CN113514416A - Medicine crystallization process quality nonlinear characterization method based on near-infrared sensor fusion - Google Patents
Medicine crystallization process quality nonlinear characterization method based on near-infrared sensor fusion Download PDFInfo
- Publication number
- CN113514416A CN113514416A CN202110607952.XA CN202110607952A CN113514416A CN 113514416 A CN113514416 A CN 113514416A CN 202110607952 A CN202110607952 A CN 202110607952A CN 113514416 A CN113514416 A CN 113514416A
- Authority
- CN
- China
- Prior art keywords
- fusion
- model
- spectrum
- infrared
- infrared sensor
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 230000004927 fusion Effects 0.000 title claims abstract description 56
- 238000002425 crystallisation Methods 0.000 title claims abstract description 40
- 230000008025 crystallization Effects 0.000 title claims abstract description 40
- 230000008569 process Effects 0.000 title claims abstract description 37
- 239000003814 drug Substances 0.000 title claims abstract description 20
- 229940079593 drug Drugs 0.000 title claims abstract description 18
- 238000012512 characterization method Methods 0.000 title claims abstract description 16
- 238000001228 spectrum Methods 0.000 claims abstract description 58
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 238000002329 infrared spectrum Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 43
- 239000001512 FEMA 4601 Substances 0.000 claims description 19
- HELXLJCILKEWJH-SEAGSNCFSA-N Rebaudioside A Natural products O=C(O[C@H]1[C@@H](O)[C@@H](O)[C@H](O)[C@@H](CO)O1)[C@@]1(C)[C@@H]2[C@](C)([C@H]3[C@@]4(CC(=C)[C@@](O[C@H]5[C@H](O[C@H]6[C@H](O)[C@@H](O)[C@H](O)[C@@H](CO)O6)[C@@H](O[C@H]6[C@H](O)[C@@H](O)[C@H](O)[C@@H](CO)O6)[C@H](O)[C@@H](CO)O5)(C4)CC3)CC2)CCC1 HELXLJCILKEWJH-SEAGSNCFSA-N 0.000 claims description 19
- HELXLJCILKEWJH-UHFFFAOYSA-N entered according to Sigma 01432 Natural products C1CC2C3(C)CCCC(C)(C(=O)OC4C(C(O)C(O)C(CO)O4)O)C3CCC2(C2)CC(=C)C21OC(C1OC2C(C(O)C(O)C(CO)O2)O)OC(CO)C(O)C1OC1OC(CO)C(O)C(O)C1O HELXLJCILKEWJH-UHFFFAOYSA-N 0.000 claims description 19
- HELXLJCILKEWJH-NCGAPWICSA-N rebaudioside A Chemical compound O([C@H]1[C@H](O)[C@@H](CO)O[C@H]([C@@H]1O[C@H]1[C@@H]([C@@H](O)[C@H](O)[C@@H](CO)O1)O)O[C@]12C(=C)C[C@@]3(C1)CC[C@@H]1[C@@](C)(CCC[C@]1([C@@H]3CC2)C)C(=O)O[C@H]1[C@@H]([C@@H](O)[C@H](O)[C@@H](CO)O1)O)[C@@H]1O[C@H](CO)[C@@H](O)[C@H](O)[C@H]1O HELXLJCILKEWJH-NCGAPWICSA-N 0.000 claims description 19
- 235000019203 rebaudioside A Nutrition 0.000 claims description 19
- 238000007500 overflow downdraw method Methods 0.000 claims description 10
- 239000000047 product Substances 0.000 claims description 10
- 230000003595 spectral effect Effects 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 239000006228 supernatant Substances 0.000 claims description 5
- 239000013307 optical fiber Substances 0.000 claims description 4
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 12
- 238000003908 quality control method Methods 0.000 abstract description 6
- 239000000523 sample Substances 0.000 description 15
- 230000008901 benefit Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 238000000746 purification Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000008030 elimination Effects 0.000 description 3
- 238000003379 elimination reaction Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004497 NIR spectroscopy Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000003153 chemical reaction reagent Substances 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- UEDUENGHJMELGK-HYDKPPNVSA-N Stevioside Chemical compound O([C@@H]1[C@@H](O)[C@H](O)[C@@H](CO)O[C@H]1O[C@]12C(=C)C[C@@]3(C1)CC[C@@H]1[C@@](C)(CCC[C@]1([C@@H]3CC2)C)C(=O)O[C@H]1[C@@H]([C@@H](O)[C@H](O)[C@@H](CO)O1)O)[C@@H]1O[C@H](CO)[C@@H](O)[C@H](O)[C@H]1O UEDUENGHJMELGK-HYDKPPNVSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000012569 chemometric method Methods 0.000 description 1
- 239000013068 control sample Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000007791 liquid phase Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000012528 membrane Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000007790 solid phase Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 229940013618 stevioside Drugs 0.000 description 1
- OHHNJQXIOPOJSC-UHFFFAOYSA-N stevioside Natural products CC1(CCCC2(C)C3(C)CCC4(CC3(CCC12C)CC4=C)OC5OC(CO)C(O)C(O)C5OC6OC(CO)C(O)C(O)C6O)C(=O)OC7OC(CO)C(O)C(O)C7O OHHNJQXIOPOJSC-UHFFFAOYSA-N 0.000 description 1
- 235000019202 steviosides Nutrition 0.000 description 1
- 239000013076 target substance Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3577—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing liquids, e.g. polluted water
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a near-infrared sensor fusion-based nonlinear characterization method for quality of a drug crystallization process and application thereof, belonging to the technical field of detection analysis and drug quality control. The detection method comprises the steps of establishing a multispectral collection platform, collecting spectrums, processing near infrared spectrums and constructing a CNN nonlinear fusion model. The invention simulates the RA crystallization process, and utilizes the multispectral collection platform and the CNN nonlinear fusion model for prediction, and the result shows that the invention successfully realizes the nonlinear characterization method aiming at the RA crystallization process, and realizes the rapid analysis of the key quality in the RA crystallization process, thereby laying a foundation for the quality control thereof, and having good practical application value.
Description
Technical Field
The invention belongs to the technical field of detection analysis and drug quality control, and particularly relates to a drug crystallization process quality nonlinear characterization method based on near infrared spectrum data fusion
Background
The information in this background section is only for enhancement of understanding of the general background of the invention and is not necessarily to be construed as an admission or any form of suggestion that this information forms the prior art that is already known to a person of ordinary skill in the art.
Near infrared spectroscopy (NIRS) is a rapid detection technology developed in recent years and belongs to molecular vibration spectroscopy. The Near Infrared (NIR) spectrum region refers to electromagnetic waves between a visible region (VIS) and a Mid Infrared (MIR) region, has a wavelength range of 780-2526 nm, and is divided into short-wave near infrared (700-1100 nm) and long-wave near infrared (1100-2500 nm). Because the spectrum is easy to obtain and stable, but the defect of serious spectrum overlapping exists, the assistance of a chemometric method is needed so as to find out useful absorption wavelengths from the overlapping and crossing of near infrared spectrum multi-component absorption to complete the analysis process.
The crystallization is used as an important medicine purification means, is applied to the separation and purification processes of bulk drugs, and is particularly suitable for the production of raw materials with target characteristics. The NIR technology is used for detecting the crystallization process, so that the understanding of the product quality forming process is enhanced, and the analysis, design and optimization of the crystallization process are enhanced, so that the quality of the raw materials and the final product can reach the design index. In some crystallization systems, researchers believe that the presence of solids tends to foul the probe, distort the spectrum, and produce large deviations from the model predictions for crystallization, and therefore care must be taken to create ideal conditions for use in the crystallization process. However, other researchers believe that the characteristic actually represents that the near infrared contains rich information of solid and liquid phases at the same time, and is not a disadvantage.
For process analysis, the first generation online analytical instrument represented by a laboratory desktop has the advantages of high accuracy and strong stability, but has the limitations of slow spectrum acquisition speed, high requirement on installation environment, low cost performance and the like. Miniaturized, low cost NIR sensors are beginning to enter the field of view of researchers. The micro near infrared spectrum sensor has the advantages of mature and stable technology, high integration level, small volume, light weight, high cost performance, strong adaptability, good selectivity and the like, but has the defect of short wavelength range.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a near-infrared sensor fusion drug crystallization process quality nonlinear characterization method, develops a new generation online analysis method based on micro-sensor fusion, effectively solves the limitation of wavelength range, reduces the cost of NIR process analysis, realizes the rapid analysis of key quality in the drug crystallization process, improves the prediction capability of a model, and has remarkable social and economic benefits. The invention is realized by the following technical scheme:
the invention provides a near-infrared sensor fusion method, which comprises the steps that near-infrared light respectively reaches two or more near-infrared sensors connected in parallel through a substance to be detected and an optical fiber branching unit, detection signals are read through an upper computer, fusion spectrum collection is completed, and the models of the near-infrared sensors are different.
In a second aspect of the present invention, a method for nonlinear characterization of quality of a near-infrared sensor fused drug crystallization process is provided, wherein the method comprises: the method comprises the steps of establishing a multispectral acquisition platform based on a near infrared sensor fusion method, acquiring a spectrum, processing a near infrared spectrum and constructing a Convolutional Neural Network (CNN) nonlinear fusion model.
In the third aspect of the invention, a method for representing the content of the supernatant in the Rebaudioside A (RA) crystallization process is constructed, and a nonlinear rebaudioside a content model is established by fusing a convolution neural network for collecting spectrum and original spectrum fusion data by utilizing two near infrared sensors capable of detecting different wavelength ranges.
The invention has the following advantages and positive effects:
(1) the near-infrared sensor fusion method has the advantages that the fusion spectrum can be collected by using the fusion spectrum measuring platform through the fusion of the near-infrared sensors, the limitation of the wavelength range can be effectively solved, the cost of NIR process analysis is reduced, and the social and economic benefits are remarkable.
(2) The invention establishes a CNN-based nonlinear characterization technology through the treatment of the fusion spectrum, and can accurately predict the quality change of the medicine in the crystallization process.
(3) According to the method, the rapid characterization of the RA content of the supernatant in the RA crystallization process is realized based on the near infrared spectrum technology, the prediction capability of the model can be effectively improved based on the CNN nonlinear fusion model, the rapid analysis of the key quality in the RA crystallization process is realized, and thus a foundation is laid for the quality control of the RA crystallization process.
(4) The method has the advantages of accurate and rapid measurement, no damage to the sample, no pollution, shortened monitoring time, and saved manpower, financial resources and material resources; the method has important significance for realizing the online monitoring and quality control of the product quality, thereby having good practical application value.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a multi-spectral collection platform;
FIG. 2 shows spectra collected by two micro-sensors (left sensor 1, right sensor 2);
FIG. 3 shows rejection of abnormal samples;
FIG. 4 is a sample set partitioning;
fig. 5 is a designed CNN model network structure;
FIG. 6 is a variation of a loss function in an iterative training process of a CNN model;
FIG. 7 is a modeling result of a CNN model fused with original spectra in series;
FIG. 8 is a modeling result of different convolution kernel sizes of the outer product fused CNN model;
fig. 9PCA fuses CNN model modeling results.
Detailed Description
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise. It is to be understood that the scope of the invention is not to be limited to the specific embodiments described below; it is also to be understood that the terminology used in the examples is for the purpose of describing particular embodiments only, and is not intended to limit the scope of the present invention.
The present invention is further illustrated by reference to specific examples, which are intended to be illustrative only and not limiting. If the experimental conditions not specified in the examples are specified, the conditions are generally as usual or as recommended by the reagents company; reagents, consumables and the like used in the following examples are commercially available unless otherwise specified.
As described above, although there are many characterization methods for the drug crystallization purification process, the problems of complicated detection method, slow detection speed, high detection cost, and the like generally exist.
In view of this, in a typical embodiment of the present invention, a near-infrared sensor fusion method is provided, where near-infrared light passes through a substance to be detected and then respectively reaches two or more near-infrared sensors connected in parallel through an optical splitter, detection signals are read by an upper computer to complete collection of fusion spectra, and the near-infrared sensors are different in model;
in an exemplary embodiment of the present invention, the near infrared sensor is a micro near infrared sensor, and preferably, the different near infrared sensors detect different wavelength ranges;
the method supports parallel connection of NIRONE sensors of different models, simultaneously detects the substances to be detected, has the advantages that the NIRONE sensors of different models can be combined at will, sensors of different models cover different wavelengths and contain different spectral characteristic information, a spectral detection system built based on the method can select different Sensor combinations according to the characteristic wavelength of a target substance, the model performance is improved through fusion of micro sensors of different wave bands, detection of the full wavelength of a desk type spectrometer can be replaced, time and labor are saved, the spectral acquisition cost is reduced, and the economic benefit is improved.
In an exemplary embodiment of the present invention, preferably, the near infrared spectrum collection mode is a transmission mode.
In an exemplary embodiment of the present invention, preferably, in the method, a temperature control accessory is used for controlling temperature, the influence of temperature on the NIR spectrum of the liquid sample is relatively large, and the temperature is controlled to ± 0.01 ℃ by the temperature control accessory, so as to further improve the accuracy of the spectrum.
In an exemplary embodiment of the present invention, a method for nonlinear characterization of quality of a near-infrared sensor fused drug crystallization process is provided, the method comprising: the construction of a multispectral collection platform and the collection of spectrums based on a near-infrared sensor fusion method are used for processing near-infrared spectrums and constructing a convolutional neural network nonlinear fusion model, wherein the construction of the multispectral collection platform comprises the fusion of near-infrared sensors.
In an exemplary embodiment of the present invention, a method for rapidly characterizing the drug content in the supernatant during rebaudioside a crystallization is provided, wherein a nonlinear rebaudioside a content model is established by fusing a convolution neural network of collected spectra and raw spectra fused data by two near infrared sensors capable of detecting different wavelength ranges;
the CNN-based nonlinear fusion model can effectively improve the prediction capability of the model and realize the rapid analysis of key quality in the RA crystallization process, thereby laying a foundation for the quality control of the RA crystallization process.
In an exemplary embodiment of the present invention, there is provided a method for rapidly characterizing the drug content of the supernatant during the crystallization of rebaudioside a, comprising the steps of:
(1) the construction of a multispectral collection platform based on a near-infrared sensor fusion method comprises the following steps: and constructing a multispectral acquisition platform based on a transmission mode. A light source enters a temperature control sample cell with a collimation function through an optical fiber, after a substance to be detected is placed in a cuvette, near infrared light respectively reaches a near infrared spectrometer sensor 1 and a near infrared spectrometer sensor 2 through the substance to be detected and then through an optical fiber branching unit, and detection signals are read through an upper computer, so that spectrum collection is completed;
(2) sample preparation: simulating a drug crystallization purification process in a laboratory, and sampling at intervals according to the content change trend of the crystallization process to obtain a crystallization process sample;
(3) spectrum collection: the sample is placed in a cuvette with an inner diameter of 1mm, and the temperature is controlled by a temperature control accessory. Collecting spectral data after setting collection parameters;
(4) abnormal spectrum union set elimination and sample set division: the raw spectra were analyzed using PCA and the sample set was screened for the presence of outliers by the degree of vergence. In order to ensure the consistency of samples used by a single sensor and a two-sensor fusion model, a union set of spectrum abnormal samples of the two sensors is removed, and sample set division and model establishment and verification are carried out on the rest samples;
(5) CNN network structure design: setting a CNN network structure, and carrying out model training;
(6) establishing an original spectrum fusion CNN model: and after the model is configured, introducing the spectral data and the primary data into a neural network for training. Optimizing the size of a convolution kernel, the number of the convolution kernels and the number of fully-connected neurons, and establishing an original spectrum fusion CNN model;
(7) establishing an outer product fusion CNN model: similarly, optimizing the size of a convolution kernel, the number of the convolution kernels and the number of fully connected neurons, and establishing an outer product fusion CNN model;
(8) establishing a main component fusion CNN model: respectively carrying out principal component analysis on the sensor 1 and the sensor 2, determining the number of fused principal components, and establishing a principal component fusion CNN model;
(9) model comparison determines the optimal fusion model: with RMSEP, Rp 2RPD is the model evaluation parameter, and a good model should have a lower RMSEP and a higher Rp 2And an RPD value.
Embodiment RA crystallization process quality nonlinear characterization based on near infrared spectrum data fusion
(1) The construction of a multispectral collection platform based on a near-infrared sensor fusion method comprises the following steps: based on the multispectral collection platform of the transmission mode that this patent was built, the wavelength range of the sensor of selecting in this example is: 1350nm-1650nm and 1550nm-1950nm, the two bands not only contain the spectral characteristic information of RA, but also contain the spectral characteristic information of water molecules, and the two kinds of information are converged, so that the prediction capability of the model is improved more easily.
(2) Sample preparation: the crystallization purification process of stevioside RA80 is simulated in a laboratory, samples are taken at intervals according to the content variation trend of the crystallization process, and the samples are filtered by a 0.22 mu m microporous filter membrane for later use. A total of 5 batches of crystallization were simulated, yielding 119 samples.
(3) Spectrum collection: the samples were loaded into cuvettes having an internal diameter of 1mm and the temperature was controlled at 30 ℃ using a temperature control attachment. And (3) carrying out spectrum collection by adopting the micro sensors 1 and 2 in combination with spectrum collection software Nirone _ V1.6.2, wherein the collection mode is transmission. Adopting default acquisition parameters: wavelength averaged 100, scan averaged 1, resolution 1nm, and scan interval 5 s. Background was collected once per batch against air background. In the collection process, 3 times of spectra are collected for each sample, the average spectrum is taken as the final original spectrum, and the spectrum collection is completed in two days. The spectrum acquisition results are shown in fig. 2.
(4) Abnormal spectrum union set elimination and sample set division: the raw spectra were analyzed using PCA and the sample set was screened for the presence of outliers by the degree of vergence. The PCA score scatter plots of the sample spectra acquired by sensor 1 and sensor 2 are shown in fig. 3, from which it can be seen that samples 1, 2, 10, 40, 94 lie outside the 95% confidence limits for the sensor 1 spectrum and samples 11, 12, 13, 22, 40, 94, 116 lie outside the 95% confidence limits for the sensor 2 spectrum. In order to ensure the consistency of samples used by the single-sensor and two-sensor fusion models, a union set of spectrum abnormal samples of the two sensors is removed, and the final residual 109 samples are divided into a correction set verification set (80: 29) for model establishment and verification.
(5) CNN network structure design: setting a CNN network structure, and selecting a LeakyRelu activation function and an Adam optimizer. The schematic diagram of the network structure and the result of the iterative training are shown in fig. 5, and under the set optimizer and the inactivation rate, the loss function is stably reduced along with the progress of the iteration, and is basically stable after 500 iterations, so that the modeling time is saved, the working efficiency is improved, and the iteration number is set to be 500.
(6) Establishing an original spectrum fusion CNN model: and after the model is configured, introducing the spectral data and the primary data into a neural network for training. The number of convolution kernels 128 and the number of fully connected neurons 64 are determined through multiple experiments, then the size of the convolution kernels is optimized, the size of the convolution kernels is determined to be 15, and an original spectrum fusion CNN model is established. RMSEC, RMSEP, R of the modelc 2、Rp 2RPD 2.520, 2.895, 0.958, 0.935, 3.919, respectively.
Table 1 shows the modeling results of different convolution kernel sizes of the original spectrum series fused CNN model;
(7) establishing an outer product fusion CNN model: similarly, the size of the convolution kernel is optimized on the basis of the number of the convolution kernels of 128 and the number of the fully-connected neurons of 64, the size of the convolution kernel is determined to be 30, and an outer product fusion CNN model is established. RMSEC, RMSEP, R of the modelc 2、Rp 2RPD is 3.858, 3.382 and 0.9 respectively02、0.911、3.355。
Table 2 shows the modeling results of different convolution kernel sizes of the outer product fusion CNN model;
(8) establishing a main component fusion CNN model: principal component analysis was performed on sensor 1 and sensor 2, respectively, and multiple experiments confirmed that the first 5 principal components of each extraction were fused (the explanatory variables were 99.88% and 99.67%, respectively). Similarly, on the basis of the number of the convolution kernels of 128 and the number of the fully-connected neurons of 64, the size of the convolution kernels is optimized, the size of the convolution kernels is determined to be 5, and a principal component fusion CNN model is established. RMSEC, RMSEP, R of the modelc 2、Rp 2RPD are 3.187, 3.029, 0.933, 0.929, 3.745, respectively.
Table 3 shows the PCA fusion (5+5) CNN model parameter optimization;
(9) model comparison determines the optimal fusion model: with RMSEP, Rp 2RPD is the model evaluation parameter, and a good model should have a lower RMSEP and a higher Rp 2And an RPD value. Original spectrum fusion CNN model RMSEP minimum, Rp 2And the RPD value is the highest and reaches 3.919, so that the accurate characterization of the RA crystallization process can be realized.
In conclusion, the accuracy of the nonlinear characterization of the quality of the drug crystallization process by the near infrared spectrum acquisition of the RA crystallization process, the union elimination of abnormal samples of the data of the two sensors and the verification of the near infrared sensor fusion by the CNN feature extraction combined with the data fusion is realized. And comparing the modeling results of the fusion models to determine that the CNN model established by the original spectrum fusion is the best method for realizing the nonlinear characterization of the RA crystallization process.
It should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention, and the present invention is not limited thereto, and although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications and equivalents can be made in the technical solutions described in the foregoing embodiments, or equivalents thereof. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention. Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A near-infrared sensor fusion method, characterized in that the method comprises: near-infrared light respectively reaches two or more than two near-infrared sensors connected in parallel through a substance to be detected and then through an optical fiber branching unit, detection signals are read through an upper computer, collection of fusion spectra is completed, and the models of the near-infrared sensors are different.
2. The near-infrared sensor fusion method of claim 1, wherein the near-infrared sensor is a micro near-infrared sensor; preferably, the near infrared light sensors detect different wavelength ranges; preferably, the near infrared spectrum acquisition mode is a transmission mode; preferably, the temperature control accessory is adopted for temperature control in the method.
3. A method for nonlinear characterization of quality of a near-infrared sensor fused drug crystallization process, the method comprising: the construction of a multispectral collection platform and the collection of spectrums based on a near-infrared sensor fusion method are used for processing near-infrared spectrums and constructing a convolutional neural network nonlinear fusion model.
4. A method for rapidly characterizing the drug content of the supernatant during the crystallization of rebaudioside a, the method comprising: a non-linear rebaudioside A content model is established by utilizing two near-infrared sensors capable of detecting different wavelength ranges to fuse a convolution neural network of collected spectrum and original spectrum fusion data.
5. The method according to claim 4, characterized in that the method comprises in particular: the method comprises the steps of establishing a multispectral collection platform, collecting spectrums, removing abnormal spectrum union sets, dividing sample sets, designing a CNN network structure, establishing an original spectrum fusion convolutional neural network model, establishing an outer product fusion CNN model, establishing a principal component fusion CNN model, and comparing the models to determine an optimal fusion model.
6. The method of claim 4, wherein the near infrared sensor detects wavelengths in the range of 1350nm to 1650nm and 1550nm to 1950nm, respectively.
7. The method of claim 5, wherein the raw spectra are subjected to outlier spectral union and sample set rejection using PCA.
8. The method as claimed in claim 5, wherein the original spectrum fusion CNN model is established by introducing the spectrum data and the primary data into a neural network for training after the model is configured, optimizing the size of the convolution kernel, the number of the convolution kernels and the number of the fully-connected neurons, and establishing the original spectrum fusion CNN model.
9. The method of claim 5, wherein the outer product fused CNN model is established by: and optimizing the size of the convolution kernel, the number of the convolution kernels and the number of the fully-connected neurons.
10. The method according to claim 5, wherein in the principal component fusion CNN model building, principal component analysis is performed on two sensors respectively, the number of fused principal components is determined, and a principal component fusion CNN model is built.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110607952.XA CN113514416B (en) | 2021-06-01 | 2021-06-01 | Near infrared sensor fusion-based nonlinear characterization method for quality of drug crystallization process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110607952.XA CN113514416B (en) | 2021-06-01 | 2021-06-01 | Near infrared sensor fusion-based nonlinear characterization method for quality of drug crystallization process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113514416A true CN113514416A (en) | 2021-10-19 |
| CN113514416B CN113514416B (en) | 2023-04-28 |
Family
ID=78065326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110607952.XA Active CN113514416B (en) | 2021-06-01 | 2021-06-01 | Near infrared sensor fusion-based nonlinear characterization method for quality of drug crystallization process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113514416B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106404713A (en) * | 2016-11-23 | 2017-02-15 | 重庆大学 | Double-detector micro near-infrared spectrometer with whole spectral band of 800-2,500nm |
| CN112098358A (en) * | 2020-09-07 | 2020-12-18 | 燕山大学 | Near infrared spectrum parallel fusion quantitative modeling method based on quaternion convolution neural network |
-
2021
- 2021-06-01 CN CN202110607952.XA patent/CN113514416B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106404713A (en) * | 2016-11-23 | 2017-02-15 | 重庆大学 | Double-detector micro near-infrared spectrometer with whole spectral band of 800-2,500nm |
| CN112098358A (en) * | 2020-09-07 | 2020-12-18 | 燕山大学 | Near infrared spectrum parallel fusion quantitative modeling method based on quaternion convolution neural network |
Non-Patent Citations (2)
| Title |
|---|
| PUNEET MISHRA ET AL.: "Sequential fusion of information from two portable spectrometers for improved prediction of moisture and soluble solids content in pear fruit", 《TALANTA》 * |
| 周昭露等: "近红外光谱技术在中药质量控制应用中的化学计量学建模:综述和展望", 《化工进展》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113514416B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Feng et al. | Investigation on data fusion of multisource spectral data for rice leaf diseases identification using machine learning methods | |
| Bangalore et al. | Genetic algorithm-based method for selecting wavelengths and model size for use with partial least-squares regression: application to near-infrared spectroscopy | |
| Yilmaz et al. | Application of Raman spectroscopy in monoclonal antibody producing continuous systems for downstream process intensification | |
| Jin et al. | Near infrared spectroscopy in combination with chemometrics as a process analytical technology (PAT) tool for on-line quantitative monitoring of alcohol precipitation | |
| CN109863239A (en) | The spectroscopy system and method for identification and quantification for pathogen | |
| CN104568822B (en) | A kind of capsule of weeping forsythia medicinal material multi objective while quick determination method | |
| CN104062257A (en) | Method for determining total flavone content of solution based on near infrared spectroscopy | |
| Véstia et al. | Predicting calcium in grape must and base wine by FT-NIR spectroscopy | |
| Laghi et al. | FTIR spectroscopy and direct orthogonal signal correction preprocessing applied to selected phenolic compounds in red wines | |
| Sousa et al. | A method for determination of COD in a domestic wastewater treatment plant by using near-infrared reflectance spectrometry of seston | |
| CN102621092B (en) | Method for detecting Danhong injection ethanol precipitation process on line | |
| CN103201616A (en) | Spectroscopic finger-printing of raw materials | |
| Power et al. | From the laboratory to the vineyard—evolution of the measurement of grape composition using NIR spectroscopy towards high-throughput analysis | |
| CN103115892A (en) | Method for preparing ginkgo leaf extract by using near infrared spectroscopy analysis technology | |
| CN104977271B (en) | A kind of safflower alcohol precipitation process active ingredient near infrared online detection method | |
| CN114026408A (en) | Method for identifying microorganisms | |
| CN104390926B (en) | Online rapid detection method of herba andrographis concentrated decolorization process | |
| CN112666119A (en) | Method and system for detecting ginseng tract geology based on terahertz time-domain spectroscopy | |
| CN108548794A (en) | A kind of biological products method for transferring near infrared model | |
| Cozzolino | State-of-the-art advantages and drawbacks on the application of vibrational spectroscopy to monitor alcoholic fermentation (beer and wine) | |
| CN113029994B (en) | Microcystin concentration inversion method based on multi-source characteristic spectrum of extracellular organic matter | |
| CN113514416B (en) | Near infrared sensor fusion-based nonlinear characterization method for quality of drug crystallization process | |
| CN103760130A (en) | Method for measuring tween-80 content in compound musk injection by using near-infrared spectrum | |
| Qiu et al. | Simultaneous sex and species classification of silkworm pupae by NIR spectroscopy combined with chemometric analysis | |
| Luo et al. | Rapid quantification of multi-components in alcohol precipitation liquid of Codonopsis Radix using near infrared spectroscopy (NIRS) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |