson correlation coefficient ;

[0030] ;

[0031] S233, generating an optimized differential coefficient based on the weighted Pearson correlation coefficient and the weight value ;

[0032] ;

[0033] wherein, is a characteristic protection factor;

[0034] S234, generating an optimized adaptive differential spectrum based on the differential coefficient ;

[0035] .

[0036] As a further improvement of the technical solution, in the S231, the characteristic channel weight factor library is constructed by the following method:

[0037] extracting historical characteristic m / z channels of key compounds from a historical grain volatile substance mass spectrum library and assigning weights ; the historical grain volatile substance mass spectrum library is updated in real time based on the target VOC determination unit;

[0038] assigning weights to m / z channels in the historical background spectrum whose fluctuation variance exceeds a threshold value .

[0039] As a further improvement of the technical solution, the target VOC determination unit comprises a mass spectrum library updating module, and the mass spectrum library updating module comprises a historical grain volatile substance mass spectrum library;

[0040] The mass spectrum library updating module is used for secondary review of the target VOC, and after the secondary review is passed, the target VOC is assigned an initial weight, and the target VOC name, characteristic m / z, initial weight and time stamp are stored in the historical grain volatile substance mass spectrum library; if the review fails, the detection log of the secondary review of the target VOC is recorded for subsequent review.

[0041] ​​As a further improvement to this technical solution, the target VOC determination unit includes a triple verification mechanism comprising a first verification mechanism, a second verification mechanism, and a third verification mechanism; wherein, the first verification mechanism is a characteristic ion screening verification, the second verification mechanism is a spectrum matching verification, and the third verification mechanism is an interference elimination verification.

[0042] As a further improvement to this technical solution, the characteristic ion screening and verification is carried out by constructing a candidate ion library. The candidate ion library is tested for uniqueness, abundance stability and temperature and humidity adaptability. The ions that pass the tests are used as the characteristic ions for the final characterization of the target VOC.

[0043] As a further improvement to this technical solution, the spectrum matching verification calls the characteristic ion channels and their weights corresponding to each target compound in the candidate ion library, and uses a weighted cosine similarity algorithm to match the ion intensity vector of the net feature spectrum on the channel with the corresponding vector of each candidate compound in the historical mass spectrometry library of grain volatiles. When the similarity score is greater than or equal to the similarity judgment threshold, the compound corresponding to the candidate ion library is confirmed as the target VOC.

[0044] As a further improvement to this technical solution, the interference elimination verification is used to perform secondary screening on the matching results of the second verification mechanism. When the overlap between the target peak of the target VOC and the interference peak of the grain warehouse interference is greater than the overlap threshold, the interference peak is removed to confirm the final target VOC.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] 1. In this real-time mass spectrometry analysis system for detecting volatile gas components in grain, characteristic peak specific weights are introduced. By strengthening the retention of key low-abundance peaks and suppressing the interference contribution of high background regions, the optimized adaptive differential spectrum obtained has a higher signal-to-noise ratio and clearer spectral details. This significantly improves the system's sensitivity and quantitative accuracy in identifying low-concentration target volatile compounds, and enhances the system's adaptability and robustness in complex grain storage environments.

[0047] 2. In the real-time mass spectrometry analysis system for detecting volatile gas components in grain, the target VOC identification module adopts a triple mechanism of characteristic ion screening verification, spectrum matching verification, and interference elimination verification, which improves the accuracy and robustness of target VOC identification, effectively avoids the problems of misjudgment and missed detection in traditional methods, and significantly improves the reliability of VOC identification and the practicality of the system in the complex environment of actual grain warehouses. Attached Figure Description

[0048] Figure 1 This is an overall flowchart of the present invention;

[0049] The meanings of the various reference signs in the figures are as follows:

[0050] 1, sample spectrum acquisition unit; 2, differential background suppression unit; 21, background spectrum acquisition module; 22, background subtraction module; 3, target VOC determination unit; 31, mass spectrum library updating module. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0052] Embodiment: Please refer to Figure 1 As shown in the figure, a real-time mass spectrum analysis system for detecting volatile gas components of grain is provided, which comprises a sample spectrum acquisition unit 1, a differential background suppression unit 2 and a target VOC determination unit 3:

[0053] The sample spectrum acquisition unit 1 is used to obtain headspace sample gas from different depths of the grain pile, and the headspace sample gas is scanned in parallel in the m / z range, so as to obtain original sample spectra of different depths. Then, the original spectra of each depth are weighted according to a preset weight, and a characteristic fingerprint spectrum is output.

[0054] Different depths include the surface layer, the middle layer and the bottom layer. Because of thermodynamic equilibrium and microbial activity in the grain pile, the concentration and composition of volatile substances may be different at different depths. For example, the middle layer may produce certain deterioration products due to more favorable temperature and humidity conditions, while the surface layer or the bottom layer is different. Therefore, multi-point sampling can obtain more accurate overall quality state.

[0055] m / z (mass-to-charge ratio) refers to the ratio of the mass of an ion to the number of charges it carries, which is a basic parameter for distinguishing different compound ions in mass spectrum analysis. The present scheme scans in the range of 20-300.

[0056] The differential background suppression unit 2 is used to acquire a real-time background spectrum, and a closed-loop adaptive multi-scale background suppression method is adopted to remove the background ion signal of the original sample spectrum based on the real-time background spectrum and the characteristic fingerprint spectrum, thereby generating a net characteristic spectrum without background.

[0057] The differential background suppression unit 2 comprises a background spectrum acquisition module 21 and a background subtraction module 22.

[0058] The background spectrum acquisition module 21 is provided with a multi-way valve capable of automatically switching the background gas and the sample gas, and the background gas is automatically acquired through the multi-way valve after the headspace sample gas is acquired. The background gas is scanned in parallel in the m / z range, and the real-time background spectrum is obtained. The sampling time interval of the background gas and the sample gas is less than or equal to 2 seconds, which ensures the time proximity with the sample spectrum and eliminates the background deviation caused by the fluctuation of environmental VOCs.

[0059] The background deduction module 22 adopts a closed-loop adaptive multi-scale background suppression method to fuse and process the input characteristic fingerprint spectrum and the real-time background spectrum, and outputs a net characteristic spectrum after background deduction.

[0060] The specific steps of outputting the net characteristic spectrum after background deduction in the background deduction module 22 are as follows:

[0061] S21, the input characteristic fingerprint spectrum and the real-time background spectrum are respectively subjected to peak alignment and normalization processing, to obtain normalized sample spectrum intensity and normalized background spectrum intensity ; and a dynamic time warping algorithm is used to finely adjust and align the main peak positions in the characteristic fingerprint spectrum and the real-time background spectrum, to eliminate the peak position deviation caused by scanning delay or drift.

[0062]

[0063]

[0064] In the formula, is the normalized sample spectrum intensity; is the sample spectrum intensity (ion count intensity) of the original characteristic fingerprint spectrum at m / z; is the normalized background spectrum intensity; is the background spectrum intensity of the real-time background spectrum at m / z;

[0065] The main peak positions are accurately aligned by the dynamic time warping, and the peak position deviation caused by scanning delay or drift is eliminated; the normalization processing eliminates the intensity difference between different sampling batches or channels, and provides a unified signal reference for subsequent differential operation.

[0066] S22, the normalized sample spectrum intensity and the normalized background spectrum intensity are used to calculate the initial differential spectrum of each m / z channel and the correlation coefficient of the sample spectrum and the background spectrum, the background deduction coefficient is determined according to the correlation coefficient , the adaptive differential spectrum is obtained based on the background deduction coefficient ;

[0067] ;

[0068] ;

[0069] ;

[0070] wherein, is the average value of the sample spectrum; is the average value of the background spectrum; is the reference similarity threshold value; is the adaptive background removal sensitivity factor, and the value range is ;

[0071] Obtaining an adaptive differential spectrum :

[0072] ;

[0073] The normalized sample spectrum and the background spectrum are directly differentiated to preliminarily strip the environmental background ion interference, so as to realize the coarse separation of the target signal and the environmental signal; the background removal coefficient α is dynamically determined based on the cross-correlation coefficient of the spectrum, so that the background removal strength can be adaptively adjusted according to the similarity between the environmental background and the sample spectrum, which not only ensures that the background can be effectively removed when the similarity is high, but also avoids excessive removal of the real signal.

[0074] S23, in the process of determining the background removal coefficient , a feature peak specificity weight is introduced, and the correlation coefficient is optimized as a weighted Pearson correlation coefficient , and then an optimized adaptive differential spectrum is generated.

[0075] The traditional correlation coefficient r regards the entire spectrum as homogeneous data, but in actual application: the key feature peak has low intensity, but the difference with the background spectrum is significant (should be strengthened to retain). Even if the intensity of the high background area is high, the sample / background similarity is also high (should be strengthened to deduct). The global r value cannot reflect the local importance difference, resulting in that the low-abundance feature peak is submerged in noise in the differential. Therefore, in S23, the specific steps of generating the optimized adaptive differential spectrum are as follows:

[0076] S231, based on a preset feature channel weight factor library, a weight value is allocated to each m / z channel .

[0077] In S231, the feature channel weight factor library is constructed by the following method:

[0078] The historical key compound characteristic m / z channel is extracted from the historical mass spectrum library of grain volatile substances, and a weight is assigned ; the historical mass spectrum library of grain volatile substances is updated in real time based on the target VOC determination unit 3; the m / z channel with a fluctuation variance exceeding a threshold in the historical background spectrum is assigned a weight .

[0079] The feature channel weight factor library is optimized once a month by a machine learning model, using historical detection data as the training set and VOC recognition accuracy as the target, to inversely fit the optimal weight of each m / z channel;

[0080] S232, the correlation coefficient is optimized as a weighted Pearson correlation coefficient ;

[0081] ;

[0082] The difference in the high-weight feature area is amplified, and the influence of the high-background area is suppressed, so that the spectrum difference in the key area is more accurately reflected;

[0083] S233, based on the weighted Pearson correlation coefficient and the weight value , an optimized differential coefficient is generated;

[0084] ;

[0085] In the formula, is a feature protection factor; the value range is , when is high, decreases, reducing value, weakening the background deduction intensity to protect the characteristic peak; when is low (high background area), increases, enhancing the background deduction intensity, achieving precise control of weak deduction in the key feature area and strong deduction in the high background area.

[0086] S234, based on the differential coefficient , an optimized adaptive differential spectrum is generated;

[0087] ;

[0088] S24, the optimized adaptive differential spectrum is subjected to discrete wavelet transform, decomposed into a plurality of scale coefficients, then a soft threshold denoising is applied on each scale, and then a wavelet inverse transform is used to reconstruct the denoised spectrum, obtaining a multi-scale smoothed reconstructed spectrum ;

[0089] The soft threshold denoising in multiple scales can suppress the high frequency noise and retain the low frequency characteristics, and the reconstructed spectrum has a significantly improved signal-to-noise ratio and clearer details.

[0090] S25, reconstructing the spectrum and the optimized adaptive differential spectrum calculating the residual spectrum If the root mean square of the residual spectrum is higher than a preset threshold, the background coefficient is adjusted. ;

[0091]

[0092] In the formula, is the root mean square of the residual spectrum; is the total number of m / z channels participating in the calculation;

[0093]

[0094] In the formula, is the residual spectrum; if R(m / z) is large, it indicates that the current background removal or denoising process is insufficient, and the parameters need to be adjusted for further optimization.

[0095] By calculating the residual of the original differential spectrum and the denoised spectrum and iteratively optimizing, the system automatically adjusts or the wavelet threshold until the root mean square error (RMSE) converges to a preset threshold, effectively eliminating the residual error caused by improper initial setting of algorithm parameters.

[0096] S26, repeating steps S21 to S25 until the root mean square of the residual spectrum is less than or equal to a preset root mean square threshold, and then obtaining the background-removed spectrum after multiple iterations.

[0097] S27, normalizing the background-removed spectrum after multiple iterations according to the original intensity to generate a background-removed net feature spectrum ;

[0098]

[0099] In the formula, is the maximum value of the feature fingerprint spectrum, ;

[0100] De-normalization: multiply by to obtain the final background-removed net feature spectrum But there is a special case: if the maximum value of the original feature fingerprint spectrum is 0 (i.e. no signal), then the entire spectrum is 0, and after inverse normalization, it is still 0. And the spectrum after inverse normalization may have negative values, which usually indicates that the background subtraction is overdone, but in this embodiment, the negative values are set to zero because the intensity cannot be negative, therefore, after inverse normalization, a post-processing step is set: truncate the negative values to 0.

[0101] Maximum value of the known feature fingerprint spectrum (arbitrary unit), after adaptive difference at a certain m / z point , then the intensity of this point after inverse normalization is If after iterative correction, 1, then after inverse normalization, it is -100, and then it is set to 0 by max(0, -100).

[0102] The multiplication of the maximum value of the feature fingerprint spectrum is because the purpose of this embodiment is to obtain the true signal in the sample after background subtraction, and this signal should have the same intensity dimension as the original sample spectrum, and the scaling of the background spectrum has been considered in the difference process by the background coefficient , and the final is relative to the normalized sample spectrum.

[0103] The background subtraction module 22 adopts a closed-loop adaptive multi-scale background suppression method, which overcomes the errors caused by the common background signal drift, environmental VOC fluctuation, noise interference and algorithm parameter fixation in real-time detection of mass spectrometry, significantly improves the recognition sensitivity and quantitative accuracy of low-concentration target volatile gas components, and meets the needs of dynamic online monitoring in grain storage environment.

[0104] The target VOC determination unit 3 is used to determine the target VOC in the grain pile based on the net feature spectrum through a triple verification mechanism, and output the target VOC name and specific m / z, and at the same time, the target VOC is audited again;

[0105] The target VOC determination unit 3 includes a mass spectrum library updating module 31, and the mass spectrum library updating module 31 includes a grain volatile historical mass spectrum library;

[0106] The mass spectrum library updating module 31 is used to audit the target VOC again, and after the second audit is passed, the target VOC is assigned an initial weight, and the target VOC name, specific m / z, initial weight and time stamp are stored in the grain volatile historical mass spectrum library; if the audit fails, the detection log of the second audit of the target VOC is recorded for subsequent review.

[0107] The secondary review specifically checks whether the VOC name and its characteristic m / z already exist in the historical mass spectrum library. If not, the "review strategy" is invoked: based on historical frequency of occurrence, such as spectrum matching score S or manual expert review as necessary, to determine whether to enter the library. After the review is passed, the initial weight is calculated, which is determined by the matching score and the proportion of historical frequency of occurrence. The target VOC name, characteristic m / z, initial weight and timestamp are written into the historical mass spectrum library. If the review is not passed, only the detection information is recorded in the log for subsequent tracing and expert review.

[0108] Therefore, the food volatile substance historical mass spectrum library contains two parts of content:

[0109] Initial static entries: preloaded standard VOC mass spectrum information (compound name, CAS number, standard reference spectrum, preset characteristic ions and their initial weights, environmental adaptability parameters);

[0110] Dynamic update entries: VOC records (including characteristic m / z, weighting value, verification log, occurrence scene) added after triple verification and secondary review during system operation.

[0111] In the target VOC determination unit 3, the triple verification mechanism includes a first verification mechanism, a second verification mechanism and a third verification mechanism;

[0112] The first verification mechanism is characteristic ion screening verification. The characteristic ion screening verification passes through the construction of a candidate ion library. The candidate ion library passes the uniqueness test, abundance stability test and temperature and humidity adaptability verification. The verified ions are used as the characteristic ions of the final target VOC.

[0113] Specifically: possible target ions m / z are screened from the preset volatile organic compound database and historical mold sample library, which are used as initial members of the candidate ion library. The initial members in the candidate ion library are subjected to uniqueness test, abundance stability test and temperature and humidity adaptability verification. The uniqueness test is to calculate the interference coefficient k between the candidate ion and other ions in the library. If k < 15%, the candidate ion passes the uniqueness test; otherwise, it is rejected. The abundance stability test is to count the ion intensity fluctuation in N repeated scans. If the fluctuation amplitude is < ± 10%, the stability test is passed. The temperature and humidity adaptability verification is repeated testing under the condition of set temperature and humidity to ensure that the signal intensity of the candidate ion is not affected by environmental changes. Finally, all the ions passing the above tests are recorded as the characteristic ions of the target VOC.

[0114] The second verification mechanism is spectrum matching verification. By calling the characteristic ion channels and their weights corresponding to each target compound in the candidate ion library, the weighted cosine similarity algorithm is used to match the ion intensity vector of the net feature spectrum on the channel with the corresponding vector of each candidate compound in the historical mass spectrometry library of grain volatiles. When the similarity score is greater than or equal to the similarity judgment threshold, the compound corresponding to the candidate ion library is confirmed as the target VOC.

[0115] First, extract candidate ion libraries from the net characteristic spectrum. The ionic strength of m / z forms a 3D sample vector;

[0116]

[0117] Similarly, for each candidate compound in the historical mass spectrometry library, its corresponding [data] is also extracted. Each m / z reference abundance forms a reference vector;

[0118]

[0119] In the formula, Indicates the first in the reference library One compound;

[0120] Weights of each channel in the candidate ion library Calculate the weighted cosine similarity between the sample vector and each reference vector;

[0121]

[0122] In the formula, The weighted cosine similarity score is used to measure the degree of similarity between the sample spectrum and the reference spectrum. For the first The weight of each feature ion reflects the importance of the ion in the matching process. The weight values ​​corresponding to the feature m / z channels in the candidate ion library are selected from the feature channel weight factor library. The net characteristic spectrum of the sample in the first... Ion intensity at each characteristic m / z channel; The first in the historical mass spectrometry library of grain volatiles The first compound The relative abundance of reference ions corresponding to each characteristic m / z channel; The number of characteristic ions participating in the matching calculation; ;

[0123] when If the net characteristic spectrum contains the k-th compound, then this compound is taken as the target VOC; where, The similarity threshold;

[0124] The third verification mechanism is interference exclusion verification, which is used for secondary screening of the matching result of the second verification mechanism, when the overlap degree of the target peak of the target VOC and the interference peak of the warehouse interference is greater than the overlap degree threshold, the interference peak is excluded, and the final target VOC is confirmed;

[0125] Even if the spectrum matching score is high, false positives may still be formed by the same m / z value of the interference (such as water vapor, ethanol), leading to distortion of the qualitative or quantitative results, the overlap degree judgment and mathematical fitting are separated, the false signal overlapping with the common interference is effectively excluded, and the final output VOC signal is ensured to be real and reliable, and the recognition accuracy and quantitative accuracy of the whole system are improved.

[0126] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application.

Claims

1. A real-time mass spectrometry analysis system for detection of volatile gas components of grain, characterized by, The application relates to a grain volatile substance detection method and device. The sample spectrum acquisition unit (1) is used for obtaining headspace sample gas from different depths of a grain pile, the headspace sample gas is scanned in parallel in an m / z range, thereby obtaining original sample spectra of different depths, then the original spectra of each depth are weighted according to preset weights, and a characteristic fingerprint spectrum is output; The differential background suppression unit (2) is used for collecting real-time background spectra, and a closed-loop adaptive multi-scale background suppression method is adopted to remove background ion signals of the original sample spectra based on the real-time background spectra and the characteristic fingerprint spectrum, thereby generating a net characteristic spectrum without background. The differential background suppression unit (2) comprises a background deduction module (22), the background deduction module (22) is used for fusing and processing the input characteristic fingerprint spectrum and real-time background spectrum by adopting a closed-loop adaptive multi-scale background suppression method, and outputs a net characteristic spectrum without background. In the background deduction module (22), the specific steps of outputting the net characteristic spectrum without background are as follows: S21, peak alignment and normalization processing are performed on the input characteristic fingerprint spectrum and real-time background spectrum respectively to obtain normalized sample spectrum intensity and normalized background spectrum intensity ; and a dynamic time warping algorithm is used to fine-tune and align the main peak positions in the characteristic fingerprint spectrum and the real-time background spectrum; S22, using the normalized sample spectrum intensity and the normalized background spectrum intensity calculating an initial difference spectrum for each m / z channel and a correlation coefficient of the sample spectrum and the background spectrum determining a background subtraction coefficient based on the correlation coefficient based on the background subtraction coefficient obtaining an adaptive difference spectrum ;​ ； ； ； wherein is the average of the sample spectrum; is the average of the background spectrum; is the baseline similarity threshold; is the adaptive background removal sensitivity factor; Acquiring adaptive differential spectrum : ； S23、in determining the background-removed coefficients In the process, the feature peak-specific weight is introduced to optimize the correlation coefficient to the weighted Pearson correlation coefficient , and further generate the optimized adaptive differential spectrum ; S24, the optimized adaptive differential spectrum Discrete wavelet transform is performed to decompose the spectrum into several scale coefficients, soft threshold denoising is applied on each scale, and inverse wavelet transform is used to reconstruct the denoised spectrum to obtain a multi-scale smoothed reconstructed spectrum ; S25, reconstructing the spectrum with the optimized adaptive differential spectrum calculating the residual spectrum , if the root mean square of the residual spectrum is higher than a preset threshold, adjusting the background coefficient ; S26, repeating steps S21 to S25 until the root mean square of the residual spectrum is less than or equal to a preset root mean square threshold, thereby obtaining the background-removed spectrum after multiple iterations ; S27, background-subtracted spectra after multiple iterations Background-subtracted, net feature spectra generated by inverse normalizing to original intensities ; In S23, the optimized adaptive differential spectrum is generated The specific steps are as follows: S231, assign a weight value to each m / z channel based on the preset feature channel weight factor library ; S232、the correlation coefficient optimized as a weighted Pearson correlation coefficient ; ； S233, based on the weighted Pearson correlation coefficient and weight values generating optimized difference coefficients ; ； In the formula, is a characteristic protection factor; S234、based on the difference coefficient generating an optimized adaptive difference spectrum ; ； In S231, the characteristic channel weight factor library is constructed in the following manner: extracting historical key compound specific m / z channels from a historical mass spectral library of grain volatiles, assigning weights ; the historical mass spectral library of grain volatiles is updated in real time based on a target VOC determination unit (3) assigning weights to m / z channels in the historical background spectrum whose variance of fluctuations exceeds a threshold ; The target VOC determination unit (3) is used for determining a target VOC in the grain pile based on the net characteristic spectrum by adopting a triple verification mechanism, and outputs a target VOC name and a characteristic m / z, and the target VOC is subjected to secondary auditing.

2. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 1, characterized in that: The differential background suppression unit (2) comprises a background spectrum acquisition module (21). The background spectrum acquisition module (21) is provided with a multi-way valve, the multi-way valve can automatically switch background gas and sample gas, the background gas is automatically collected through the multi-way valve after the headspace sample gas is obtained, and the background gas is scanned in parallel in an m / z range, thereby obtaining real-time background spectra.

3. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 2, characterized in that: The target VOC determination unit (3) comprises a mass spectrum library updating module (31), and the mass spectrum library updating module (31) comprises a grain volatile substance historical mass spectrum library. The mass spectrum library updating module (31) is used for performing secondary auditing on the target VOC, and after the secondary auditing is passed, an initial weight is assigned to the target VOC, and the target VOC name, the characteristic m / z, the initial weight and a time stamp are stored in the grain volatile substance historical mass spectrum library. If the auditing is not passed, a detection log of the secondary auditing of the target VOC is recorded for subsequent review.

4. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 3, characterized in that: In the target VOC determination unit (3), the triple verification mechanism comprises a first verification mechanism, a second verification mechanism and a third verification mechanism; wherein the first verification mechanism is characteristic ion screening verification, the second verification mechanism is spectrum matching verification, and the third verification mechanism is interference exclusion verification.

5. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 4, characterized in that: The characteristic ion screening verification passes through a candidate ion library, the candidate ion library passes through uniqueness inspection, abundance stability test and temperature and humidity adaptability verification, and the verified ions are used as characteristic ions for finally representing the target VOC.

6. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 5, characterized in that: The spectrum matching verification is performed by calling the characteristic ion channel and its weight corresponding to each target compound in the candidate ion library, using a weighted cosine similarity algorithm to match the ion intensity vector of the net characteristic spectrum on the channel with the corresponding vector of each candidate compound in the historical mass spectrum library of grain volatiles, and when the similarity score is greater than or equal to the similarity determination threshold, the compound corresponding to the candidate ion library is confirmed as the target VOC.

7. The real-time mass spectrometer system for detecting volatile gas components of grain according to claim 6, characterized in that: The interference exclusion verification is used for secondary screening of the matching results of the second verification mechanism, and when the overlap degree of the target peak of the target VOC and the interference peak of the warehouse interference is greater than the overlap degree threshold, the interference peak is excluded, and the final target VOC is confirmed.

Images