Solid-liquid two-phase flow chemical component online analysis method based on multi-source synchronous sensing measurement

Abstract

The invention relates to a solid-liquid two-phase flow chemical component online analysis method based on multi-source synchronous sensing measurement, which comprises the following specific steps: step 1, synchronously measuring a solid-liquid mixture modeling sample by multiple sensors, and obtaining multi-source heterogeneous data representing key physical and chemical attributes of the solid-liquid mixture modeling sample; 2, carrying out preprocessing and feature extraction on the multi-source heterogeneous data to obtain new feature vector data; 3, constructing a chemical component content quantitative analysis model based on multi-source information fusion, taking the feature vector as input, taking the known chemical component content of the modeling sample as a label response, and performing iterative fitting prediction training by utilizing a supervised learning method to obtain an ideal model; 4, in actual detection application, multi-source heterogeneous data are collected online through multiple sensors, new feature vector data are obtained through feature extraction in the step 2, the new feature vector data are input into the ideal model for prediction calculation, and the measured value of the chemical component content is obtained. When the method is used for carrying out online analysis on the solid-liquid two-phase flow sample, complex physical and chemical matrix effect influences can be eliminated, and more accurate chemical component content information can be obtained.

Description

Technical Field

[0001] The invention belongs to the field of physics-measurement-automatic analysis, and specifically is an online analysis method of chemical components of solid-liquid two-phase flow based on multi-source synchronous sensing measurement. Background Art

[0002] As a very important material carrier and transportation form, solid-liquid two-phase flow is used in many fields such as mineral mining and processing, hydrometallurgy, petrochemical industry, and food and drug production. The chemical composition content of solid-liquid two-phase flow is an important characterization parameter for studying its internal state, transmission mechanism and change law. Online analysis of chemical composition content is of great significance for production process adjustment and product quality control.

[0003] The chemical composition content of solid phase materials in solid-liquid two-phase flow is currently mainly detected by offline laboratory analysis. In order to achieve high-precision measurement, it is often necessary to separate various interferences in the sample through pre-treatment such as sampling, separation, drying, and grinding to eliminate the influence of physical and chemical matrices. However, the complex pre-treatment and analysis process consumes a lot of time, resulting in the chemical composition information obtained seriously lagging behind the production process and cannot be used to guide process adjustments and quality control. Online chemical composition analysis technology based on spectral information is a hot field that has emerged in recent years. However, the online detection of solid-liquid two-phase flow spectra is affected by complex physical and chemical matrix effects, and the quantitative analysis accuracy in practical applications is not yet sufficient to guide production.

[0004] Therefore, more accurate chemical composition analysis methods are needed to realize online detection of solid-liquid two-phase flow in the fields of mineral processing, pharmaceuticals, environmental engineering, etc., and provide digital and information support for refined modeling and control of production processes, process innovation, and industrial upgrading. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to solve the problem that during the on-line detection of solid-liquid two-phase flow, the chemical composition analysis accuracy of the solid phase material cannot meet the application requirements due to the influence of changes in complex physical and chemical matrices.

[0006] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:

[0007] The solid-liquid two-phase flow chemical composition online analysis method based on multi-source synchronous sensing measurement establishes an accurate chemical composition content measurement model through the following measurement and analysis steps, which is used for online real-time analysis and measurement of different solid-liquid mixed substances. The method includes the following steps:

[0008] Step 1: Multi-sensor synchronous measurement of solid-liquid mixture modeling samples to obtain multi-source heterogeneous data characterizing its key physicochemical properties;

[0009] Step 2: Preprocess and extract features of multi-source heterogeneous data to obtain new feature vector data;

[0010] Step 3: Construct a quantitative analysis model of chemical component content based on multi-source information fusion, taking the feature vector as input and the known chemical component content of the modeling sample as the label response, and use the supervised learning method to perform iterative fitting prediction training to obtain an ideal model;

[0011] Step 4: In actual detection applications, multi-source heterogeneous data are collected online through multiple sensors, and feature extraction in step 2 is performed to obtain new feature vector data, which is input into the ideal model for predictive calculation to obtain the measured value of the chemical component content.

[0012] The multi-source heterogeneous data are I, D, and P; wherein I is spectral data, D is particle size information data, and P is other physical parameter data; the number of types of I, D, and P is a natural number, and the number of types of I and D is at least 1, and the number of types of P is at least 0.

[0013] When the type of spectral data I is 1, the spectral data is a laser induced breakdown spectrum; when the type of spectral data I is greater than 1, the spectral data is a combination of laser induced breakdown spectrum and Raman spectrum, X-ray fluorescence spectrum, and infrared absorption spectrum.

[0014] The particle size information data is the particle size value data obtained by directly detecting the solid-liquid mixture through a particle size sensor, or is multi-dimensional indirect information data that can indirectly reflect the particle size value.

[0015] The multi-dimensional indirect information data indirectly reflecting the particle size value is collected in any of the following ways: particle size distribution statistics obtained by a distance sensor; particle size distribution data obtained by an image sensor; particle size statistics obtained by laser scattering; particle size distribution data obtained by an ultrasonic sensor.

[0016] Each sensor needs to perform multiple measurements on the same sample and complete basic data preprocessing by calculating the average value of multiple measurements.

[0017] Feature extraction is achieved in any of the following ways:

[0018] a. Direct concatenation of original multi-source heterogeneous data, Features = [I, D, P];

[0019] b. Extract features from the original multi-source heterogeneous data and then splice them according to the sensor source, Features = [f Pre1 (I), f Pre (D), f Pre3 (P)];

[0020] c. Treat all original multi-source heterogeneous data as a whole and extract features uniformly, Features = f Pre (I, D, P);

[0021] where f Pre (·),f Pre1 (·),f Pre2 (·),f Pre (·) is the feature extraction function, and Features is the feature vector data after feature extraction.

[0022] The modeling process of the quantitative analysis model is a supervised learning process that takes the characteristic vector data Features of the spectrum, particle size, and other physical parameters of multiple modeling samples as input and the chemical composition content C as response. It is a process of determining the specific form and parameters of g(·) in C=g(Features) through reverse conduction iterative training.

[0023] The acquisition of laser-induced breakdown spectroscopy involves the following steps:

[0024] The solid-liquid mixture is made to form a vertically downward solid-liquid two-phase flow through a sampling device;

[0025] Use high-energy pulsed laser to vertically strike solid-liquid two-phase flow to generate plasma;

[0026] The emission spectrum of the plasma is collected by a spectrometer.

[0027] The present invention has the following advantages and beneficial effects:

[0028] 1. The chemical composition online analysis method proposed in the present invention can select different spectral detection, particle size analysis and physical parameter measurement technologies to obtain raw data according to the application scenario and chemical composition analysis requirements, and then perform data preprocessing and modeling analysis. It is widely applicable to various solid-liquid two-phase flow online analysis fields.

[0029] 2. The online analysis method of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement proposed in the present invention can eliminate the complex influence of physical and chemical matrix changes of solid-liquid two-phase flow on spectral detection during the online analysis process, and effectively improve the quantitative analysis accuracy of chemical composition content. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the on-line analysis method of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement;

[0031] Figure 2 Implement a flow chart for the example;

[0032] Figure 3 This is a schematic diagram of the application effect of the example. DETAILED DESCRIPTION

[0033] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation method of the present invention is described in detail below by taking a quantitative analysis process of iron ore slurry grade as an example. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the invention, so the present invention is not limited to the specific implementation disclosed below.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0035] like Figure 1 , an online analysis method of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement, comprising the following steps:

[0036] Step 1: Synchronously obtain the characterization information of the spectrum, particle size, concentration or other key physical parameters of the solid-liquid mixture modeling sample through multiple sensors such as spectrum and particle size, and obtain multi-source heterogeneous data characterizing the key physical and chemical properties of the solid-liquid mixture for modeling;

[0037] The modeling samples are multiple solid-liquid two-phase flow samples with certain differences in chemical composition and known specific content information.

[0038] Among them, synchronously acquiring characterization information through multiple sensors such as spectrum and particle size means using different sensing technologies to directly detect solid-liquid two-phase flow samples to obtain data without the need for complex sample preparation and offline analysis.

[0039] Among them, the multi-source heterogeneous original modeling data includes at least one spectral data and one sensor data that can characterize the particle size information. The characterization information of other physical parameters needs to be introduced based on the interference to the results.

[0040] Among them, the spectral characterization information can be a single atomic or molecular spectrum, or multiple spectra obtained by synchronous measurement. When only laser-induced breakdown spectroscopy is included, the acquisition of the spectrum includes the following steps: using a sampling device such as a sampler to form a vertical downward flow column of the solid-liquid mixture; using a high-energy pulsed laser to hit the slurry flow column multiple times to form a plasma; collecting the emission spectrum of the plasma; after the sample measurement is completed, the slurry channel is cleaned with clean water, and the next sample measurement is switched after cleaning.

[0041] The particle size characterization information is obtained by any of the following methods: particle size distribution statistics obtained by a distance sensor; particle size distribution data obtained by an image sensor; particle size statistics obtained by laser scattering; and particle size distribution data obtained by an ultrasonic sensor.

[0042] Among them, other physical parameters may include one or more of the physical quantities that can be measured and quantified online, such as solid phase concentration in solid-liquid two-phase flow, fluid viscosity, color, etc., or may be omitted.

[0043] Step 2: Preprocess and extract features of multi-source heterogeneous data to obtain new feature information;

[0044] Data preprocessing and feature extraction can be achieved by any of the following methods: direct concatenation of the original data, that is, Features = [I, D, P]; feature extraction and concatenation based on the sensor source, that is, Features = [f Pre1 (I), f Pre (D), f Pre3 (P)]; all the original data are treated as a whole and feature extraction is performed uniformly, that is, Features = f Pre (I, D, P). Where I is the original spectral data, D is the data representing the particle size information, and P is other data. Pre (·),f Pre (·),f Pre2 (·),f Pre3 (·) is the feature extraction function, and Features is the feature information after feature extraction.

[0045] Step 3: Using the processed new feature information as input and the known chemical component content reference labels of the modeled samples as responses, a quantitative analysis model of chemical component content based on multi-source information fusion is established;

[0046] Among them, the chemical component reference label only needs to contain the content information of the chemical component to be predicted by quantitative analysis, and does not need to contain the composition information of all chemical components in the sample.

[0047] The modeling process of the quantitative analysis model is a supervised learning process that takes the characteristic information Features such as spectrum and particle size of multiple modeling samples as the input matrix and the content information C of the chemical components as the response matrix. It is a process of determining the specific structure and parameters of g(·) in C=g(Features) through training.

[0048] Step 4: Using the same sensing technology as in step 1 and under the same measurement conditions as in step 1, the spectrum, particle size, concentration or other physical parameters of the sample to be analyzed are measured online synchronously to obtain online measurement raw data;

[0049] Step 5: Use the same data preprocessing and feature extraction methods as in step 2 to process the online measurement raw data; the extracted features are determined by the data preprocessing and feature extraction model established in step 2. The structure and parameters of the model are determined in the offline modeling process. Repeated calculation and updating are not required in the online analysis process to ensure the consistency of the dimension and range of the input data of the quantitative analysis model.

[0050] Step 6: Input the characteristic information obtained in step 5 into the quantitative analysis model established in step 3 for calculation to obtain the measured value of the chemical composition content. The quantitative analysis model, structure and parameters are determined in the offline modeling process, and there is no need to repeat the calculation and update in the online analysis process. The calculated chemical composition content prediction value C Prt =g(Feature OL ), where Feature OL It is the data feature of the online measurement data after processing.

[0051] Example: Online quantitative analysis of iron ore grade based on LIBS spectroscopy, direct diameter measurement and concentration pot.

[0052] Online quantitative analysis process of iron ore grade based on LIBS spectroscopy, direct diameter measurement and concentration pot, such as Figure 2 As shown, the specific implementation steps are:

[0053] (1-1) Import n slurry training samples with known total iron content into the LIBS slurry grade analyzer, perform LIBS measurement, and obtain the original spectral data Spc n×p (p is the spectral pixel dimension);

[0054] (1-2) Import the slurry training sample into the particle size analyzer to measure the particle size and obtain the particle size distribution data D n×q (q is the number of sampling points of particle size distribution);

[0055] (1-3) Import the slurry training sample into the concentration pot and measure the slurry mass M n×1 ;

[0056] (2-1) Identify the characteristic spectral lines of key elements such as Fe and Si in the LIBS spectrum and extract the characteristic spectral line intensity I s_n×k (k is the number of selected characteristic spectral lines);

[0057] (2-2) Calculate the median particle size D based on the particle size distribution data 50_n×1 ;

[0058] (2-3) The measured quality data M n×1 , concentration pot volume V and mineral density ρ n×1 Bring in

[0059]

[0060] Calculate the slurry concentration H n×1 ;

[0061] (2-4) The characteristic spectral line intensity I of all slurry training samples is s_n×k , median particle size D 50_n×1 and concentration H n×1 Splice into feature data matrix Features n×(k+2) ;

[0062] (3) Feature data matrix Features n×(k+2) As input, the total iron content C TFe_n×1 For the response, train the PLS regression model and obtain the regression coefficient matrix β (k+3)×1 , complete offline modeling;

[0063] (4) In the online analysis, the original LIBS spectrum Spc of the slurry sample to be tested is measured in the same measurement method as (1-1) to (1-3). 1×p , particle size distribution D 1×q and mass M;

[0064] (5) According to the feature extraction and calculation methods of (2-1) to (2-3), the characteristic spectral line intensity I of the sample to be tested is obtained. s1×k , median particle size D 50 and slurry concentration H, and concatenate the feature data vector Features according to (2-4). 1×(k+2) ;

[0065] (6) According to

[0066]

[0067] Calculate total iron content C TFe_Prt , where [β0, β1,…,β k+2 ] T =β (k+3)×1 , the output is the grade prediction value of online analysis.

[0068] The grade of 50 iron ore flotation concentrate samples was analyzed online according to the above method. The analysis results were compared with the traditional quantitative analysis results based on LIBS spectrum. Figure 3 The online quantitative analysis of iron ore grade based on LIBS spectroscopy, direct diameter measurement and concentration pot method obtained a higher determination coefficient (R) than the traditional LIBS top two analysis. 2 ), lower root mean square error (RMSE) and mean absolute error (MAE), indicating that the method of this patent invention has higher quantitative analysis accuracy than traditional online analysis methods.

[0069] The present invention is described above by way of example in conjunction with the accompanying drawings. Obviously, the present invention is not limited to the above embodiments, and may also have many variations. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, or the concept and technical solution of the invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. An online analysis method for chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement, characterized in that: Through the following measurement and analysis steps, an accurate chemical composition content measurement model is established for online real-time analysis and measurement of different solid-liquid mixtures. The method includes the following steps: Step 1: Multi-sensor synchronous measurement of solid-liquid mixture modeling samples to obtain multi-source heterogeneous data characterizing its key physicochemical properties; Step 2: Preprocess and extract features of multi-source heterogeneous data to obtain new feature vector data; Step 3: Construct a quantitative analysis model of chemical component content based on multi-source information fusion, taking the feature vector as input and the known chemical component content of the modeling sample as the label response, and use the supervised learning method to perform iterative fitting prediction training to obtain an ideal model; Step 4: In actual detection applications, multi-source heterogeneous data are collected online through multiple sensors, and feature extraction in step 2 is performed to obtain new feature vector data, which is input into the ideal model for predictive calculation to obtain the measured value of the chemical component content.

2. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 1 is characterized in that: The multi-source heterogeneous data are I, D, and P; Where I is spectral data, D is particle size information data, and P is other physical parameter data; the number of I, D, and P types is a natural number, and the number of I and D types is at least 1, and the number of P types is at least 0.

3. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 2 is characterized in that: When the type of spectral data I is 1, the spectral data is a laser induced breakdown spectrum; when the type of spectral data I is greater than 1, the spectral data is a combination of laser induced breakdown spectrum and Raman spectrum, X-ray fluorescence spectrum, and infrared absorption spectrum.

4. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 2 is characterized in that: The particle size information data is the particle size value data obtained by directly detecting the solid-liquid mixture through a particle size sensor, or is multi-dimensional indirect information data that can indirectly reflect the particle size value.

5. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 4 is characterized in that: The multi-dimensional indirect information data indirectly reflecting the particle size value is collected in any of the following ways: particle size distribution statistics obtained by a distance sensor; particle size distribution data obtained by an image sensor; particle size statistics obtained by laser scattering; particle size distribution data obtained by an ultrasonic sensor.

6. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 1, characterized in that: Each sensor needs to perform multiple measurements on the same sample and complete basic data preprocessing by calculating the average value of multiple measurements.

7. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 1, characterized in that: Feature extraction is achieved in any of the following ways: a. Direct concatenation of original multi-source heterogeneous data, Features = [I, D, P]; b. Extract features from the original multi-source heterogeneous data and then splice them according to the sensor source, Features = [f Pre1 (I),f Pre2 (D),f Pre3 (P)]; c. Treat all original multi-source heterogeneous data as a whole and extract features uniformly, Features = f Pre (I,D,P); where f Pre (·),f Pre1 (·),f PreP (·),f Pre3 (·) is the feature extraction function, and Features is the feature vector data after feature extraction.

8. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 1, characterized in that: The modeling process of the quantitative analysis model is a supervised learning process that takes the characteristic vector data Features of the spectrum, particle size, and other physical parameters of multiple modeling samples as input and the chemical composition content C as response. It is a process of determining the specific form and parameters of g(·) in C=g(Features) through reverse conduction iterative training.

9. The method for online analysis of chemical composition of solid-liquid two-phase flow based on multi-source synchronous sensing measurement according to claim 3, characterized in that: The acquisition of laser-induced breakdown spectroscopy involves the following steps: The solid-liquid mixture is made to form a vertically downward solid-liquid two-phase flow through a sampling device; Use high-energy pulsed laser to vertically strike solid-liquid two-phase flow to generate plasma; The emission spectrum of the plasma is collected by a spectrometer.

Images