CN-121994735-A - In-situ catalytic reaction infrared spectrum characterization data intelligent analysis system

CN121994735ACN 121994735 ACN121994735 ACN 121994735ACN-121994735-A

Abstract

The invention relates to an in-situ catalytic reaction infrared spectrum characterization data intelligent analysis system which comprises a preprocessing module, a characteristic extraction module, a machine learning identification module and a data analysis module, wherein the preprocessing module is used for carrying out real-time dynamic preprocessing on in-situ infrared spectrum data, the characteristic extraction module is used for extracting characteristics of the preprocessed spectrum data to obtain spectrum characteristic parameters, the machine learning identification module is used for identifying the spectrum characteristic parameters through a machine learning model to obtain a chemical formula or a chemical chain, and the data analysis module is used for obtaining dynamic change trend of in-situ reaction products according to the chemical formula or the chemical chain. The invention carries out real-time automatic data preprocessing and feature recognition aiming at the in-situ catalytic reaction process, improves the reliability and scientificity of data preprocessing and feature recognition by utilizing a self-correction circulation function, and combines database inquiry and deep learning technologies to realize intelligent analysis of in-situ infrared spectrum data.

Inventors

YU SHUWEN
JIN YAN
ZHAO DENGCAI
XU FEIFEI
LIU YUAN

Assignees

中国科学院大连化学物理研究所
榆林中科洁净能源创新研究院

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (10)

1. An in situ catalytic reaction infrared spectrum characterization data intelligent analysis system, which is characterized by comprising: the preprocessing module is used for carrying out real-time dynamic preprocessing on the in-situ infrared spectrum data through the preprocessing module; the feature extraction module is used for obtaining spectral feature parameters through feature extraction of the preprocessed spectral data; the machine learning identification module is used for identifying the spectral characteristic parameters through a machine learning model to obtain the chemical structure information of the included catalytic reaction intermediate and the product; And the data analysis module is used for acquiring the dynamic change trend of the in-situ reaction product according to the change of the chemical structure along with the reaction condition.
2. The intelligent analysis system for in-situ catalytic infrared spectrum characterization data according to claim 1, wherein the preprocessing module comprises: 1) The spectrum data difference subtraction preprocessing module is used for making a standard curve for the infrared spectrum of the initial state of the sample before the catalytic reaction, making a difference between the spectrum data in the reaction process and the standard curve, and segmenting the spectrum range according to the data difference subtraction result; 2) The baseline correction module is used for selecting different correction models for carrying out baseline removal processing on each spectrum data according to the segmentation result to form corrected spectrum data; 3) The smooth noise reduction module is used for carrying out smooth noise reduction on each piece of corrected spectrum data; 4) The wave searching and segmenting module is used for identifying the position of a wave trough point by calculating the first derivative and the second derivative of spectrum data based on the data after smooth noise reduction and automatically segmenting the spectrum according to the position of the wave trough point; 5) The fitting peak dividing module is used for carrying out peak dividing fitting on each segment of spectrum data after segmentation and outputting peak characteristic parameters; 6) The loop feedback evaluation module is used for comparing the characteristic difference between the peak-dividing fitting curve and the original infrared spectrum, comprising peak position, peak area and peak characteristic parameter, obtaining the peak area ratio and peak height ratio of the fitting curve and the original infrared spectrum, respectively serving as a determining coefficient and a peak height weight factor; returning to the step 3) to adjust the size of a filter window in the smoothing noise reduction module when the weight factor obtained by the (i+1) th iteration is still lower than the threshold value; And returning to the step 2) to adjust the parameters of different correction models until the weight factor is not lower than the threshold value when the weight factor obtained by the (i+2) th iteration is still lower than the threshold value.
3. The intelligent analysis system for in-situ catalytic infrared spectroscopy data of claim 1, wherein the baseline correction module comprises a plurality of correction models: The polynomial fitting model is used for fitting the data through a least square method to obtain polynomial fitted data; The segmented baseline fitting model is used for dividing the data into a plurality of segments, and obtaining segmented baseline fitted data by adopting different fitting methods for each segment; The moving window smoothing model is used for sliding on the data sequence through a window with a set length, and the statistical value of each data point in the window is used as a smoothing value of a window center point, so that smoothed data are obtained; And the punishment least square method model is used for enabling the square error between the predicted value and the actual value of the regression model to be minimum through the regression model, and further obtaining the regressed data.
4. The intelligent analysis system for in-situ catalytic infrared spectrum characterization data according to claim 1, wherein the fitting peak splitting module performs the following steps: 1) According to symmetry and distribution conditions of the spectrum characteristics, carrying out Gaussian fitting and Lorentz fitting parallel operation to evaluate applicability of Gaussian fitting and Lorentz fitting according to the degree of fitting minimum deviation: The formula of the Gaussian fitting is: the formula of the lorentz fit is: Wherein a is the peak height, b is the peak position, c is the peak width, y (x) represents the fitted value, x represents the input spectral wavenumber; 2) Evaluating the fitting effect by two evaluation criteria, namely, obtaining a decision coefficient R 2 and a peak height ratio H for each fitting mode of Gaussian fitting and Lorentz fitting; The determining coefficient R 2 is used for measuring the fitting degree of the fitting curve and the actual data: Where y i is the actual data, i.e., the spectral data after trough finding segmentation, g i is the fitting data, Is the average of the actual data; peak height ratio: Wherein h fitted is the height of the peak after fitting, and h original is the height of the peak in the spectral data after trough searching segmentation; 3) When the determined coefficient R 2 and the peak height ratio H respectively reach the set range, the fitting effect is considered to be good, the peak characteristic parameters including peak height a, peak position b and peak width c are output, otherwise, the step 1) is returned to for the re-identification and peak separation treatment of the shoulder peak until the determined coefficient R 2 and the peak height ratio H respectively reach the set range; 4) And selecting a fitting mode with smaller determining coefficient R 2 and peak height ratio H, and outputting peak characteristic parameters.
5. The intelligent analysis system for the in-situ catalytic reaction infrared spectrum characterization data according to claim 1, wherein the feature extraction module obtains peak areas according to peak feature parameters, and the peak heights, the peak positions, the peak widths and the peak areas are combined to form the spectrum feature parameters for machine learning model identification.
6. The intelligent analysis system for in-situ catalytic infrared spectrum characterization data according to claim 1, wherein the machine learning identification module comprises: The machine learning model comprises a support vector machine, a random forest, a BP neural network and CNN, wherein the machine learning model is used for identifying the chemical structure information of the included catalytic reaction intermediate and the product according to the input spectral characteristic parameters for each model; The model training module is used for carrying out model training on each machine learning model in advance, wherein the input of the model is in-situ infrared spectrum with a label, the label comprises sample information, temperature information and time information, and the output of the model is the contained catalytic reaction intermediate and product chemical structure information; And the model evaluation optimization module is used for evaluating and optimizing the machine learning model and updating the machine learning model.
7. The in situ catalytic infrared spectrum characterization data intelligent analysis system according to claim 1, wherein the data analysis module performs the steps of: 1) Storing the in-situ infrared spectrum of the catalytic reaction into a database one by one, and sorting according to label information; 2) After each infrared spectrum is analyzed, the characteristics of organic functional groups under each reaction temperature condition are determined; 3) According to the temperature change and the time change, the change rule and trend of the organic functional group content in the catalytic reaction process, namely the dynamic change trend of the in-situ reaction product, are respectively and automatically generated.
8. An intelligent analysis method for in-situ catalytic reaction infrared spectrum characterization data is characterized by comprising the following steps: for in-situ infrared spectrum data, carrying out real-time dynamic preprocessing by a preprocessing module; the spectrum data after pretreatment is subjected to feature extraction to obtain spectrum feature parameters; Identifying the spectral characteristic parameters through a machine learning model to obtain chemical structure information of the contained catalytic reaction intermediate and product; and obtaining the dynamic change trend of the in-situ reaction product according to the change of the chemical structure along with the reaction condition.
9. The method for intelligently analyzing the in-situ catalytic infrared spectrum characterization data according to claim 8, wherein the in-situ catalytic infrared spectrum data is subjected to real-time dynamic preprocessing by a preprocessing module, and the method comprises the following steps of: 1) The spectrum data difference subtraction preprocessing module takes an infrared spectrum of a sample initial state before catalytic reaction as a standard curve, takes a difference between spectrum data in the reaction process and the standard curve, and segments a spectrum range according to a data difference subtraction result; 2) The baseline correction module selects different correction models for carrying out baseline removal processing on each spectrum data according to the segmentation result to form corrected spectrum data; 3) The smooth noise reduction module carries out smooth noise reduction on each piece of corrected spectrum data; 4) The wave searching and segmenting module is used for identifying the position of a wave trough point by calculating the first derivative and the second derivative of spectrum data based on the data after smooth noise reduction, and automatically segmenting the spectrum according to the position of the wave trough point; 5) The fitting peak-dividing module carries out peak-dividing fitting on each segment of spectrum data after segmentation and outputs peak characteristic parameters; 6) The loop feedback evaluation module obtains the ratio of the peak area and the peak height of the fitting curve to the original infrared spectrum as a determining coefficient and a peak height weight factor respectively by comparing the characteristic difference of the peak-by-peak fitting curve and the original infrared spectrum, including peak position, peak area and peak characteristic parameters; returning to the step 3) to adjust the size of a filter window in the smoothing noise reduction module when the weight factor obtained by the (i+1) th iteration is still lower than the threshold value; And returning to the step 2) to adjust the parameters of different correction models until the weight factor is not lower than the threshold value when the weight factor obtained by the (i+2) th iteration is still lower than the threshold value.
10. The method for intelligently analyzing the in-situ catalytic reaction infrared spectrum characterization data according to claim 8, wherein the step of identifying the spectral feature parameters through a machine learning model to obtain the chemical structure information of the included catalytic reaction intermediates and products comprises the following steps: the machine learning model comprises a support vector machine, a random forest, a BP neural network and CNN, wherein for each model, the included catalytic reaction intermediate and product chemical structure information is identified according to the input spectral characteristic parameters; The model training module performs model training on each machine learning model in advance, the input of the model is in-situ infrared spectrum with a label, the label comprises sample information, temperature information and time information, and the output of the model is the chemical structure information of the contained catalytic reaction intermediate and product.

Description

In-situ catalytic reaction infrared spectrum characterization data intelligent analysis system Technical Field The invention relates to the technical field of infrared spectrum data intelligent analysis, in particular to a pretreatment and feature identification method for in-situ catalytic reaction infrared spectrum. Background The infrared spectrum analysis technology has wide application in the fields of energy catalysis, biomedicine and the like. At present, the infrared spectrum data analysis mainly depends on expert experience and theoretical calculation, but because the spectrum data contains more information, a plurality of peak positions exist in a specific functional group, the peak positions of similar functional groups coincide, so that the data analysis difficulty is high, and the requirements on physical and chemical knowledge such as molecular vibration theory are higher. For example, carbon monoxide adsorption on the metal surface has complicated conditions such as linear adsorption, bridge adsorption, polyatomic adsorption and the like, and particularly in the complicated in-situ catalytic reaction detection process, the adsorption state is influenced by continuous temperature change and chemical reaction, so that peak positions and peak shapes shift, spectrum base line drift, local noise increase, signal intensity conversion and the like are caused, and the data analysis difficulty is further increased due to the unknown changes. Meanwhile, a large amount of spectrum data can be generated in real time in the in-situ reaction process, and high requirements are placed on timeliness of data analysis. In the field of energy catalysis research, real-time, automatic and intelligent data analysis on the in-situ infrared spectrum of dynamic change is urgently needed. Disclosure of Invention Aiming at the in-situ infrared spectrum real-time data analysis in the energy catalysis research, the invention provides an in-situ infrared spectrum intelligent analysis system with strong self-adaption and error correction capability. Establishing an in-situ infrared spectrum database, analyzing each infrared spectrum in the database separately, preprocessing in-situ infrared spectrum data through a segmentation and difference feedback model, extracting and correcting the characteristics of the infrared spectrum data by using knowledge experience, merging and storing the preprocessed and analyzed spectrum data and original data to form a new database and training a spectrum recognition model, and performing reinforcement training on weak spectrum signals in the in-situ reaction process to further improve the accuracy of data analysis and feature recognition. The technical scheme adopted by the invention for achieving the purpose is that the in-situ catalytic reaction infrared spectrum characterization data intelligent analysis system comprises: the preprocessing module is used for carrying out real-time dynamic preprocessing on the in-situ infrared spectrum data through the preprocessing module; the feature extraction module is used for obtaining spectral feature parameters through feature extraction of the preprocessed spectral data; the machine learning identification module is used for identifying the spectral characteristic parameters through a machine learning model to obtain the chemical structure information of the included catalytic reaction intermediate and the product; And the data analysis module is used for acquiring the dynamic change trend of the in-situ reaction product according to the change of the chemical structure along with the reaction condition. The preprocessing module comprises: 1) The spectrum data difference subtraction preprocessing module is used for making a standard curve for the infrared spectrum of the initial state of the sample before the catalytic reaction, making a difference between the spectrum data in the reaction process and the standard curve, and segmenting the spectrum range according to the data difference subtraction result; 2) The baseline correction module is used for selecting different correction models for carrying out baseline removal processing on each spectrum data according to the segmentation result to form corrected spectrum data; 3) The smooth noise reduction module is used for carrying out smooth noise reduction on each piece of corrected spectrum data; 4) The wave searching and segmenting module is used for identifying the position of a wave trough point by calculating the first derivative and the second derivative of spectrum data based on the data after smooth noise reduction and automatically segmenting the spectrum according to the position of the wave trough point; 5) The fitting peak dividing module is used for carrying out peak dividing fitting on each segment of spectrum data after segmentation and outputting peak characteristic parameters; 6) The loop feedback evaluation module is used for comparing the characteristic difference between the