Methods for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional spectra and their applications in complex sample analysis

Abstract

The present invention discloses a method for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional spectra and its application in complex sample analysis. By exchanging two one-dimensional spectral sequences in a plurality of one-dimensional spectra, a one-dimensional spectrum of an optimized sequence is obtained, thereby constructing a two-dimensional or high-dimensional asynchronous spectrum. By changing the one-dimensional spectral sequence, the present invention significantly enhances the cross-peak intensity of the two-dimensional or high-dimensional asynchronous spectrum by more than 100 times, while the standard deviation of the corresponding noise intensity hardly changes. Even in binary and ternary mixed materials with highly overlapping elution curves of mixture components and severe noise interference, according to the method for strengthening the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectrum described in the present invention, the best one-dimensional spectral sequence can still be quickly obtained, the signal-to-noise ratio of the asynchronous spectrum is strongly enhanced, and the spectra of pure substances of each component are successfully separated, providing a new way for two-dimensional and high-dimensional correlation spectral analysis methods in complex sample analysis.

Description

Technical Field

[0001] The invention belongs to the field of infrared spectrum analysis, and in particular relates to a method for enhancing the signal-to-noise ratio of a two-dimensional / high-dimensional asynchronous spectrum and an application thereof in complex sample analysis. Background Art

[0002] Complex sample analysis has been widely used in materials science, especially in the field of commodity production. Many industrial development and research systems at home and abroad use analytical technology to track the latest research results and development trends in the industry, obtain the latest advanced technical information, and grasp accurate and timely technology and market information. So in a sense, analytical work is a scout for my country's materials industry and the expansion of materials applications.

[0003] When analyzing material samples, the analysis object is often a complex mixture sample. Usually, chromatography-spectroscopy and other experimental methods are used for analysis. Using two-dimensional / high-dimensional asynchronous correlation spectra to process the data matrix generated by chromatography-spectroscopy and other experimental methods can still obtain the spectra of pure substances of different components when the different components are not separated.

[0004] However, when analyzing a data matrix containing a mixture of multiple components, the interference of noise in two-dimensional or high-dimensional spectra will be greatly aggravated. In order to improve the quality of two-dimensional or high-dimensional correlation spectra, people have used a variety of methods. In general, the signal-to-noise of two-dimensional spectra can be improved from two dimensions, namely by reducing noise and enhancing signals.

[0005] On the one hand, the Savitzky-Golay algorithm and similar methods were first chosen to smooth the original spectra. Wu Yuqing et al. proposed an orthogonal orthogonal signal correlation method. Jung's team developed eigenvector reconstruction and smoothing factor analysis based on the combination of smoothing and eigenvector reconstruction. Various filtering methods have also been used to reduce the noise of two-dimensional spectra. For example, Ozaki and Berry have demonstrated that the wavelet filter method shows excellent performance in noise reduction. He Anqi used Butterworth filters to suppress noise in two-dimensional spectra.

[0006] In the noise reduction methods, an inevitable problem is signal distortion, which may mislead the analysis of 2D spectra, which is why it is emphasized to maintain a balance between reducing noise and suppressing signal distortion.

[0007] On the other hand, the enhancement of the signal part can be achieved by modifying the method of generating the two-dimensional spectrum. Li Xiaopei and colleagues showed that when the modified one-dimensional spectrum was used to construct the two-dimensional asynchronous spectrum, the intensity of the cross peak increased by about 100 times; and exchanging the order of the one-dimensional spectra to construct the two-dimensional correlation spectrum can also enhance the intensity of the cross peak. He Anqi et al. showed that the construction of the two-dimensional spectrum without subtracting the reference spectrum can also effectively improve the signal-to-noise ratio of the two-dimensional asynchronous spectrum, and successfully applied it to the study of hydrogen bond reorganization in a mixture of methanol / ether / THF. Recently, this method has been used in the data matrix analysis of chromatographic spectra and similar experiments. The advantage of the signal enhancement method is that there is no signal distortion in the generated two-dimensional asynchronous spectrum.

[0008] Although the above-mentioned studies can improve the signal-to-noise ratio of two-dimensional spectra to a certain extent, they still cannot reach the ideal state. Moreover, the operation is not only time-consuming and labor-intensive, but also not suitable for all application scenarios. In particular, when analyzing the composition of samples containing multiple components, there is a serious overlap in the elution curves, and the spectrum of the pure components in the complex sample cannot be obtained.

[0009] Due to the above reasons, there is an urgent need to further improve the research methods of two-dimensional or high-dimensional correlation spectra to avoid finding difficult and complex separation conditions for multiple sample components and advanced separation instrumentation required for multidimensional chromatography technology, so as to obtain spectra of pure components from complex samples conveniently and quickly. Summary of the invention

[0010] In order to overcome the above problems, the inventors have made a keen study on the method of enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectroscopy, and have studied the method of enhancing the signal-to-noise ratio of two-dimensional / high-dimensional spectroscopy and its application in complex sample analysis. By exchanging two one-dimensional spectral sequences in multiple one-dimensional spectra, a one-dimensional spectrum of the optimized sequence is obtained, thereby constructing a two-dimensional or high-dimensional asynchronous spectrum. The present invention changes the one-dimensional spectral sequence, and while the cross-peak intensity of the two-dimensional or high-dimensional asynchronous spectrum is significantly enhanced by more than 100 times, the standard deviation of the corresponding noise intensity hardly changes. Even in binary and ternary mixed materials with highly overlapping effluent curves of the mixture components and severe noise interference, according to the method of the strong two-dimensional / high-dimensional asynchronous spectroscopy signal-to-noise ratio of the present invention, the best one-dimensional spectral sequence can still be quickly obtained, the signal-to-noise ratio of the asynchronous spectrum is strongly enhanced, and the spectra of the pure substances of each component are successfully separated, which provides a new way for two-dimensional and high-dimensional correlation spectral analysis methods in complex sample analysis, thereby completing the present invention.

[0011] Specifically, the purpose of the present invention is to provide the following aspects:

[0012] On the one hand, a method for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectra is provided, characterized in that the method comprises: cutting the line at the systematic missing peak of the two-dimensional or high-dimensional asynchronous spectrum constructed by changing the one-dimensional spectral sequence to obtain the spectrum of the pure substance.

[0013] In another aspect, there is provided an application of the method according to the first aspect for obtaining a spectrum of a pure substance in a complex sample.

[0014] The beneficial effects of the present invention include:

[0015] (1) The method provided by the present invention for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectra can effectively obtain the cross-peak intensity of substantial two-dimensional or high-dimensional asynchronous spectra, improve the signal-to-noise ratio of complex sample systems, and improve the efficiency of obtaining pure substance spectra from complex systems.

[0016] (2) The present invention provides a method for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectra. Even in binary and ternary mixed materials where the elution curves of the mixture components are highly overlapped and the noise interference is serious, the optimal one-dimensional spectral sequence can still be obtained with high efficiency, the signal-to-noise ratio of the asynchronous spectrum can be strongly enhanced, and the spectra of the pure substances of each component can be successfully separated.

[0017] (3) The method provided by the present invention for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectroscopy is simple and efficient, and provides a new idea for obtaining spectra of pure substances from complex samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a sequence for obtaining an optimized one-dimensional spectrum according to a preferred embodiment of the present invention is shown;

[0019] Figure 2 -A shows the data matrix constructed from 100 FTIR spectra in Example 1;

[0020] Figure 2 -B shows the two-dimensional asynchronous spectrum diagram in Example 1;

[0021] Figure 2 -C shows the FTIR spectra of pure water and isopropanol in Example 1;

[0022] Figure 3 The two-dimensional asynchronous spectrum diagram constructed by artificially adding noise in a simulated harsh environment in Example 1 is shown;

[0023] Figure 4 -A shows a two-dimensional asynchronous spectrum constructed from the best one-dimensional spectrum sequence after adding noise in Example 1;

[0024] Figure 4 -B shows the FTIR spectra of pure substances of water and isopropanol obtained by cutting in the two-dimensional asynchronous spectrum constructed by the best one-dimensional spectrum sequence after adding noise in Example 1;

[0025] Figure 5 The data matrix constructed from 140 FTIR spectra in Example 2 is shown;

[0026] Figure 6 -A shows a typical two-dimensional asynchronous spectrum diagram of Example 2;

[0027] Figure 6 -B and Figure 6 -C respectively show the two-dimensional cross-section Ξ(x, 3902, z) and Ξ(x, 2618, z) diagrams of the three-dimensional asynchronous spectrum in Example 2;

[0028] Figure 6 -D Trace 1, Trace 2 and Trace 3 respectively show the cross-sections Ξ(x,3902,1744), Ξ(x,2618,3902) and Ξ(x,2618,2918) of the three-dimensional asynchronous spectrum in Example 2;

[0029] Figure 6 -D Trace 4, Trace 5 and Trace 6 show the FTIR spectra of the pure substances acetonitrile vapor, butanone vapor and water vapor in Example 2, respectively;

[0030] Figure 7 The two-dimensional cross-sectional diagram of the three-dimensional asynchronous spectrum constructed by artificially adding noise in a simulated harsh environment in Example 2 is shown;

[0031] Figure 8 -A and 8-B show the two-dimensional cross-sectional diagram of the three-dimensional asynchronous spectrum constructed from the best one-dimensional spectrum sequence after adding noise in Example 2;

[0032] Figure 8 - In C, Trace 1, Trace 2 and Trace 4 respectively show the cross-sections of the three-dimensional asynchronous spectrum Ξ(x,3902,1744), Ξ(x,2618,3902) and Ξ(x,2618,2918);

[0033] Fig. 9 -A shows the pure substance FTIR spectra of simulated substances P, Q, and R of Experimental Example 1;

[0034] Fig. 9 -B shows the elution curve of the mixture of P, Q, and R in Experimental Example 1;

[0035] Fig.10 The chromatographic-spectral experimental result diagram of the simulated P, Q, and R mixture of Experimental Example 1 is shown;

[0036] Fig.11 A typical two-dimensional asynchronous spectrum of Experimental Example 1 is shown;

[0037] Fig.12 -A and 12-B respectively show the two-dimensional cross-sectional views of the three-dimensional asynchronous spectrum of Experimental Example 1;

[0038] Fig.12 -C shows the FTIR spectra of pure substances P, Q, and R;

[0039] Fig.13 A two-dimensional cross-sectional diagram of a three-dimensional asynchronous spectrum constructed from the best one-dimensional spectrum sequence after adding noise in Experimental Example 1 is shown;

[0040] Fig.14 The absolute intensity comparison diagram of the cross peaks in Experimental Example 1 is shown. DETAILED DESCRIPTION

[0041] The present invention will be further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present invention will become more clear and distinct.

[0042] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0043] In the process of analyzing the components of unknown materials, the selection of separation conditions requires a lot of time and effort to find options. Especially when the sample amount is very small, the error tolerance of the analysis test is reduced. The two-dimensional correlation spectrum method can obtain the corresponding spectrum of the component through the systematic missing peaks in the n-dimensional spectrum when the unknown material is not separated. However, there is always a large amount of unavoidable interference noise in the original one-dimensional spectrum used to generate two-dimensional or high-dimensional asynchronous spectra. When generating two-dimensional asynchronous spectra, the noise will be significantly amplified. And when constructing three-dimensional or four-dimensional asynchronous spectra, the impact of the noise problem will become more and more serious. In many cases, the interference of the noise is so severe that the systematic missing peaks are difficult to identify, which makes it very difficult to extract the spectrum of the pure component, which greatly limits the application of two-dimensional correlation spectrum in complex sample materials.

[0044] When constructing a 2D asynchronous spectrum from n 1D spectra arranged in a given order, the cross-peak intensity can be enhanced by selecting the optimal sequence of the n 1D spectra. A simple approach is to evaluate each possible order in order to obtain the optimal arrangement of n spectra that produces the most substantial cross-peaks. However, this approach is not feasible when n is not a small number. The number of possible sequences of n 1D spectra is n factorial (n!). As n increases, the value of n! explodes and quickly becomes an astronomical number. In a typical hyphenated experiment of a mixed sample, dozens to more than a hundred 1D spectra are organized into a data matrix. The corresponding possible sequences are greater than 10 100Therefore, even with the most advanced computer technology, it is impossible to check every possible sequence. Therefore, random search is the only available option.

[0045] Among the traditional methods, the most commonly used is the random exchange method, which randomly selects r (where r≤n) one-dimensional spectra from a given sequence of n one-dimensional spectra. Then, the positions of the r one-dimensional spectra are randomly changed, and then a two-dimensional or high-dimensional asynchronous spectrum is constructed to extract the spectrum of the pure substance. This method is time-consuming and laborious because the probability of successfully increasing the absolute value of the intensity of the cross-peak at the specified position of the two-dimensional or high-dimensional asynchronous spectrum is low. The noise level in the obtained two-dimensional or high-dimensional asynchronous spectrum is roughly the same as the noise level in the two-dimensional or high-dimensional asynchronous spectrum obtained by evaluating each possible sequence in other traditional methods, and will not be significantly improved.

[0046] The inventors have found that by swapping the positions of two random one-dimensional spectra in a given sequence of n one-dimensional spectra, the cross-peak intensity of substantial two-dimensional or high-dimensional asynchronous spectra can be effectively obtained, the signal-to-noise ratio of complex sample systems can be improved, and the efficiency of obtaining pure substance spectra from complex systems can be improved.

[0047] The present invention is described in detail below.

[0048] On the one hand, a method for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectrum is provided, the method comprising: changing a one-dimensional spectral sequence to construct a two-dimensional or high-dimensional asynchronous spectrum, cutting the line at the systematic missing peak of the obtained two-dimensional or high-dimensional asynchronous spectrum, and obtaining the spectrum of a pure substance.

[0049] According to the present invention, the changing one-dimensional spectrum sequence is to select the best sequence of one-dimensional spectrum sequence from the one-dimensional spectrum of a given sequence, thereby constructing a two-dimensional or high-dimensional asynchronous spectrum.

[0050] In a preferred embodiment, the one-dimensional spectrum sequence of the optimal sequence is obtained by random permutation:

[0051] In step 1), a given sequence of n (n ≥ 11) one-dimensional spectra is called a specified sequence (seq0), and a two-dimensional or high-dimensional asynchronous spectrum is constructed using the n one-dimensional spectra organized in the specified sequence. The absolute intensity of the target cross peak of the obtained two-dimensional or high-dimensional asynchronous spectrum is called the specified intensity (|sig0|). The upper limit of unsuccessful exchange (expressed as NU value) is set to the specified value, and the number of exchanges (expressed as NT) is set to zero.

[0052] Step 2), randomly swap the positions of the two one-dimensional spectra in step 1) to generate a new sequence (seq1) of n one-dimensional spectra, and use the n one-dimensional spectra in the new sequence to construct a new two-dimensional or high-dimensional asynchronous spectrum to obtain the absolute intensity (sig1) of the target cross peak of the two-dimensional or high-dimensional asynchronous spectrum.

[0053] Step 3), if the absolute value of |sig1| is greater than |sig0|, it is said that a successful exchange has been achieved, then the value of |sig0| is set to |sig1|, and seq0 is replaced by seq1, and the value of NT is set to zero; if |sig1| is not greater than |sig0|, it is said that an unsuccessful exchange has been encountered, and then the value of NT is increased by 1.

[0054] Step 4), if the NT value is lower than the NU value, go to step 2), if the NT value is equal to NU, the optimized sequence is set to seq0, the optimized absolute intensity of the target cross peak is set to |sig0|, and the optimization process is terminated. The specific steps are as follows Figure 1 shown.

[0055] Optionally, the method for obtaining the one-dimensional spectrum sequence of the best sequence is a random process, and the optimization results of the same initial sequence may be different. In the above process, when a local maximum is encountered, the above steps are restarted from the same initial sequence. If the best sequence is obtained, the optimization is stopped; otherwise, the above steps are restarted from the same initial sequence.

[0056] Furthermore, 11 to 300, preferably 11 to 200 one-dimensional spectra are selected from n one-dimensional spectra of a given sequence to construct a two-dimensional or high-dimensional asynchronous spectrum, and then the positions of two one-dimensional spectra are randomly exchanged in the selected one-dimensional spectrum, and the one-dimensional spectrum of the best sequence is selected by the absolute intensity of the target cross-peak, thereby effectively obtaining a substantial two-dimensional or high-dimensional asynchronous spectrum and a target cross-peak intensity.

[0057] According to the present invention, the data matrix is ​​constructed by vectors of signal intensities of 11 to 300 one-dimensional spectra at different wavelengths. Preferably, the data matrix is ​​constructed by vectors of signal intensities of 11 to 200 one-dimensional spectra at different wavelengths.

[0058] According to the present invention, 11 to 300 one-dimensional spectra, preferably 11 to 200 one-dimensional spectra are selected from the one-dimensional spectrum of the binary mixture to construct a two-dimensional asynchronous spectrum, that is, when the binary mixture is detected by a thermogravimetric-infrared coupling instrument, 11 to 300, preferably 11 to 200 one-dimensional infrared spectra of the binary gaseous mixture at different time points are recorded, and the vectors of the signal intensities of the recorded one-dimensional infrared spectra at different wavelengths are recorded as the constructed data matrix A, and the two-dimensional asynchronous spectrum is generated by formula (1):

[0059] Ψ(x,y)=A T (x)NA(y) Formula (1);

[0060] In formula (1): Ψ(x, y) represents the intensity at the midpoint (x, y) of the two-dimensional asynchronous spectrum;

[0061] T Represents the transpose of a matrix;

[0062] N is the n-order Hilbert-Noda matrix, which is specifically expressed as:

[0063]

[0064] According to the present invention, 11 to 300 one-dimensional spectra, preferably 11 to 200 one-dimensional spectra are selected from the one-dimensional spectrum of the ternary mixture to construct a three-dimensional asynchronous spectrum, that is, when the ternary mixture is detected using a thermogravimetric-infrared coupling instrument, 11 to 300, preferably 11 to 200 one-dimensional spectra of the ternary gaseous mixture at different time points are recorded, the recorded one-dimensional infrared spectra are divided into m groups, each group contains p one-dimensional spectra, and the vectors of the signal intensities of the p one-dimensional spectra at different wavelengths are used to construct a data matrix A i , a two-dimensional asynchronous spectrum is constructed by formula (2), then m sets of FTIR spectra can construct m two-dimensional asynchronous spectra (Ψ i (x, y), where i = {1, 2, …, m}).

[0065] Wherein, the formula (2) is as follows:

[0066] Ψ i (x,y)=A i T (x)N p A i (y) Formula (2);

[0067] In formula (2), Ψ(x, y) is a two-dimensional asynchronous spectrum;

[0068] T Represents the transpose of a matrix;

[0069] N is the p-order Hilbert-Noda matrix, which is specifically expressed as:

[0070]

[0071] Take a line segment at y=y0 from each two-dimensional asynchronous spectrum, so that a total of m lines segment are obtained, and the m lines segment are assembled into a data matrix B y0 , a two-dimensional asynchronous spectrum is constructed by formula (3). This two-dimensional asynchronous spectrum is the two-dimensional cross-section of the three-dimensional asynchronous spectrum at wavelength y=y0, denoted as Ξ(x, y0, z).

[0072] Ξ(x,y0,z)=(B y0 (x) T N m B y0 (z) Formula (3);

[0073] In formula (3), Ξ(x, y0, z) is the two-dimensional cross section of the three-dimensional asynchronous correlation spectrum at y = y0;

[0074] T Represents the transpose of a matrix;

[0075] N is the m-order Hilbert-Noda matrix, which is specifically expressed as:

[0076]

[0077] When y0 traverses all wavelengths, a series of two-dimensional cross sections of the three-dimensional asynchronous spectrum are obtained. By assembling these cross sections along the Z axis, a three-dimensional asynchronous spectrum is obtained.

[0078] According to the present invention, the intercept taken on the systematic missing peak of the two-dimensional or high-dimensional asynchronous spectrum is a linear combination of the spectral signals of the components of the mixture sample.

[0079] Furthermore, in a two-dimensional cross section of a two-dimensional or high-dimensional asynchronous spectrum, when a certain point (corresponding wavelength is x1, x2, ... x n ) is 0, but in constructing a two-dimensional or high-dimensional asynchronous spectrum, the one-dimensional wavelength is x1, x2, …x n When the signal strength at is not 0, this is the system missing peak

[0080] For example, suppose a binary mixture contains substance U and substance V, substance U has a wavelength x in the one-dimensional spectrum. U1 , x U2 , …, x Uk There is an independent peak at x, that is, substance U has an independent peak at wavelength x. Ui ,i∈{1,2,…,k} is not 0, but the spectral signal intensity of substance V is 0. Then, in the corresponding two-dimensional asynchronous spectrum, at (x Ui ,x Uj ), where i, j∈{1,2,…,k} will have a systematic missing peak; similarly, the wavelength x of the one-dimensional spectrum of substance V V1 , x V2 ,..,x vl There is an independent peak at x, that is, substance V has an independent peak at a wavelength of x. Vj ,j∈{1,2,…,l} is not 0, but the spectral signal intensity of substance U is 0. Then, in the corresponding two-dimensional asynchronous spectrum, at (x Vi ,x Vj ), where i, j∈{1,2,…,l} will have a system missing peak.

[0081] For example, a two-dimensional asynchronous spectrum is constructed using a series of spectra containing binary mixtures of U and V. Let x U1 , xU2,.., x Uk is the peak position of the independent peak in the one-dimensional infrared spectrum of substance U, then the pure substance spectrum of substance V can be obtained by y=x in the two-dimensional asynchronous spectrum. Ui (where i∈{1,2,…,k}) is obtained by taking the intercept line. Similarly, let the wavelength x of the material V in the one-dimensional spectrum be V1， x V2,.., x vl If there is an independent peak at , the spectrum of substance U can be obtained by y = x in the two-dimensional asynchronous spectrum. Vi (where i∈{1,2,…,l}) is obtained by intercepting the line.

[0082] The three-dimensional asynchronous spectrum constructed by a series of one-dimensional spectra of the ternary mixture of substance J, substance K and substance L is K On the two-dimensional cross section (Ξ(x,y K ,z)), such as at wavelength x J ,z L At, that is, Ξ(x J ,y K ,z L ) intensity is 0, and in the series of one-dimensional spectra for constructing three-dimensional asynchronous spectra, x J ,y K ,z L If the spectral signal intensity at is not 0, then the three-dimensional asynchronous spectrum is called K On the two-dimensional cross section, (x J ,y K ,z L ) has a system missing peak.

[0083] Furthermore, at wavelength y K is the independent peak of substance K, then in the three-dimensional asynchronous spectrum at y=y K There is no spectral signal contribution of substance K on the two-dimensional cross section. J If the peak at Ξ(x,y K ,z) obtained z=x J There is no spectral signal contribution of J on the cross section, so on Ξ(x,y K ,z) obtained z=x J The intercept of is the pure material spectrum of substance L; similarly, if the wavelength x L is an independent peak of substance L, then at Ξ(x,y K ,z) obtained z=x L There is no spectral signal contribution of substance L on the intersection of Ξ(x,y K ,z) obtained z=x J The intercept of is the pure material spectrum of substance J.

[0084] In another aspect, there is provided application of the method according to the first aspect for acquiring a spectrum of a pure substance in a complex sample.

[0085] Example

[0086] The present invention is further described below by specific examples. However, these examples are merely exemplary and do not constitute any limitation to the protection scope of the present invention.

[0087] The spectra in the following examples were collected on a TG-IR spectrometer (model: TGA8000-Frontier, manufacturer: Perkin-Elmer).

[0088] Example 1

[0089] (1) 15.6 mL of pure water and 15.6 mL of isopropanol (analytical grade, produced by Beijing Tongguang Fine Chemical Co., Ltd.) were mixed to obtain a water / isopropanol mixture, which was loaded onto a TG-IR spectrometer and heated from 30°C to 150°C at a heating rate of 30°C / min. The sample was then kept at 150°C, and an FTIR spectrum of the effluent was recorded every 6 seconds.

[0090] 100 FTIR spectra corresponding to 100 outflow time points are selected from the recorded spectra. The vectors formed by the signal intensities of the 100 FTIR spectra at different wavelengths are the constructed data matrix, denoted as A. Figure 2 -A shows the data matrix constructed from 100 FTIR spectra. According to formula (1), a two-dimensional asynchronous spectrum is constructed, and the obtained two-dimensional asynchronous spectrum is as follows Figure 2 -B, it can be seen from the figure that there are a lot of systematic missing peaks in the two-dimensional asynchronous spectrum, for example, at 3800cm -1 Up to 4000cm -1 There are at least 20 independent peaks related to gaseous water in the spectral band, resulting in more than 400 systematic missing peaks in the two-dimensional asynchronous spectrum.

[0091] exist Figure 2 -B, 8 system missing peaks are selected and divided into two groups: the first group of system missing peaks appear at (3903, 1717), (3903, 3903), (1717, 3903) and (1717, 1717), forming a 2×2 square matrix; the second group of system missing peaks appear at (2978, 951), (2978, 2978), (951, 2978) and (951, 951), forming another 2×2 square matrix.

[0092] From the two-dimensional asynchronous spectrum y = 3903 cm -1 and y = 2978 cm -1Take a horizontal section to obtain the FTIR spectra of pure water and isopropanol. The results are as follows: Figure 2 - As shown in Trace4 and Trace1 in C. The shape of the obtained intercept is the same as the infrared spectrum shape of pure substances of gaseous water and gaseous isopropanol; Figure 2 -C also shows the FTIR spectra of pure substances gaseous isopropanol (Trace2) and gaseous water (Trace5). It can be seen that the shapes of Trace1 and Trace2 are basically the same, and the shapes of Trace4 and Trace5 are basically the same.

[0093] (2) To further verify whether the random permutation method can effectively improve the signal-to-noise ratio of two-dimensional asynchronous spectroscopy, the data matrix in step (1) is processed as follows:

[0094] The order of the 100 FTIR spectra was disrupted, and noise was introduced into each data point of each FTIR spectrum. The fluctuation amplitude of the noise was 1% of the maximum signal intensity in the 100 infrared spectra. A two-dimensional asynchronous spectrum was constructed in a similar manner to step (1), as shown in FIG. Figure 3 As shown in the figure, it can be seen that the noise interference is so severe that the cross peaks and system missing peaks cannot be identified.

[0095] Two of the 100 shuffled FTIR spectra (denoted as designated sequence seq0) are randomly swapped to construct 1000 random sequences, thereby obtaining 1000 data matrices. Each of these 1000 data matrices is used as Figure 1 Enter the data shown, press Figure 1 The method shown in FIG. 1 was used to optimize the two-dimensional asynchronous spectrum, and NU was set to 200. The absolute intensity of the cross peak at (3903, 2978) of each constructed two-dimensional asynchronous spectrum was observed. The final optimized two-dimensional asynchronous spectrum is shown in FIG. Figure 4 -A, it is obvious that the signal-to-noise ratio of the two-dimensional asynchronous spectrum is greatly improved due to the obvious enhancement of the intensity of the cross peak. Figure 4 In -A, two groups of missing peaks were identified: (3903, 1717), (3903, 3903), (1717, 3903) and (1717, 1717) as well as (2978, 951), (2978, 2978), (951, 2978) and (951, 951), at y = 3903 cm -1 and y = 2978 cm -1 Take a vertical section to obtain the FTIR spectra of pure water and isopropanol. The results are as follows: Figure 4 -B.

[0096] As can be seen from the above, even in the most complex and noisy conditions, it is still possible to obtain the FTIR spectra of pure substances from binary mixtures. The reproduced spectra of water and isopropanol ( Figure 4-B) Shape and Figure 2 - Much the same as shown in C.

[0097] Example 2

[0098] (1) 30 mL of pure water, 30 mL of pure acetonitrile (analytical grade, produced by Tianjin Kangkede Technology Co., Ltd.) and 40 mL of pure butanone (analytical grade, produced by Sinopharm Chemical Reagent Co., Ltd.) were mixed to obtain a ternary mixture, which was loaded onto a TG-IR spectrometer and heated from 30°C to 150°C at a heating rate of 30°C / min. The sample was then kept at 150°C and an FTIR spectrum of the gaseous escaping product was recorded every 6 seconds.

[0099] 140 FTIR spectra corresponding to 140 outflow times are randomly selected from the recorded spectra. The vectors formed by the signal intensities of the 100 FTIR spectra at different wavelengths are the constructed data matrix, denoted as B. Figure 5 A data matrix constructed from 140 FTIR spectra is shown.

[0100] The 140 FTIR spectra are divided into 7 groups, each with 20 FTIR spectra. A data matrix is ​​constructed with each group of FTIR spectra. A two-dimensional asynchronous spectrum is constructed using formula (2). Thus, 7 groups of FTIR spectra can construct 7 two-dimensional asynchronous spectra. Figure 6 -A shows a typical two-dimensional asynchronous spectrum, where it can be observed that: Segment and A large number of systematic missing peaks were observed in the section, among which, The system peak of the section is missing and The system missing peaks are 2 different groups of system missing peaks.

[0101] In each of the seven constructed two-dimensional asynchronous spectra, y = 3902 cm -1 Take a line segment at , thus obtaining 7 lines segment, construct the vectors of the signal strength of these 7 lines at different wavelengths into a data matrix, then construct a two-dimensional asynchronous spectrum according to formula (3), this two-dimensional asynchronous spectrum is the three-dimensional asynchronous spectrum at y = 3902cm -1 The two-dimensional cross section at ((x,3902,z)) is as follows: Figure 6 As shown in Figure 2-B, it can be seen that the system lacks a peak at (1744, 1744).

[0102] Take z=1744cm from Ξ(x,3902,z) -1 The spectrum of gaseous acetonitrile is obtained by cutting the line, such as Figure 6 -D in Trace 1, Figure 6-D Trace 4 shows the FTIR spectrum of pure acetonitrile vapor. It can be seen that the shape of Trace 1 is basically consistent with Trace 4.

[0103] Similarly, in each of the seven two-dimensional asynchronous spectra constructed, y = 2618 cm -1 Take the cutoff line at the position, construct the 7 cutoff lines into a data matrix, and then construct a two-dimensional asynchronous spectrum, such as Figure 6 The obtained two-dimensional asynchronous spectrum is shown in Figure 3. The three-dimensional asynchronous spectrum (Ξ(x, y, z)) at y = 2618 cm -1 The two-dimensional cross section at (denoted as Ξ(x,2618,z)).

[0104] exist Figure 6 -C, The systematic peak loss of the segment still exists. In addition, it can be observed that The system peak of the section is missing due to No system missing peak, in The system peak of the section is missing and The missing peaks in the segment do not belong to the same group. Therefore, z = 3902 cm -1 and z = 2918 cm -1 The FTIR spectra of gaseous butanone and gaseous water were obtained by using the intercept line at Figure 6 -D is shown in Trace2 and Trace3. Figure 6 -D, Trace 5 and Trace 6 show the FTIR spectra of pure butanone vapor and water vapor, respectively. It can be seen that the shape of Trace 2 is basically consistent with Trace 5, and the shape of Trace 3 is basically consistent with Trace 6.

[0105] (2) In order to further verify whether the random permutation method can effectively improve the signal-to-noise ratio of the three-dimensional asynchronous spectrum, the above data matrix is ​​processed as follows:

[0106] The order of the 140 FTIR spectra was disrupted, and noise was introduced into each data point of each FTIR spectrum. The fluctuation amplitude of the noise was 1% of the maximum signal intensity in the 140 infrared spectra. A two-dimensional cross section of the three-dimensional asynchronous spectrum was constructed in a similar manner to step (1). Figure 7 -A shows the three-dimensional asynchronous spectrum at y = 3902 cm -1 The two-dimensional cross section of Figure 7 -B shows the three-dimensional asynchronous spectrum at y = 2618 cm -1 From the two-dimensional cross section at , it can be seen that neither cross peaks nor systematic missing peaks can be identified in the two-dimensional cross section of the three-dimensional asynchronous spectrum.

[0107] The 140 shuffled FTIR spectra (denoted as designated sequence seq0) were divided into seven groups, each with 20 FTIR spectra, and seven two-dimensional asynchronous spectra were constructed in a similar manner to step (1). In each two-dimensional asynchronous spectrum, y = 3902 cm -1 The seven cross-lines were taken at 1744, 3902, 1058 to construct a two-dimensional asynchronous spectrum and measure the absolute intensity of the cross-peaks at (1744, 3902, 1058).

[0108] Afterwards, two of the 140 shuffled FTIR spectra were randomly swapped to construct a total of 100 random sequences, thereby obtaining 100 data matrices. Each of these 100 data matrices was used as Figure 1 Enter the data shown, press Figure 1 The method shown was optimized, NU was set to 1000, and the absolute intensity of the cross peaks at (1744, 3902, 1058) was measured.

[0109] Measure at y = 2618 cm in a manner similar to step (2) above -1 and z = 1174 cm -1 The cross peak at 160° is removed to improve the signal-to-noise ratio of Ξ(x,2618,z).

[0110] The final optimized three-dimensional asynchronous spectrum is at y = 3902 cm -1 The two-dimensional cross section at Figure 8 -A, at y = 2618cm -1 The two-dimensional cross section at Figure 8 -B.

[0111] Figure 8 -Trace1 in C shows (x,3902,z) at z=1744cm -1 The horizontal section taken at is the FTIR spectrum of gaseous acetonitrile. Figure 8 -Trace2 in C shows (x,2618,z) at z=3902cm -1 The horizontal section taken at is the FTIR spectrum of gaseous butanone. Figure 8 -Trace3 in C shows that (x, 2618, z) is at z = 2918 cm -1 The horizontal section taken at is the spectrum of gaseous water.

[0112] As can be seen from the above, the random permutation method can significantly enhance the signal-to-noise ratio of the three-dimensional asynchronous spectrum on the two two-dimensional cross sections. Since the intensity of the cross peak is significantly enhanced, the signal-to-noise ratio of the two-dimensional asynchronous spectrum is greatly improved. Figure 6 -D is roughly the same as shown.

[0113] Experimental example

[0114] Experimental Example 1

[0115] The FTIR spectra of pure substances and the elution curves of mixtures of three components P, Q and R were simulated by chromatography-spectroscopy method, where P, Q and R were three simulated substances, and the FTIR spectra of pure substances of P, Q and R were obtained as follows: Fig. 9 -A, the elution curve of the P, Q, R mixture is as follows Fig. 9 As shown in Figure 1-B, due to the serious overlap of the elution curves of P, Q, and R, the chromatography-spectrometry method can only obtain the spectra of the mixtures of P, Q, and R in different proportions, but not the spectra of the pure substances of P, Q, and R. The experimental results of the mixture of three components P, Q, and R obtained by the chromatography-spectrometry method are shown in Figure 1-B. Fig.10 shown.

[0116] Use the following steps to obtain the pure material spectra of P, Q, and R:

[0117] Step 1: In the chromatography-spectroscopy combined experiment, a total of 140 one-dimensional spectra are generated. The 140 one-dimensional spectra are divided into 7 groups, each with 20 spectra. A two-dimensional asynchronous spectrum is constructed using the 20 one-dimensional spectra in each group in a manner similar to step (1) of Example 2, and a total of 7 two-dimensional asynchronous spectra are constructed. Fig.11 A typical two-dimensional asynchronous spectrum is shown. It can be seen that two systematic missing peaks appear at (50, 50) and (120, 120). The two systematic missing peaks do not belong to the same group, otherwise there will also be systematic missing peaks at (50, 120) and (120, 50).

[0118] Step 2: From 7 two-dimensional asynchronous spectra at y = 120 cm -1 Take horizontal sections at , and get 7 sections in total. The vectors of the signal intensities of these 7 sections at different wavelengths are constructed into a data matrix. Then, according to formula (3), a two-dimensional asynchronous spectrum is constructed using the data matrix. The two-dimensional asynchronous spectrum is the three-dimensional asynchronous spectrum at y = 120 cm -1 The two-dimensional cross section at Fig.12 -A, similarly, from 7 two-dimensional asynchronous spectra at y = 50 cm -1 Take a horizontal section at y = 50 cm to construct a two-dimensional asynchronous spectrum. The two-dimensional asynchronous spectrum is the three-dimensional asynchronous spectrum at y = 50 cm -1 The two-dimensional cross section at Fig.12 -B.

[0119] Step 3, Fig.12 -A, it can be seen that there is a new system missing peak at (190, 190). From the three-dimensional asynchronous spectrum at y = 120 cm -1 Take z=190cm on the two-dimensional cross section at-1 The horizontal section of , the obtained section is the pure substance spectrum of P; Fig.12 In B, it can be seen that there are new systematic missing peaks at (265, 265) and (348, 348). These two systematic missing peaks do not belong to the same group, otherwise systematic missing peaks can also be observed at (265, 348) and (348, 265). From the three-dimensional asynchronous spectrum at y = 50 cm -1 Take z = 265 cm on the two-dimensional cross section at -1 and z = 348 cm -1 The horizontal section of the obtained lines are the pure material spectra of Q and R, respectively, such as Fig.12 -C.

[0120] Afterwards, noise was added to the data generated by the analytical chromatography-spectroscopy experiment. The fluctuation amplitude of the noise was 1% of the maximum signal intensity in the 140 spectrum. The above steps 1-3 were repeated to obtain a three-dimensional asynchronous spectrum at z = 120 cm -1 and z = 50 cm -1 Two two-dimensional cross-sections (respectively as Fig.13 -A and Fig.13 -B) It can be seen that due to the interference of noise, neither cross peaks nor systematic missing peaks can be identified in the two-dimensional cross section of the three-dimensional asynchronous spectrum.

[0121] The signal-to-noise ratios of the three components P, Q, and R are improved by the random swap method of the present invention and the traditional random swap method respectively as follows:

[0122] Random swap method: Taking the absolute intensity of the cross peak of the three-dimensional asynchronous spectrum at (190, 50, 265) as the optimization target, the 140 one-dimensional spectra are arranged in a random sequence, and the two-dimensional cross section of the three-dimensional asynchronous spectrum is generated in a similar way to steps 1-3 above. Then, the cross section is assembled along the Z axis to obtain the three-dimensional asynchronous spectrum, and the absolute intensity of the cross peak of the three-dimensional asynchronous spectrum at (190, 50, 265) is tested. Then, press Figure 1 The one-dimensional spectrum sequence is optimized in the same way, and the NU value is set to 1000. When the absolute intensity of the cross peak of the three-dimensional asynchronous spectrum at (190, 50, 265) is less than 0.0072, the corresponding sequence is a qualified initial sequence. This initial sequence is used as the initial input data to obtain the absolute intensity of the cross peak of the three-dimensional asynchronous spectrum at (190, 50, 265). The above process is repeated 100 times to obtain 100 absolute intensities of the cross peaks of the three-dimensional asynchronous spectrum at (190, 50, 265) optimized by the qualified initial sequence. After arranging the absolute intensities of the cross peaks in ascending order, the results are shown in Figure 1. Fig.14 As shown in the curve a in the middle.

[0123] Random exchange method: The one-dimensional spectrum sequence optimization is performed in a similar way to the random swap method. The difference is that during the optimization process, the sequence of 140 one-dimensional spectra is randomly exchanged and the NU value is set to 5000. The absolute intensities of the cross peaks are finally arranged in ascending order. The results are shown in Fig.14 As shown in curve b.

[0124] Depend on Fig.14 It can be seen that the absolute intensity of the cross peak at (190, 50, 265) of the three-dimensional asynchronous spectrum constructed by arranging the 140 one-dimensional spectra in the sequence optimized by the random exchange method is much greater than the absolute intensity of the cross peak obtained by the random exchange method. The maximum absolute intensity of the cross peak obtained by the random exchange method is only 0.045% of the maximum value obtained by the random exchange method. The above results fully prove that the optimization effect of the random exchange method is significantly better than that of the random exchange method.

[0125] The present invention is described in detail above in combination with preferred embodiments and exemplary examples. However, it should be noted that these specific embodiments are only illustrative explanations of the present invention and do not constitute any limitation on the protection scope of the present invention. Without exceeding the spirit and protection scope of the present invention, various improvements, equivalent substitutions or modifications may be made to the technical content of the present invention and its embodiments, which all fall within the protection scope of the present invention. The protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for enhancing the signal-to-noise ratio of two-dimensional / high-dimensional asynchronous spectroscopy, characterized in that: The method comprises: changing a one-dimensional spectrum sequence to construct a two-dimensional or high-dimensional asynchronous spectrum, cutting a line at a system missing peak of the obtained two-dimensional or high-dimensional asynchronous spectrum to obtain a spectrum of a pure substance, The changing of the one-dimensional spectrum sequence is to select the best sequence of one-dimensional spectrum from the one-dimensional spectrum of a given sequence, thereby constructing a two-dimensional or high-dimensional asynchronous spectrum. The one-dimensional spectrum sequence of the optimal sequence is obtained by random permutation method. From n one-dimensional spectra of a given sequence, n≥11, 11 to 300 one-dimensional spectra are selected to construct a two-dimensional or high-dimensional asynchronous spectrum. Then, the positions of two one-dimensional spectra are randomly exchanged in the selected one-dimensional spectra, and the one-dimensional spectrum of the best sequence is selected by the absolute intensity of the target cross-peak.

2. The method according to claim 1, characterized in that Selecting the best sequence of one-dimensional spectra from a given sequence of one-dimensional spectra comprises the following steps: Step 1), constructing a two-dimensional or high-dimensional asynchronous spectrum according to n one-dimensional spectra of a given sequence, obtaining the absolute intensity of the target cross peak, the upper limit of unsuccessful exchange is set to a specified value, and the number of exchanges is set to zero; Step 2), randomly swapping the positions of the two one-dimensional spectra in step 1) and constructing a new two-dimensional or high-dimensional asynchronous spectrum to obtain the absolute intensity of the target cross peak; Step 3), when the absolute value of the absolute intensity of the target cross peak in step 2) is greater than the absolute value of the absolute intensity of the target cross peak in step 1), it is said that a successful exchange has been achieved, and then the absolute value of the absolute intensity of the target cross peak in step 1) is set to the absolute value of the absolute intensity of the target cross peak in step 2), and the one-dimensional spectrum in step 1) is replaced by the one-dimensional spectrum of the new sequence in step 2), and the value of the number of exchanges is set to zero; otherwise, it is said that an unsuccessful exchange is encountered, and the value of the number of exchanges is increased by 1; Step 4), when the exchange times are lower than the upper limit of unsuccessful exchanges, go to step 2); otherwise, terminate the optimization process, and the optimized sequence is now the best one-dimensional spectral sequence.

3. The method according to claim 1, characterized in that The binary or ternary mixture is detected by thermogravimetric-infrared coupling instrument, and the one-dimensional infrared spectra of the binary mixture at different time points are randomly recorded. 11 to 300 one-dimensional spectra are selected from them, and a data matrix is ​​generated by exchanging two one-dimensional spectral sequences. An asynchronous spectrum is generated according to the data matrix.

4. The method according to claim 3, characterized in that The data matrix is ​​constructed by vectors of signal intensities of 11 to 300 one-dimensional spectra at different wavelengths.

5. The method according to claim 3 or 4, characterized in that: The data matrix of the binary mixture is denoted as A, and the two-dimensional asynchronous spectrum is generated by equation (1): Ψ(x,y) = A T (x)NA(y) Equation (1); In formula (1): Ψ(x,y) represents the intensity at the point (x,y) in the two-dimensional asynchronous spectrum; T represents the transpose of the matrix; N represents the n-th order Hilbert-Noda matrix.

6. The method according to claim 3 or 4, characterized in that: The one-dimensional spectrum of the ternary mixture is divided into m groups, each group contains p one-dimensional spectra, and the p one-dimensional spectra are constructed into a data matrix, and then a two-dimensional asynchronous spectrum is constructed, resulting in a total of m two-dimensional asynchronous spectra. Line sections are taken at different wavelengths of each two-dimensional asynchronous spectrum to obtain two-dimensional cross sections of the three-dimensional asynchronous spectrum at different wavelengths. The obtained two-dimensional cross sections are assembled along the same direction to obtain a three-dimensional asynchronous spectrum.

7. Use of the method according to any one of claims 1 to 6 for acquiring spectra of pure substances in complex samples.

Images