CN104568805B - A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber - Google Patents
A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber Download PDFInfo
- Publication number
- CN104568805B CN104568805B CN201510040384.4A CN201510040384A CN104568805B CN 104568805 B CN104568805 B CN 104568805B CN 201510040384 A CN201510040384 A CN 201510040384A CN 104568805 B CN104568805 B CN 104568805B
- Authority
- CN
- China
- Prior art keywords
- optical fiber
- component
- reaction mechanism
- infrared
- reaction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Abstract
The invention discloses a kind of method of use Non-negative Matrix Factorization method combination middle infrared optical fiber probe research reaction mechanism, comprise the following steps：Step A. will collect the ir data of reaction system online in infrared optical fiber probe intercalation reaction device；The infrared mixed spectra data obtained are carried out denoising and baseline correction by step B. by wavelet transformation；Step C. determines the component number in system with Subspace comparison, completes pretreatment；The preprocessed data that step D. is obtained, which is updated in NMF programs, to be iterated until convergence, obtains the concentration curve corresponding with its of the infrared spectrogram of each component in the system after final optimization pass；The infrared spectrogram that step E. is resolved with reference to every kind of material derives its structure, and derives reaction mechanism with reference to concentration curve.The present invention only can derive reaction mechanism without extra intermediate agent for capturing by online infrared mixed spectra.The invention metering system is simple, data accurate, and measurable response is numerous, can quickly and accurately derive reaction mechanism.
Description
Technical field
The present invention relates to a kind of method of use Non-negative Matrix Factorization method combination middle infrared optical fiber probe research reaction mechanism,
Belong to field of spectral analysis technology.
Background technology
Increase with today's society to chemical products demand, people start to have turned one's attention to correlated response mechanism
In research work, quality control and production process optimization are realized with this, and then reach raising yield, reduce cost, reduction pair
The purpose of environmental pollution.
Current study on mechanism mainly use intermediate agent for capturing, isotope tracing, quantum chemical method and
The methods such as on-line measurement.Spectral technique means can provide frequent on the structural information and change in concentration of material simultaneously because of it
It is used to monitor the reaction mechanism mechanism of reaction.We can rapidly obtain the substantial amounts of information about chemically reacting by spectrographic technique.
Because by the infrared spectrum of analysis of compounds can obtain it is many reflection molecular structures information, these information can be used for into
The measure and Mixture component analysis of row molecular structure of compounds, thus we select to use infrared spectrum side in the present invention
Method detects reaction mechanism.
FTIR spectrum method in situ（in-situ FT-IR）It can detect as reaction carries out the structure of matter and its electricity
The major advantage of the change of energy subband structure, this in-situ techniques and off-line method is inside monitoring reaction system, so as to eliminate sample
The risk that product change or polluted.
The content of the invention
It is an object of the invention to infrared in line spectrum in NMF arithmetic analysis, so as to draw the reaction in reaction system
The respective spectrum of thing, intermediate, product and its corresponding relative concentration curve, and then intuitively derive reaction mechanism.
Implementation process of the present invention is as follows：
A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber, including following step
Suddenly：
（1）Online middle infrared optical fiber is popped one's head in and the MCT detectors in infrared survey instrument（Include the liquid for refrigeration
Nitrogen）It is connected, middle infrared optical fiber probe is inserted into acquisition mid-infrared light modal data in reaction system, the wave-number range of measurement is
4 000~600 cm-1.The chemical reaction temperature carried out in reactor is between -120 DEG C to 180 DEG C, reaction system pressure
No more than 40 MPa, chloroazotic acid etc. is not contained in solution with corrosive substance；
（2）Collected ir data is subjected to denoising and base with Matlab softwares using wavelet transform function
Line is corrected, and completes pretreatment；
（3）Based on principal component analysis（PCA）With simplified self mode method（Simplisma）Two kinds of algorithms, respectively by it is different into
Fraction is separated to pretreatment gained infrared spectrum, is then compared separating obtained Subspace difference value, difference is most
Small place is component number in the reaction system finally determined；
（4）Infrared spectrum obtained by pretreatment is brought directly to be iterated in NMF programs until convergence, number of components selection
For step（3）In determined by component number in system, you can obtain the infrared spectrogram of each component in the system after final optimization pass
With corresponding concentration curve；
（5）From concentration curve determine each component be respectively reactant, intermediate also or product, in conjunction with every kind of component
Infrared spectrogram derive its counter structure, and then draw the reaction mechanism of the reaction.
Above-mentioned steps（2）In, carry out filter using wavelet transform function and make an uproar, choose wavelet function DB5, gone after being divided into 4 layers
Except high-frequency noise, so as to realize that filter is made an uproar；Then baseline correction is carried out, using DB5 wavelet functions, to removing the data point after noise
1-10 layers are not layered as and are reconstructed.
In the present invention, inventor will decay and be totally reflected（ATR）Method is used to monitor whole building-up process.However, in
The multidimensional data that infrared probe is directly collected from complex reaction system can not be directly resolved, thus, inventor carries
Non-negative Matrix Factorization method is gone out to realize Data Dimensionality Reduction, reaction mechanism is successfully derived whereby.
Simple add operation, which is used only, in NMF programs in an iterative process can directly obtain spectrum and concentration information, grasp
Make simple, quick, reliable results.With traditional independent component analysis（ICA）Algorithm is compared, NMF algorithms in an iterative process only
Carry out simple add operation, thus the algorithm can greatly reduce in middle infrared spectrum obtained by parsing in feature spectral peak with
Machine overturns probability, and then can reduce analysis spectrogram, derive intermediate structure and determine the difficulty of reaction mechanism；In addition, NMF algorithms
There is no negative value in the relative concentration drawn, thus with physical significance and interpretation definitely.
Brief description of the drawings
The initial data that Fig. 1 obtains for on-line monitoring；
Fig. 2 be filtered by wavelet transformation make an uproar, reconstruct and after carrying out baseline correction gained data；
Fig. 3 is principal component number determined by after two kinds of algorithms gained Subspace differences are compared；
Fig. 4 is each component material relative concentration in the system calculated based on Non-negative Matrix Factorization method；
Fig. 5 is the infrared spectrogram for each component isolated based on Non-negative Matrix Factorization method；
Fig. 6 is that middle infrared optical fiber probe incorporated Algorithms of Non-Negative Matrix Factorization derives 3- amino 4- together with the anti-of oxamidino furazan
Answer mechanism;
The discreet value of each component relative concentration in the reaction system that Fig. 7 obtains for Evolving methods；
Fig. 8 is each component material relative concentration in the system calculated based on MCR-ALS algorithms；
Fig. 9 is the infrared spectrogram for each component isolated based on MCR-ALS algorithms；
Figure 10 is each component material relative concentration in the system calculated based on ICA algorithm；
Figure 11 is the infrared spectrogram for each component isolated based on ICA algorithm.
Embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
Embodiment 1
Carrying out analyzing step to infrared spectrum by NMF algorithms is：
（1）Synthesized by the 3- amino 4- of online middle infrared optical fiber probe monitors together with oxamidino furazan
The Vertex-70 type infrared spectrometer fibre-optical probe annexes that this example is produced using German Brooker company are supervised online
Survey reaction process.16 mL deionized waters are added into the three-neck flask equipped with thermometer and magnetic stirring apparatus, insertion is online infrared
Fibre-optical probe, water is scanned as background.Then 4.0 g malononitrile are added, heating stirring is to being completely dissolved.Start monitoring
Course of reaction, treats that solution temperature is down to -5 DEG C, sequentially adds 4.2 g natrium nitrosums, 0.3 mL glacial acetic acids, control glacial acetic acid
Rate of addition makes reacting liquid temperature be no more than 10 DEG C, at room temperature the min of coreaction 150.Solution colour is changed into shallow successively from colourless
Green, yellow, eventually become red.Then the NaoH aqueous solution (4.5 g NaOH and 4 mL H are sequentially added into flask2O)、
8.9 g hydroxylamine hydrochlorides, adjust pH to 10 or so.After reaction is carried out about 30 minutes, reaction solution is heated to 150 points of boiling reaction
Clock, stops monitoring reaction.Cooling, filtering, in atmosphere washing, dry yellow solid.
（2）Filter is carried out to original spectral data to make an uproar and baseline correction
Being popped one's head in by middle infrared optical fiber, obtain is a three-dimensional light spectral moment being made up of 1763 × 1424 absorbances
Battle array, as shown in figure 1, resulting spectrum picture is unsmooth, and the baseline of each bar ir data is not consistent, it is necessary to carry out
Filter is made an uproar and baseline correction.Filter is carried out first by wavelet transform function to make an uproar：Wavelet function DB5 is chosen, is removed after being divided into 4 layers high
Frequency noise, so as to realize that filter is made an uproar；Then baseline correction is carried out：DB5 wavelet functions are reused, to removing the data point after noise
1-10 layers are not layered as and is reconstructed, is as a result shown, filter when being divided into 9 layers and being reconstructed is made an uproar best with baseline correction effect
（As shown in Figure 3）.
（3）Determine the number of components in system
Based on different number of components（From 1 to 10）Principal component analysis is used respectively（PCA）With simplified self mode (Simple-
To-Use Interactive Self-Modeling, Simplisma) both algorithms carry out to matrix obtained by upper step respectively
Parsing, then compares the gap between the subspace matrices of their each gained, then obtains Subspace difference degree and become with dimension
Change figure, as shown in figure 3, when flex point occurs in D (N), it is exactly the number of components in system to reach corresponding dimension during low ebb value.As a result show
It is 5 to show the number of components in this example.
（4）Use Non-negative Matrix Factorization（NMF）Algorithm is iterated under conditions of rank number is 5 to spectrum obtained by pretreatment
Until its convergence, respectively obtains each component material relative concentration corresponding to the reaction system each component（Fig. 4）With isolate
The infrared spectrogram of each component（Fig. 5）.It can be drawn by Fig. 4：Component a is reactant, and component b, c, d are intermediate product, group
It is final product to divide e.
（5）The structure of matter is derived with reference to the characteristic peak in the spectrogram of each component in Fig. 5
The component A isolated characteristic peak：The absworption peak of C ≡ N strong stretching vibration（2229,2362 cm-1）Inhaled with C-C
Receive peak（1228 cm-1）It is the IR spectrum of reactant malononitrile to show component A；The component B isolated characteristic peak：In component A
Present in C ≡ N and C-C absworption peak still suffered from component B, and N=O absworption peaks（〜1554cm-1）Appearance show instead
Intermediate 1, i.e. 2- nitrosos malononitrile should be generated；The component C isolated characteristic peak：C ≡ N absworption peak（〜2229cm-1）Still exist, and C=N（1660cm-1）The appearance of absworption peak illustrates that the nitroso in intermediate 1 there occurs oxime glycosylation reaction
Generate intermediate 2, i.e. 2- oximidos malononitrile；The component D isolated characteristic peak：C=N absworption peaks in component D（1664cm-1）
Strength ratio component C in correspondence absworption peak intensity it is stronger, this shows to form more oximido functional groups in component D；Separation
The component E gone out characteristic peak：C=N absworption peak is from 1660cm-1Move to 1622cm-1Expression has belonged to the furazan ring of product
Through being formed.
With reference to the relative concentration curve and correspondence pure component spectrum isolated, reaction mechanism as shown in Figure 6 is derived,
The reaction mechanism and document（The 500 grams of level synthesis of 3- amino -4- amino isonitrosos furazan, energetic material, 2006,14 (1):27-28）
Middle drawn mechanism is consistent.
Embodiment 2
Infrared spectrum is parsed by MCR-ALS algorithms, the step of the present embodiment（1）, step（2）, step（3）
It is same as Example 1.
Step（4）The use of MCR-ALS methods in rank number to be spectrum obtained by pretreatment to be handled under conditions of 5：It is first
Evolving factor analysis first is carried out to pretreated data（EFA）Each component relative concentration in operation, obtained reaction system
Discreet value, as shown in Figure 7；Then EFA acquired results are brought into MCR-ALS programs and be iterated, respectively obtain this anti-
Answer each component material relative concentration corresponding to system each component（Fig. 8）With the infrared spectrogram for each component isolated（Fig. 9）.
It can not judge that each component is belonging respectively to reactant, intermediate or product by Fig. 8.
With spectrogram obtained by NMF algorithms（Fig. 5）Compare, the spectrogram that MCR-ALS is parsed（Fig. 9）In many characteristic peaks not
Substantially and spectral peak flop phenomenon is serious, this Structure Deduction to each material causes very big obstruction.Obviously, MCR- is only passed through
ALS has been difficult the mechanism for being inferred to the reaction.
In addition, EFA programs need about hour to complete to estimate each component relative concentration in reaction system,
This has had a strong impact on the efficiency of real-time analysis detection；And NMF only needs to call a NMF program can be at one point in calculating process
Spectrum and concentration information are directly obtained in clock, simple to operate, quick, reliable results.
Embodiment 3
Infrared spectrum is parsed by ICA algorithm, the step of iterative process, the present embodiment（1）, step（2）, step
Suddenly（3）It is same as Example 1.
Step（4）The use of ICA algorithm in number of components to be spectrum obtained by pretreatment to be handled under conditions of 5, respectively
Obtain each component material relative concentration corresponding to the reaction system each component（Figure 10）With the infrared light for each component isolated
Spectrogram（Figure 11）.It can be drawn by Fig. 7：Component a is reactant, and component b, c, d are intermediate products, and component e is final production
Thing.
With spectrogram obtained by NMF algorithms（Fig. 5）Compare, the spectrogram that ICA is parsed（Figure 11）Middle partial spectrum peak is not obvious
And spectral peak flop phenomenon is serious, and certain obstacle is caused to the Structure Deduction of intermediate product.In addition, ICA parse it is relative
Concentration map（Figure 10）In occur in that the concentration negative value of no practical significance, and the relative concentration figure that NMF is parsed（Fig. 4）In be non-
Negative value, it is more identical with actual conditions.
Above content is to combine the further description that optimal technical scheme is done to the present invention, it is impossible to assert invention
Specific implementation is only limitted to these explanations.For general technical staff of the technical field of the invention, the present invention is not being departed from
Concept thereof under, can also make it is simple deduce and replace, should all be considered as protection scope of the present invention.
Claims (4)
1. a kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber, it is characterised in that bag
Include following steps：
（1）Online middle infrared optical fiber probe is connected with the MCT detectors in infrared survey instrument, middle infrared optical fiber is popped one's head in and inserted
Enter into reaction system and obtain mid-infrared light modal data, the wave-number range of measurement is 4 000~600 cm-1；
（2）Collected ir data is subjected to denoising and baseline school with Matlab softwares using wavelet transform function
Just, pretreatment is completed；
（3）It is infrared to pretreatment gained by heterogeneity number respectively based on principal component analysis and simplified self mode two kinds of algorithms of method
Spectrum is separated, and is then compared separating obtained Subspace difference value, and what is as finally determined at difference minimum is anti-
Answer component number in system；
（4）Infrared spectrum obtained by pretreatment is brought directly to be iterated in NMF programs until convergence, number of components selection is step
Suddenly（3）In determined by component number in system, obtain in the system after final optimization pass the infrared spectrogram of each component and corresponding
Concentration curve；
（5）From concentration curve determine each component be respectively reactant, intermediate also or product, in conjunction with the red of every kind of component
External spectrum figure derives its counter structure, and then draws the reaction mechanism of the reaction.
2. the side of research reaction mechanism according to claim 1 of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber
Method, it is characterised in that：MCT detectors include the liquid nitrogen for refrigeration.
3. the side of research reaction mechanism according to claim 1 of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber
Method, it is characterised in that：Step（1）In, the chemical reaction temperature carried out in reactor is between -120~180 DEG C, reaction system
Pressure is no more than in 40 MPa, solution and does not contain corrosive substance.
4. the side of research reaction mechanism according to claim 1 of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber
Method, it is characterised in that：Step（2）In, carry out filter using wavelet transform function and make an uproar, wavelet function DB5 is chosen, after being divided into 4 layers
High-frequency noise is removed, so as to realize that filter is made an uproar；Then baseline correction is carried out, using DB5 wavelet functions, to removing the data after noise
1-10 layers are layered as respectively and are reconstructed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510040384.4A CN104568805B (en) | 2015-01-27 | 2015-01-27 | A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510040384.4A CN104568805B (en) | 2015-01-27 | 2015-01-27 | A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104568805A CN104568805A (en) | 2015-04-29 |
| CN104568805B true CN104568805B (en) | 2017-09-05 |
Family
ID=53085382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510040384.4A Expired - Fee Related CN104568805B (en) | 2015-01-27 | 2015-01-27 | A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104568805B (en) |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3706437B2 (en) * | 1996-07-13 | 2005-10-12 | 株式会社堀場製作所 | Analysis method of multi-component aqueous solution |
| US9311897B2 (en) * | 2010-12-28 | 2016-04-12 | Indian Institute Of Technology Kanpur | Convergent matrix factorization based entire frame image processing |
| CN103196864B (en) * | 2013-04-12 | 2015-07-08 | 西北大学 | Method of researching reaction mechanism by utilizing MCR-ALS (Multivariate Curve Resolution-Alternating Least Squares) combined with infrared online spectrum |
-
2015
- 2015-01-27 CN CN201510040384.4A patent/CN104568805B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN104568805A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cui et al. | W-jet tagging: optimizing the identification of boosted hadronically-decaying W bosons | |
| US7250306B2 (en) | Method for distinguishing between protein variants | |
| Papadopoulos et al. | Spectroscopy of mobility-selected biomolecular ions | |
| Zhang et al. | Mid-infrared spectroscopy for coffee variety identification: comparison of pattern recognition methods | |
| CN106769974B (en) | Gas concentration Two dimensional Distribution on-line detecting system and method | |
| CN104568805B (en) | A kind of method for research reaction mechanism of being popped one's head in based on Non-negative Matrix Factorization method combination middle infrared optical fiber | |
| CN103196864B (en) | Method of researching reaction mechanism by utilizing MCR-ALS (Multivariate Curve Resolution-Alternating Least Squares) combined with infrared online spectrum | |
| CN110441254A (en) | A kind of near-infrared frequency comb spectrometer of plastics for identification | |
| Cho et al. | UV resonance Raman and absorption studies of angiotensin II conformation in lipid environments | |
| Jiang et al. | High precision qualitative identification of yeast growth phases using molecular fusion spectra | |
| CN103515185B (en) | Ionization apparatus that a kind of Mass Spectrometer Method samples of juice middle peasant is residual and detection method | |
| CN106526040B (en) | Differentiate the method for fresh cocoon/dried cocoon raw silk based on GC or GC-MS technologies | |
| Di-Na et al. | Study on fast recognition of biotoxins and biological modifiers using data fusion algorithm | |
| Raimondo et al. | Classification of the Ibuprofen Active Pharmaceutical Ingredients by chemical patterns combining HPLC, 1H-NMR spectroscopy and Chemometrics: Traceability of legal medicines. | |
| CN105445220A (en) | Simple identification method for optical isomer | |
| CN106022025A (en) | Airflow decomposition unlocking method based on mobile terminal | |
| CN110308109A (en) | A method of perchlorate content range in tealeaves is differentiated based near infrared spectrum | |
| CN107525855B (en) | The screening method of antifungal drug chemical risk substance in a kind of washing product | |
| CN106124451B (en) | A method of update the system and detection through the detection of packaging bag On-line near infrared analyzer | |
| Wang et al. | Identification of tartrazine and sunset yellow by fluorescence spectroscopy combined with radial basis function neural network | |
| CN208350633U (en) | A kind of celadon processing feedstock analysis detection device | |
| CN203553100U (en) | Ionization apparatus and mass spectrometry system for conducting mass spectrometric detection on pesticide residue in fruit juice sample | |
| CN220104835U (en) | Chinese herbal medicine type and producing area identification device | |
| CN220040214U (en) | Chinese medicinal material type and producing area identification system based on spectrum technology | |
| Slabey | Contribution of the Cyclopropyl Ring to Molar Refraction1 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170905 Termination date: 20180127 |