CN114117954B - Dynamic real-time visualization method for three-dimensional reaction field in reactor - Google Patents

Dynamic real-time visualization method for three-dimensional reaction field in reactor Download PDF

Info

Publication number
CN114117954B
CN114117954B CN202111316547.9A CN202111316547A CN114117954B CN 114117954 B CN114117954 B CN 114117954B CN 202111316547 A CN202111316547 A CN 202111316547A CN 114117954 B CN114117954 B CN 114117954B
Authority
CN
China
Prior art keywords
reactor
data
dimensional
reaction
model
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.)
Active
Application number
CN202111316547.9A
Other languages
Chinese (zh)
Other versions
CN114117954A (en
Inventor
杨宏燕
韩华云
韩红桂
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN202111316547.9A priority Critical patent/CN114117954B/en
Publication of CN114117954A publication Critical patent/CN114117954A/en
Application granted granted Critical
Publication of CN114117954B publication Critical patent/CN114117954B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Molecular Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Algebra (AREA)
  • Fluid Mechanics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Physical Or Chemical Processes And Apparatus (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a dynamic real-time visualization method for a three-dimensional reaction field in a reactor. Based on an operation database access module, historical operation data and a real-time operation database of the reactor are constructed; determining typical reference working conditions of the reactor based on a typical working condition analysis module; based on a numerical simulation module, performing off-line three-dimensional CFD numerical simulation calculation of a reactor internal reaction field under a reference operation condition; based on a data clustering and standardization module, dividing a data set of the established reactor CFD three-dimensional simulation database by adopting a data mining and principal component analysis method, processing parameters, integrating redundancy, reducing dimension, and obtaining 8-10 principal components after processing; and calculating to obtain the three-dimensional reaction field data of the reactor under the full-operation working condition by adopting a regression prediction algorithm and a neural network model based on a deep learning prediction module. The invention solves the problem of 'black box' in the reactor in the field of petrochemical industry, and accurately optimizes the operation and intelligently regulates and controls the operation process of the reactor.

Description

Dynamic real-time visualization method for three-dimensional reaction field in reactor
Technical Field
The invention relates to a dynamic real-time visualization method for a three-dimensional reaction field in a reactor, which is suitable for the crossing field of petrochemical reactor three-dimensional visualization and process control.
Background
The reactor in the fields of petrochemical industry and the like is a complex nonlinear chemical reaction system and is often used for meeting various requirements of synthetic materials, medical reagents, dyes and the like. Because the reaction conditions inside the reactor directly determine the quality and performance of reaction products, but the reaction process is complex, most of the reactions are carried out in a high-temperature and high-pressure environment, three-dimensional reaction field information including flow fields, gas-phase reactant concentrations, particle concentration distribution and the like is difficult to obtain, and intelligent control and optimal adjustment aiming at the reaction process cannot be realized.
In order to obtain multi-field coupling data inside the reactor and ensure that the flow and the reaction process inside the reactor are within a controllable range, reaction field information inside the reactor needs to be accurately known, but because complex operating conditions such as tightness, high temperature and high pressure, low visibility and the like exist inside the reactor, the traditional measuring methods such as endoscopes and component concentration measuring sensors have limited measurable parameters and area ranges, the measuring precision and the measuring efficiency are required to be improved, the influence of human factors is large, and accurate three-dimensional reaction field information is difficult to obtain. In recent years, data-driven methods represented by various machine learning methods are applied to data prediction of industrial processes, but conventional deep learning prediction methods based on neural networks are often highly dependent on input of historical operating data sets, neglect reaction mechanism models and have inherent defects in the aspects of predictable data types and accuracy. With the continuous maturity of the three-dimensional numerical simulation (CFD) technology, the actual reaction mechanism can be coupled, and the flow, heat transfer and chemical reaction processes in the reactor can be accurately calculated and obtained.
Aiming at solving the problems, the invention provides a dynamic real-time visualization method for a three-dimensional reaction field in a reactor, aiming at solving the problem of a black box in the reactor commonly existing in the field of petrochemical industry and meeting the requirement of acquiring the information of complex flow and reaction process in the reactor in real time.
Disclosure of Invention
The invention provides a dynamic real-time visualization method for a three-dimensional reaction field in a reactor, aiming at the problem of 'black boxes' in the reactor commonly existing in the field of petrochemical industry and aiming at meeting the requirement of obtaining complex flow and reaction process information in the reactor in real time.
A dynamic real-time visualization method for a three-dimensional reaction field in a reactor comprises the following modules:
the system comprises an operation database access module, a typical working condition analysis module, a numerical simulation module, a data clustering and standardization module, a deep learning prediction module and a three-dimensional visualization module which are totally six modules;
the operation database access module, the typical working condition analysis module and the numerical simulation module are all offline modules; the data clustering and standardizing module, the deep learning prediction module and the three-dimensional visualization module are all online modules;
the method comprises the following specific steps:
the method comprises the following steps: historical operation data and real-time operation database of reactor are constructed based on operation database access module
a) Extracting historical operation data of the reactor from a historical operation database of the reactor according to a certain time interval delta t, wherein the historical operation data comprises reactor inlet parameters, reactor internal monitoring parameters and reactor outlet parameters;
reactor inlet parameters, including the composition of the reactor inlet feed, the flow rate Q of the inlet feed I Inlet feed velocity v I Temperature T of the inlet feed I Wherein, I represents different inlet reactant species;
reactor internal monitoring parameters, including temperature value T of the reaction zone x,y,z Pressure value P x,y,z Gas component concentration C x,y,z Wherein x, y and z represent the three-dimensional space coordinate position in the reactor under a Cartesian coordinate system;
reactor outlet parameters including gas flow Q at the reactor outlet O Average temperature T O Average pressure P O ；
b) Extracting real-time operation data of the reactor according to a certain time interval delta t in the operation process of the reactor, and analyzing the working condition;
reactor inlet parameters including composition of reactor inlet feed, inlet feedFlow rate Q of I Velocity v of the inlet feed I Temperature T of the inlet feed I Wherein I represents a different inlet reactant type;
reactor internal monitoring parameters, including temperature value T of the reaction zone x Pressure value P x Gas component concentration C x Wherein x, y and z represent the three-dimensional space coordinate position in the reactor under a Cartesian coordinate system;
reactor outlet parameters including gas flow Q at the reactor outlet O Average temperature T O Average pressure P O ；
c) And storing historical operating data and real-time operating data of the reactor in a database, constructing the operating database of the reactor, removing data abnormal points and carrying out data standardization processing. The method comprises the following specific steps:
processing by using a K-means algorithm to reduce data dimensionality; judging whether the data contains null values, 0 values or abnormal values, and if so, filling the data by using a Lagrange interpolation method; then, the data is normalized. Wherein the normalization processing formula is expressed as:
Figure BDA0003343843930000031
in the formula (I), the compound is shown in the specification,
Figure BDA0003343843930000032
representing a production parameter of the kth sample normalized in n-dimensional data>
Figure BDA0003343843930000033
N-dimensional metadata representing that the kth sample is arranged in time series, K is the number of data sets, based on>
Figure BDA0003343843930000034
N-dimensional data representing the kth sample, k =1,2, …, N being the number of datasets;
step two: determining typical reference conditions of a reactor based on a typical condition analysis module
a) Because mass data exist in the actual operation of the reactor, the number of the operation working conditions is huge, and therefore representative reference operation working condition data in the reactor needs to be selected preferably. Specifically, calculating to obtain data of a reference operation condition in the reactor operation process by adopting a K-means mean value clustering algorithm in the reactor operation database constructed in the step one;
step three: based on a numerical simulation module, off-line three-dimensional CFD computational fluid dynamics numerical simulation computation of a reactor internal reaction field under a standard operation condition is carried out
a) Calculating and solving the chemical reaction process among the materials by using Chemkin software according to the components of the materials at the inlet of the reactor and the chemical reaction conditions to obtain the reaction composition of each element, the corresponding chemical equation and the reaction kinetic parameters of the element reaction in each step;
b) Establishing a three-dimensional calculation domain physical model of the reactor according to the operation process, structure and size parameters of the reactor;
c) Adopting gambit computational domain meshing software to perform meshing on a three-dimensional computational domain model of the reactor, and performing local encryption on meshes on an inlet, an outlet and a structurally complex region of the reactor to obtain a high-precision simulation result; setting boundary condition types of an inlet, an outlet and a wall surface of the reactor, wherein the boundary condition of the inlet of the reactor is a mass inlet, the boundary condition of the outlet of the reactor is a pressure outlet, and the boundary condition of the wall surface of the reactor is a constant temperature wall surface;
d) Setting inlet feeding components and inlet feeding flow Q of the three-dimensional calculation domain model of the reactor according to the reactor reference working condition operation data obtained in the step two I Inlet feed velocity v I Temperature T of the feed I Wherein I represents a different inlet reactant type;
e) Selecting a turbulent flow model, a multiphase flow model and a radiation and reaction mechanism model by using a high-performance computer and a CFD numerical simulation platform and adopting the chemical elementary reaction obtained in the step three a), and performing reaction by using the reactor computational domain grid model obtained in the step three c)Solving and simulating the three-dimensional chemical reaction process in the reactor, and simulating to obtain the speed (v) corresponding to the coordinate (x, y, z) of the three-dimensional space in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The distribution data of (2); wherein, the turbulent flow model, the multiphase flow model, the radiation model and the reaction mechanism model are respectively expressed as:
in a turbulent flow model, a k-epsilon two-pass model is used for simulating a turbulent flow process in a reactor;
in the multiphase flow model, a Mixture model based on an Euler-Euler method is adopted, wherein a continuity equation and a momentum equation are respectively expressed as follows:
Figure BDA0003343843930000041
Figure BDA0003343843930000042
in the formula, ρ m As density, v is the Hamiltonian, v m In order to be the mass-average velocity,
Figure BDA0003343843930000043
in the form of a vector of mass-averaged velocities, μ m Is the viscosity coefficient of the mixture, T is the temperature of the phase, is based on>
Figure BDA0003343843930000044
Representing the form of a mass-averaged velocity vector at a temperature T, F being the volume force, <' > based on>
Figure BDA0003343843930000045
In the form of a vector of volume forces, B is the total number of phases, k is the kth phase, g m Is based on gravity acceleration>
Figure BDA0003343843930000046
As acceleration of gravityVector form, α k Volume fraction of the k-th phase, p k Is the density of the k-th phase, v dr,k Is the slip speed of the k-th phase>
Figure BDA0003343843930000047
In the form of a slip velocity vector of the k-th phase;
in the radiation model, P-1 radiation model is used to simulate radiation heat transfer, and the model also includes the mutual radiation among ions and describes radiation flux
Figure BDA0003343843930000048
The equation of (c) can be expressed as follows:
Figure BDA0003343843930000049
wherein beta is the absorption coefficient, C s Is the scattering coefficient, h is the coefficient of the linear anisotropy phase function, v is the hamiltonian, G is the incident radiation;
in the reaction mechanism model, the chemical reaction rate constants follow the arrhenius formula and are expressed as follows:
Figure BDA0003343843930000051
in the formula, k a For reaction rate constant, Z is the molar gas constant, T a Is a thermodynamic temperature, Z a Is the apparent activation energy, A is the pre-exponential factor, e is the natural constant;
f) Based on the CFD three-dimensional numerical simulation result obtained in the step three e), performing data verification and CFD calculation model optimization by adopting the reactor operation database constructed in the step one until the calculation error of the CFD three-dimensional numerical simulation result is less than 5%;
g) Repeating the step three a) to the step three f), calculating to obtain all CFD three-dimensional numerical simulation results of all working conditions in the reactor reference operation working condition, and establishing a simulation database of the reactor reference operation working condition reaction process;
step four: based on a data clustering and standardization module, a reactor CFD three-dimensional simulation database established in the third step is subjected to data set division, parameter processing, integration redundancy and dimensionality reduction by adopting a data mining and principal component analysis method, and 8-10 principal components are obtained after processing
a) Aiming at a CFD three-dimensional simulation database of reactors, in order to reduce the number of subsequent data modeling, the data quantity randomly extracted in the reference operation condition of each reactor is 2-3 ten thousand groups, and in the data extraction process, the data extraction coordinates of each working condition are ensured to be the same;
b) The extracted CFD simulation data information includes: the coordinate position (x, y, z) of each calculation node in the reactor and the fluid velocity (v) corresponding to each node x ，v y ，v z B), rotation phi x,y,z Reaction product component concentration C x,y,z Temperature T x,y,z And pressure T x,y,z ；
c) Because the data extraction amount of each working condition is huge, the direct grouping use can cause the problems of low prediction precision and long time, therefore, in order to improve the calculation precision and speed of the full operation working condition of the reactor, a data mining and principal component analysis method is adopted to carry out correlation analysis on the data; specifically, a K-means mean clustering algorithm and a dimensionality reduction algorithm are combined to divide a data set and process parameters so as to integrate redundant data and reduce dimensionality of the data, so that a high-efficiency data set is obtained, and main parameters influencing the chemical reaction process in the reactor are obtained. Obtaining 8 to 10 main components after data processing;
step five: based on a deep learning prediction module, three-dimensional reaction field data under the full-operation working condition of the reactor are obtained by adopting a regression prediction algorithm and a neural network model for calculation
a) Combining a regression prediction algorithm and a neural network model to establish a main parameter rapid prediction model of the internal reaction process of the reactor, wherein the parameters comprise: velocity (v) corresponding to three-dimensional space arbitrary position coordinate (x, y, z) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z Distribution data of. The neural network model adopts a BP neural network model, the regression prediction algorithm adopts a lasso regression prediction algorithm, and in the lasso regression prediction algorithm, the loss function of the lasso model is assumed to be as follows:
Figure BDA0003343843930000061
wherein J (β) is a loss function, β j The regression parameter is s, the total data quantity is s, the characteristic variable is L, the calculated data quantity is p, the actual output is G, and the regular coefficient is lambda; beta is a j The partial derivatives of the first part are solved by using a sub-gradient method to obtain:
Figure BDA0003343843930000062
optimizing and continuously iterating each time to obtain an optimal regression coefficient;
in the BP neural network model, the output layer is the speed (v) corresponding to the reaction field parameters (including the coordinates (x, y, z) of any position in three-dimensional space) in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z ) Recording the number of nodes of each hidden layer, limiting the number of nodes of an input layer and the number of iterations, defining the weight from the input layer to the hidden layer and the weight from the hidden layer to the output layer, the threshold value of the hidden layer and the threshold value of the output layer, and setting the learning rate and the minimum error, wherein the learning rate is set to be in the range of 0.1 to 0.6, and the minimum error is 1%;
b) Taking the inlet parameters of the reactor in the database as an input data set based on the database of the reference operation conditions of the reactor obtained in the fourth step, wherein the inlet parameters comprise: composition of inlet feed, flow rate of inlet feed Q I Velocity v of the inlet feed I Temperature T of the inlet feed I (ii) a Training a neural network model by using 8 to 10 principal components obtained by data processing as an output data set, wherein the principal components comprise arbitrary bits in a three-dimensional spaceVelocity (v) corresponding to position coordinate (x, y, z) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z ；
c) The above processes are circulated until the error meets the precision requirement or the learning times is more than the set maximum times, and the training process is ended; the accuracy requirement is that the error of a predicted value is within 5 percent, and the maximum learning frequency to be set is 1000 times;
d) And adopting the trained neural network model to carry out three-dimensional reaction field prediction under all working conditions of the reactor, and specifically taking the difference value between the inlet parameter of the reactor under the atypical working condition to be predicted and the reference operation working condition as the input variable of the neural network prediction model. Taking the principal component difference value of the non-reference operation condition and the reference operation condition as an output variable of a prediction model, and carrying out prediction on principal components in a reaction field;
step six: based on a three-dimensional visualization module, three-dimensional visualization under full-operation working condition of the reactor is realized
a) Acquiring a reactor three-dimensional reaction field database based on the reactor three-dimensional reaction field data under the non-reference operation condition obtained by model prediction in the fifth step;
b) Integrating the three-dimensional reaction field database under the non-reference operation condition with the three-dimensional reaction field database under the reference operation condition to form a three-dimensional reaction field database under the full operation condition;
c) Based on a three-dimensional reaction field database under the full operation condition, drawing a three-dimensional reaction field image by using OpenGL software, and constructing a speed (v) corresponding to any position coordinate (x, y, z) in a three-dimensional space x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The three-dimensional distribution cloud picture and the numerical scale are displayed on a computer-side software interface in real time, and are displayed in a three-dimensional visual mode for operators of the reactor to visually observe and check;
according to the three-dimensional visualization result of the reactor, the operating personnel of the reactor can visually know the distribution conditions of the eddy current distribution, the high-temperature area position and the insufficient reaction area in the reactor in real time, so that the operating personnel can conveniently adjust the operating conditions of the reactor in a targeted manner, and the required high-concentration and high-quality reaction product can be obtained while the safe and reliable operation of the reactor is realized.
According to the dynamic real-time visualization method for the three-dimensional reaction field in the reactor, the three-dimensional reaction field in the reactor is established by the system, the three-dimensional real-time monitoring of the reaction in the reactor is realized, the three-dimensional reaction field is displayed on a visualization interface in real time, the state of a black box in the reactor is broken, a technical method for knowing and visualizing the state in the reactor is provided for the technology and operating personnel of the reactor, and the intelligent control and optimized operation capability of the reaction process are improved.
The beneficial effects of the invention are as follows:
(1) Based on chemical reaction dynamics simulation, and in combination with a computational fluid dynamics three-dimensional numerical simulation method, numerical simulation is carried out on three-dimensional flow, heat transfer and chemical reaction processes in the reactor, so that three-dimensional reaction field parameter distribution of any position of the internal space of the reactor under a typical reference operation condition is obtained, and a three-dimensional reaction field database under a typical operation condition based on a reaction mechanism model is provided;
(2) By combining a numerical simulation database based on a reaction mechanism model, a historical operation database of the reactor and real-time operation data of the reactor, the prediction of the reaction field of the reactor under the full operation condition is realized by utilizing a neural network algorithm, so that a large number of working conditions, operation steps and calculation time generated by the traditional CFD three-dimensional numerical calculation are saved, and high-precision three-dimensional reaction field data under the full operation condition of the reactor can be obtained in real time;
(3) Based on the obtained reaction field data under the full-operation working condition of the reactor, the reaction field under the full-operation working condition of the reactor can be displayed on a computer-end display in real time through visual software, three-dimensional visualization is realized, the technology and operators of the reactor are facilitated, the cause analysis is carried out on the problems occurring in the operation process of the reactor, a solution is provided, meanwhile, the problems about to occur in the operation process of the reactor are predicted, and a corresponding optimization control scheme is proposed in advance.
Drawings
FIG. 1 is a logic framework diagram of a dynamic real-time visualization method of a three-dimensional reaction field in a reactor according to the present invention.
Fig. 2 is a schematic diagram of a neural network of a dynamic real-time visualization method of a three-dimensional reaction field in a reactor according to the present invention.
FIG. 3 is a logic control diagram of a neural network for a dynamic real-time visualization method of a three-dimensional reaction field inside a reactor according to the present invention.
FIG. 4 shows that the method of the present invention is applied to a coal gasifier to predict the three-dimensional axial velocity v in the gasifier under the condition of changing the incident angle of air distribution z And (4) distributing cloud pictures.
FIG. 5 shows that the method of the present invention is applied to a coal gasifier to predict the three-dimensional temperature T in the gasifier under the condition of changing the incident angle of air distribution x,y,z And (4) distributing cloud pictures.
Detailed Description
In order to clarify the objects, technical solutions and advantages of the present invention, the following embodiments further illustrate the method for dynamically visualizing the three-dimensional reaction field inside the reactor in real time. It should be noted that the specific embodiments described herein are only for explaining the present invention and do not limit the present invention.
In order to explain the embodiments of the present invention, a specific embodiment of the present invention will be described below by taking a coal gasifier as an example.
As shown in fig. 1, a method for dynamically visualizing a three-dimensional reaction field inside a reactor in real time includes the following modules:
the system comprises an operation database access module, a typical working condition analysis module, a numerical simulation module, a data clustering and standardization module, a deep learning prediction module and a three-dimensional visualization module, wherein the total six modules are provided;
the operation database access module, the typical working condition analysis module and the numerical simulation module are all offline modules; the data clustering and standardizing module, the deep learning prediction module and the three-dimensional visualization module are all online modules;
the method comprises the following specific steps:
as shown in fig. 1, step one: historical operation data and real-time operation database of coal gasifier reactor are constructed based on operation database access module
a) Extracting historical operation data of a reactor from a historical operation database of the coal gasifier reactor according to a certain time interval delta t =10s, wherein the historical operation data comprises coal gasifier reactor inlet parameters, coal gasifier reactor internal monitoring parameters and coal gasifier reactor outlet parameters;
coal gasifier reactor inlet parameters, including the composition of the reactor inlet feed (including air, CH) 4 Pulverized coal), inlet feed flow Q I (including air flow, steam flow, pulverized coal feed flow), velocity of inlet feed v I (including air incidence velocity, steam incidence velocity, pulverized coal incidence velocity), inlet feed temperature T I (including air temperature, steam temperature, coal dust temperature), wherein I represents different inlet reactant species;
internal monitoring parameters of the coal gasifier reactor, including the temperature value T of the violent reaction zone x,y,z Pressure value P x,y,z Gas component concentration C x,y,z (including CO, H) 2 、CO 2 、N 2 、H 2 O、H 2 S and CH 4 Concentration), wherein x, y, z represent the three-dimensional space coordinate position of the reactor interior under a Cartesian coordinate system;
coal gasifier reactor outlet parameters, including gas flow Q at the reactor outlet O (mainly CO flux and H) 2 Flow rate), average temperature T O Average pressure P O ；
b) Extracting real-time operation data of a reactor according to a certain time interval delta t =10s in the operation process of the coal gasifier flute reactor, and analyzing working conditions;
coal gasifier reactor inlet parameters, including the composition of the reactor inlet feed (including air, CH) 4 Pulverized coal), inlet feed flow Q I (including air flow, steam flow, feedPulverized coal flow), velocity v of inlet feed I (including air incidence velocity, steam incidence velocity, pulverized coal incidence velocity), inlet feed temperature T I (including air temperature, steam temperature, coal dust temperature), wherein I represents different inlet reactant species;
internal monitoring parameters of the coal gasifier reactor, including the temperature value T of the violent reaction zone x,y,z Pressure value P x,y,z Gas component concentration C x,y,z (including CO, H) 2 、CO 2 、N 2 、H 2 O、H 2 S and CH 4 Concentration), wherein x, y and z represent the three-dimensional space coordinate position in the reactor under a Cartesian coordinate system;
coal gasifier reactor outlet parameters including gas flow Q at the reactor outlet O (mainly CO flux and H) 2 Flow rate), average temperature T O Average pressure P O ；
c) And storing historical operating data and real-time operating data of the coal gasifier reactor in a database, constructing an operating database of the reactor, removing data abnormal points and carrying out data standardization processing. The method comprises the following specific steps:
processing by using a K-means algorithm to reduce the data dimension; judging whether null values, zero values or abnormal values exist in the data, and if so, supplementing the data by using a Lagrange interpolation method; then, carrying out standardization processing on the data; wherein the normalization processing formula is expressed as:
Figure BDA0003343843930000101
in the formula (I), the compound is shown in the specification,
Figure BDA0003343843930000102
represents a production parameter of the k-th sample normalized with respect to the n-dimensional data>
Figure BDA0003343843930000103
Representing n-dimensional metadata in which K-th sample is arranged in time series, K being dataSet number >>
Figure BDA0003343843930000104
N-dimensional data representing the kth sample, k =1,2, …, N being the number of datasets;
as shown in fig. 1, step two: determining typical reference working conditions of coal gasifier reactor based on typical working condition analysis module
a) Because the coal gasifier reactor has massive data in actual operation, the number of operation working conditions is huge, and therefore representative reference operation working condition data in the reactor needs to be selected preferably. Specifically, based on the coal gasifier reactor operation database constructed in the step one, calculating to obtain representative reference operation condition data in the reactor operation process by adopting a K-means mean value clustering algorithm;
as shown in fig. 1, step three: based on a numerical simulation module, off-line three-dimensional CFD computational fluid dynamics numerical simulation calculation of the internal reaction field of the coal gasifier reactor under the reference operation condition is carried out
a) According to the composition (comprising air flow and CH) of the coal gasifier reactor inlet material 4 Flow, coal dust flow) and chemical reaction conditions, and obtaining each elementary reaction composition, a corresponding chemical equation and reaction kinetic parameters of each step of elementary reaction by adopting Chemkin software to calculate and solve the chemical reaction process among all materials. In the reaction, after the pulverized coal is heated to separate out volatile components, the coal coke and the gas are subjected to out-of-phase reaction, and the specific out-of-phase reaction is considered to comprise:
C(s)+0.5O 2 (g)＝CO(g) (2)
C(s)+H 2 O(g)＝CO(g)+H 2 (g) (3)
C(s)+CO 2 (g)＝2CO(g) (4)
C(s)+2H 2 (g)＝CH 4 (g) (5)
corresponding reaction rate constant k a Apparent activation energy Z a And the value of the reaction order q is as follows:
Figure BDA0003343843930000111
b) Establishing a three-dimensional calculation domain physical model of the reactor according to the operation process, structure and size parameters of the coal gasifier reactor;
c) Adopting Gambit computational domain meshing software to perform meshing on a three-dimensional computational domain model of a coal gasifier reactor, and performing local encryption on meshes of an inlet, an outlet and a structurally complex region of the coal gasifier reactor to obtain a high-precision simulation result; setting boundary condition types of an inlet, an outlet and a wall surface of the reactor, wherein the boundary condition of the inlet of the reactor is a mass inlet, the boundary condition of the outlet of the reactor is a pressure outlet, and the boundary condition of the wall surface of the reactor is a constant temperature wall surface;
d) Setting inlet feeding components and inlet feeding flow Q of the three-dimensional calculation domain model of the reactor according to the reference working condition operation data of the coal gasifier reactor obtained in the step two I Velocity v of the inlet feed I Temperature T of the feed I Wherein I represents a different inlet reactant type;
e) Utilizing a high-performance computer and a CFD numerical simulation platform, adopting the chemical elementary reaction obtained in the step three a), selecting a turbulent flow model, a multiphase flow model and a radiation and reaction mechanism model, utilizing the coal gasifier reactor computational domain grid model obtained in the step three c) to carry out solving simulation on the three-dimensional chemical reaction process in the reactor, and simulating to obtain the speed (v) corresponding to the coordinate (x, y, z) of the three-dimensional space in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The distribution data of (2). Wherein, the turbulent flow model, the multiphase flow model, the radiation model and the reaction mechanism model are respectively expressed as:
in a turbulent flow model, a k-epsilon two-pass model which can better simulate non-rotational flow and can realize the simulation of a turbulent flow process in a reactor is used;
in the multiphase flow model, a Mixture model based on an Euler-Euler method is adopted, wherein a continuity equation and a momentum equation are respectively expressed as follows:
Figure BDA0003343843930000121
Figure BDA0003343843930000122
in the formula, ρ m As density, v is the Hamiltonian, v m In order to be the mass-average velocity,
Figure BDA0003343843930000123
in the form of a vector of mass-averaged velocities, μ m Is the viscosity coefficient of the mixture, T is the temperature of the phase, is based on>
Figure BDA0003343843930000124
Representing the form of a mass-averaged velocity vector at a temperature T, F being the volume force, <' > based on>
Figure BDA0003343843930000125
In the form of a vector of volume forces, B is the total number of phases, k is the kth phase, g m Is acceleration of gravity, </or>
Figure BDA0003343843930000126
In the form of a vector of gravitational acceleration, α k Volume fraction of the k-th phase, p k Is the density of the k-th phase, v dr,k Is the slip speed of the k-th phase>
Figure BDA0003343843930000127
In the form of a slip velocity vector of the k-th phase; />
In the radiation model, P-1 radiation model is used to simulate radiation heat transfer, and the model also includes the mutual radiation among ions and describes radiation flux
Figure BDA0003343843930000128
The equation of (c) can be expressed as follows:
Figure BDA0003343843930000129
wherein beta is the absorption coefficient, C s Is the scattering coefficient, h is the coefficient of the linear anisotropy phase function, v is the hamiltonian, G is the incident radiation;
in the reaction mechanism model, the chemical reaction rate constants follow the arrhenius equation and are expressed as follows:
Figure BDA00033438439300001210
in the formula, k a For reaction rate constant, Z is the molar gas constant, T a Is a thermodynamic temperature, Z a Is apparent activation energy, A is a pre-exponential factor, e is a natural constant;
f) Based on the CFD three-dimensional numerical simulation result obtained in the step three e), performing data verification and CFD calculation model optimization by adopting the coal gasifier reactor operation database constructed in the step one until the calculation error of the CFD three-dimensional numerical simulation result is less than 5%;
g) Repeatedly performing the step three a) to the step three f), calculating to obtain all CFD three-dimensional numerical simulation results of all working conditions in the reference operation working condition of the coal gasifier reactor, and establishing a simulation database of the reaction process of the reference operation working condition of the coal gasifier reactor;
as shown in fig. 1, step four: based on a data clustering and standardization module, a CFD three-dimensional simulation database of the coal gasifier reactor established in the third step is subjected to data set division, parameter processing, redundancy integration and dimensionality reduction by adopting a data mining and principal component analysis method, and 8 principal components are obtained after processing
a) Aiming at a CFD three-dimensional simulation database of a coal gasifier reactor, in order to reduce the number of subsequent data modeling, the data quantity randomly extracted in the reference operation condition of each coal gasifier reactor is 2 ten thousand groups, and in the data extraction process, the data extraction coordinates of each working condition are ensured to be the same;
b) The extracted CFD simulation data information includes: coordinate position (x, y, z) of each calculation node in the coal gasifier reactor and fluid velocity (v) corresponding to each node x ，v y ，v z B), rotation phi x,y,z Temperature T x,y,z Pressure P x,y,z CO concentration, CO 2 Concentration, H 2 Concentration, N 2 Concentration, H 2 S concentration, H 2 Concentration of O, CH 4 Concentration;
c) Because the data extraction amount of each working condition is huge, the direct grouping use can cause the problems of low prediction precision and long time, therefore, in order to improve the calculation precision and speed of the full operation working condition of the coal gasifier reactor, a data mining and principal component analysis method is adopted to carry out correlation analysis on the data; particularly, a K-means mean value clustering algorithm and a dimensionality reduction algorithm are combined to divide a data set and process parameters so as to integrate redundant data and reduce dimensionality of the data, a high-efficiency data set is obtained, and main parameters influencing the chemical reaction process in the coal gasifier reactor are obtained. After data processing, 8 main components are obtained, including flow velocity (v) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z CO concentration, H 2 Concentration, N 2 Concentration, H 2 Concentration of O, CH 4 Concentration;
as shown in fig. 1, fig. 2 and fig. 3, step five: based on a deep learning prediction module, three-dimensional reaction field data under the full-operation working condition of the reactor are obtained by adopting a regression prediction algorithm and a neural network model for calculation
a) Combining a regression prediction algorithm and a neural network model to establish a main parameter rapid prediction model of the internal reaction process of the coal gasifier reactor, wherein the parameters comprise: flow velocity (v) corresponding to three-dimensional space arbitrary position coordinate (x, y, z) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z CO concentration, H 2 Concentration, N 2 Concentration, H 2 Concentration of O, CH 4 Distribution data of concentration. Wherein, the neural network model adopts BP neural networkThe model adopts a lasso regression prediction algorithm as a regression prediction algorithm, and in the lasso regression prediction algorithm, the loss function of the lasso model is assumed as follows:
Figure BDA0003343843930000141
wherein J (β) is a loss function, β j The regression parameter is s, the total data quantity is s, the characteristic variable is L, the calculated data quantity is p, the actual output is G, and the regular coefficient is lambda; beta is a beta j The first RSS part is subjected to partial derivation by using a sub-gradient method to obtain:
Figure BDA0003343843930000142
optimizing and continuously iterating each time to obtain an optimal regression coefficient;
in the BP neural network model, the output layer is the speed (v) corresponding to the reaction field parameters (including the coordinates (x, y, z) of any position in three-dimensional space) in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z ) Recording the number of nodes of each hidden layer, limiting the number of nodes of an input layer and the number of iterations, defining the weight from the input layer to the hidden layer and the weight from the hidden layer to an output layer, the threshold value of the hidden layer and the threshold value of the output layer, and setting the learning rate and the minimum error; wherein, the learning rate setting range is between 0.1 and 0.6, and the minimum error is 1 percent;
b) Taking the inlet parameters of the reactor in the database as an input data set based on the database of the reference operation conditions of the coal gasifier reactor obtained in the fourth step, wherein the inlet parameters comprise: reactor inlet feed composition (including air, CH) 4 Pulverized coal), inlet feed flow Q I (including air flow, steam flow, pulverized coal feed flow), velocity of inlet feed v I (including air incidence velocity, steam incidence velocity, pulverized coal incidence velocity), inlet feed temperature T I (including empty)Gas temperature, steam temperature, pulverized coal temperature), 8 principal components obtained through data processing are used as an output data set to train a neural network model, and the principal components comprise: flow velocity (v) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z CO concentration, H 2 Concentration, N 2 Concentration, H 2 Concentration of O, CH 4 Concentration;
c) The above processes are circulated until the error meets the precision requirement or the learning times are more than the set maximum times, and the training process is finished; the accuracy requirement is that the error of a predicted value is within 5 percent, and the maximum learning frequency to be set is 1000 times;
d) Adopting the trained neural network model to carry out three-dimensional reaction field prediction under all working conditions of the coal gasifier reactor, and specifically taking the difference value between the reactor inlet parameter of the atypical working condition to be predicted and the reference operation working condition as the input variable of the neural network prediction model; using the principal component difference value of the non-reference operation condition and the reference operation condition as an output variable of a prediction model, and carrying out prediction of principal components in a reaction field;
as shown in fig. 1, 4 and 5, step six: based on three-dimensional visualization module, three-dimensional visualization under full-operation working condition of coal gasifier reactor is realized
a) Acquiring three-dimensional reaction field data of the coal gasifier reactor under the non-reference operation condition based on the model prediction in the step five to obtain a three-dimensional reaction field database of the reactor;
b) Integrating the three-dimensional reaction field database under the non-reference operation condition with the three-dimensional reaction field database under the reference operation condition to form a three-dimensional reaction field database of the coal gasifier under the full operation condition;
c) Based on a coal gasifier three-dimensional reaction field database under the full-operation working condition, drawing a three-dimensional reaction field image by using OpenGL software, and constructing a flow velocity (v) corresponding to any position coordinate (x, y, z) in a three-dimensional space x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z CO concentration, H 2 Concentration, N 2 Concentration, H 2 Concentration of O, CH 4 The three-dimensional distribution cloud picture and the numerical scale of the concentration are displayed on a software interface at a computer end in real time for three-dimensional visual display for operators of the reactor to visually observe and check;
according to the three-dimensional visualization result of the coal gasifier reactor, the operating personnel of the reactor can visually know the distribution conditions of the eddy current distribution, the high-temperature area position and the insufficient reaction area in the reactor in real time, the operating personnel can conveniently and pertinently adjust the operating conditions of the reactor, and the required high-concentration and high-quality reaction products can be obtained while the safe and reliable operation of the reactor is realized.
FIG. 4 shows that the method of the present invention is applied to a coal gasifier to predict the three-dimensional axial velocity v in the gasifier under the condition of changing the incident angle of air distribution z And (4) distributing cloud pictures. FIG. 5 shows the predicted three-dimensional temperature T in the coal gasifier under the condition of changing the incident angle of the air distribution x,y,z And (4) distributing cloud pictures. Wherein, the working conditions of 0 degree and 50 degrees are typical reference working conditions, the cloud chart in the figure is the CFD simulation result obtained through the step three, the working conditions of 10 degrees, 20 degrees, 30 degrees and 40 degrees are non-reference working conditions, and the cloud chart in the figure is the prediction result obtained through the step five.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that various changes and modifications can be made by those skilled in the art without departing from the spirit of the invention, and these changes and modifications are all within the scope of the invention. Therefore, the protection scope of the present patent should be subject to the appended claims.

Claims (1)

1. A dynamic real-time visualization method for a three-dimensional reaction field in a reactor is characterized by comprising the following modules:
the system comprises an operation database access module, a typical working condition analysis module, a numerical simulation module, a data clustering and standardization module, a deep learning prediction module and a three-dimensional visualization module, wherein the total six modules are provided;
the operation database access module, the typical working condition analysis module and the numerical simulation module are all offline modules; the data clustering and standardizing module, the deep learning prediction module and the three-dimensional visualization module are all online modules;
the method comprises the following specific steps:
the method comprises the following steps: historical operation data and real-time operation database of reactor are constructed based on operation database access module
a) Extracting historical operating data of the reactor from a historical operating database of the reactor according to a certain time interval delta t, wherein the historical operating data comprises reactor inlet parameters, reactor internal monitoring parameters and reactor outlet parameters;
reactor inlet parameters, including the composition of the reactor inlet feed, the flow rate Q of the inlet feed I Velocity v of the inlet feed I Temperature T of the inlet feed I Wherein, I represents different inlet reactant species;
reactor internal monitoring parameters, including temperature values T of the reaction zone x,y,z Pressure value P x,y,z Gas component concentration C x,y,z Wherein x, y and z represent the three-dimensional space coordinate position in the reactor under a Cartesian coordinate system;
reactor outlet parameters including gas flow Q at the reactor outlet O Average temperature T O Average pressure P O ；
b) Extracting real-time operation data of the reactor according to a certain time interval delta t in the operation process of the reactor, and analyzing the working condition;
reactor inlet parameters including composition of the reactor inlet feed, flow rate Q of the inlet feed I Velocity v of the inlet feed I Temperature T of the inlet feed I Wherein I represents a different inlet reactant type;
reactor internal monitoring parameters, including temperature values T of the reaction zone x Pressure value P x Gas component concentration C x Wherein x, y and z represent the three-dimensional space coordinate position in the reactor under a Cartesian coordinate system;
reactor outlet parameters including the gas flow Q at the reactor outlet O Average temperature T O Average pressure P O ；
c) Storing historical operating data and real-time operating data of the reactor in a database, constructing an operating database of the reactor, removing data abnormal points and carrying out data standardization processing; the method comprises the following specific steps:
processing by using a K-means algorithm to reduce data dimensionality; judging whether the data contains null values, 0 values or abnormal values, and if so, filling the data by using a Lagrange interpolation method; then, carrying out standardization processing on the data; wherein the normalization processing formula is expressed as:
Figure FDA0003343843920000021
in the formula (I), the compound is shown in the specification,
Figure FDA0003343843920000022
representing a production parameter of the kth sample normalized in n-dimensional data>
Figure FDA0003343843920000023
N-dimensional metadata representing the arrangement of the kth sample in time series, K being the number of data sets, and->
Figure FDA0003343843920000024
N-dimensional data representing the kth sample, k =1,2, …, N being the number of datasets;
step two: determining typical reference conditions of a reactor based on a typical condition analysis module
a) Calculating to obtain data of a reference operation condition in the reactor operation process by adopting a K-means mean value clustering algorithm in the reactor operation database constructed in the step one;
step three: based on a numerical simulation module, off-line three-dimensional CFD computational fluid dynamics numerical simulation calculation of the reactor internal reaction field under the reference operation condition is carried out
a) Calculating and solving the chemical reaction process among the materials by using Chemkin software according to the components of the materials at the inlet of the reactor and the chemical reaction conditions to obtain the reaction composition of each element, the corresponding chemical equation and the reaction kinetic parameters of the element reaction in each step;
b) Establishing a three-dimensional calculation domain physical model of the reactor according to the operation process, structure and size parameters of the reactor;
c) Adopting Gambit computational domain meshing software to perform meshing on a three-dimensional computational domain model of the reactor, and performing local encryption on meshes of an inlet, an outlet and a structurally complex region of the reactor to obtain a high-precision simulation result; setting boundary condition types of an inlet, an outlet and a wall surface of the reactor, wherein the boundary condition of the inlet of the reactor is a mass inlet, the boundary condition of the outlet of the reactor is a pressure outlet, and the boundary condition of the wall surface of the reactor is a constant temperature wall surface;
d) Setting inlet feeding components and inlet feeding flow Q of the three-dimensional calculation domain model of the reactor according to the reactor reference working condition operation data obtained in the step two I Velocity v of the inlet feed I Temperature T of the feed I Wherein, I represents different inlet reactant types;
e) Utilizing a high-performance computer and a CFD numerical simulation platform, adopting the chemical elementary reaction obtained in the step three a), selecting a turbulent flow model, a multiphase flow model and a radiation and reaction mechanism model, utilizing the reactor computational domain grid model obtained in the step three c) to carry out solving simulation on the three-dimensional chemical reaction process in the reactor, and simulating to obtain the speed (v) corresponding to the coordinate (x, y, z) of the three-dimensional space in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The distribution data of (2); the turbulent flow model, the multiphase flow model, the radiation model and the reaction mechanism model are respectively expressed as follows:
in the turbulent flow model, a k-epsilon two-pass model is used for simulating the turbulent flow process in the reactor;
in the multiphase flow model, a Mixture model based on an Euler-Euler method is adopted, wherein a continuity equation and a momentum equation are respectively expressed as follows:
Figure FDA0003343843920000031
Figure FDA0003343843920000032
in the formula, ρ m In order to be the density of the mixture,
Figure FDA0003343843920000033
as Hamiltonian, v m For the mass mean speed, is>
Figure FDA0003343843920000034
In the form of a vector of mass-averaged velocities, μ m In combination with a viscosity index, T is the temperature of the phase>
Figure FDA0003343843920000035
Representing the form of a mass-averaged velocity vector at a temperature T, F being the volume force, <' > based on>
Figure FDA0003343843920000036
In the form of a vector of the volume force, B is the total number of phases, k is the kth phase, g m Is based on gravity acceleration>
Figure FDA0003343843920000037
In the form of a vector of gravitational acceleration, α k Volume fraction of the k-th phase, p k Is the density of the k-th phase, v dr,k For the slip speed of the k-th phase>
Figure FDA0003343843920000038
Is slip speed of the k-th phaseVector form;
the radiation heat transfer is simulated in a radiation model using a P-1 radiation model, which also includes inter-ion radiation describing the radiation flux
Figure FDA0003343843920000039
The equation of (c) can be expressed as follows:
Figure FDA00033438439200000310
wherein beta is the absorption coefficient, C s Is the scattering coefficient, h is the linear anisotropy phase function coefficient,
Figure FDA00033438439200000311
is Hamiltonian, G is incident radiation;
in the reaction mechanism model, the chemical reaction rate constants follow the arrhenius formula and are expressed as follows:
Figure FDA0003343843920000041
in the formula, k a For reaction rate constant, Z is the molar gas constant, T a Is a thermodynamic temperature, Z a Is the apparent activation energy, A is the pre-exponential factor, e is the natural constant;
f) Based on the CFD three-dimensional numerical simulation result obtained in the step three e), performing data verification and CFD calculation model optimization by adopting the reactor operation database constructed in the step one until the calculation error of the CFD three-dimensional numerical simulation result is less than 5%;
g) Repeating the step three a) to the step three f), calculating to obtain all CFD three-dimensional numerical simulation results of all working conditions in the reactor reference operation working condition, and establishing a simulation database of the reactor reference operation working condition reaction process;
step four: based on a data clustering and standardization module, a reactor CFD three-dimensional simulation database established in the third step is subjected to data set division, parameter processing, integration redundancy and dimensionality reduction by adopting a data mining and principal component analysis method, and 8-10 principal components are obtained after processing
a) Aiming at a CFD three-dimensional simulation database of reactors, in order to reduce the number of subsequent data modeling, the data quantity randomly extracted in the reference operation condition of each reactor is 2-3 ten thousand groups, and in the data extraction process, the data extraction coordinates of each working condition are ensured to be the same;
b) The extracted CFD simulation data information includes: the coordinate position (x, y, z) of each calculation node in the reactor and the fluid velocity (v) corresponding to each node x ，v y ，v z B), rotation phi x,y,z Reaction product component concentration C x,y,z Temperature T x,y,z And pressure T x,y,z ；
c) Performing correlation analysis on the data by adopting a data mining and principal component analysis method; specifically, a K-means mean clustering algorithm and a dimensionality reduction algorithm are combined to carry out data set division and parameter processing so as to integrate redundant data and carry out data dimensionality reduction to obtain 8-10 principal components;
step five: based on a deep learning prediction module, three-dimensional reaction field data under the full-operation working condition of the reactor is obtained by adopting a regression prediction algorithm and a neural network model for calculation
a) Combining a regression prediction algorithm and a neural network model to establish a main parameter rapid prediction model of the internal reaction process of the reactor, wherein the parameters comprise: velocity (v) corresponding to three-dimensional space arbitrary position coordinate (x, y, z) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The distribution data of (2); the neural network model adopts a BP neural network model, the regression prediction algorithm adopts a lasso regression prediction algorithm, and in the lasso regression prediction algorithm, the loss function of the lasso model is assumed as follows:
Figure FDA0003343843920000051
in the formula (I), the compound is shown in the specification,j (β) is the loss function, β j The regression parameter is s, the total data quantity is s, the characteristic variable is L, the calculated data quantity is p, the actual output is G, and the regular coefficient is lambda; beta is a beta j The first part is subjected to partial derivation by using a sub-gradient method to obtain:
Figure FDA0003343843920000052
optimizing and continuously iterating each time to obtain an optimal regression coefficient;
in the BP neural network model, the output layer is the speed (v) corresponding to the reaction field parameters (including the coordinates (x, y, z) of any position in three-dimensional space) in the reactor x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z ) Recording the number of nodes of each hidden layer, limiting the number of nodes of an input layer and the number of iterations, defining the weight from the input layer to the hidden layer and the weight from the hidden layer to an output layer, the threshold value of the hidden layer and the threshold value of the output layer, and setting the learning rate and the minimum error; wherein, the learning rate setting range is between 0.1 and 0.6, and the minimum error is 1 percent;
b) Taking the inlet parameters of the reactor in the database as an input data set based on the database of the reference operation conditions of the reactor obtained in the fourth step, wherein the inlet parameters comprise: composition of inlet feed, flow rate of inlet feed Q I Inlet feed velocity v I Temperature T of the inlet feed I (ii) a Taking 8 to 10 principal components obtained by data processing as an output data set to train a neural network model, wherein the principal components comprise velocities (v) corresponding to three-dimensional space arbitrary position coordinates (x, y, z) x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z ；
c) The above processes are circulated until the error meets the precision requirement or the learning times is more than the set maximum times, and the training process is ended; the accuracy requirement is that the error of a predicted value is within 5 percent, and the maximum learning frequency to be set is 1000 times;
d) Adopting the trained neural network model to carry out three-dimensional reaction field prediction under all working conditions of the reactor, and specifically taking the difference value between the inlet parameter of the reactor under the atypical working condition to be predicted and the reference operation working condition as the input variable of the neural network prediction model; using the principal component difference value of the non-reference operation condition and the reference operation condition as an output variable of a prediction model, and carrying out prediction of principal components in a reaction field;
step six: based on a three-dimensional visualization module, the three-dimensional visualization under the full operation condition of the reactor is realized
a) Acquiring a reactor three-dimensional reaction field database based on the reactor three-dimensional reaction field data under the non-reference operation condition obtained by model prediction in the step five;
b) Integrating the three-dimensional reaction field database under the non-reference operation condition with the three-dimensional reaction field database under the reference operation condition to form a three-dimensional reaction field database under the full operation condition;
c) Based on a three-dimensional reaction field database under the full operation condition, drawing a three-dimensional reaction field image by using OpenGL software, and constructing a speed (v) corresponding to any position coordinate (x, y, z) in a three-dimensional space x ，v y ，v z B), temperature T x,y,z Pressure P x,y,z Reaction product component concentration C x,y,z The three-dimensional distribution cloud picture and the numerical scale are displayed on a computer software interface in real time to carry out three-dimensional visual display.
CN202111316547.9A 2021-11-08 2021-11-08 Dynamic real-time visualization method for three-dimensional reaction field in reactor Active CN114117954B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111316547.9A CN114117954B (en) 2021-11-08 2021-11-08 Dynamic real-time visualization method for three-dimensional reaction field in reactor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111316547.9A CN114117954B (en) 2021-11-08 2021-11-08 Dynamic real-time visualization method for three-dimensional reaction field in reactor

Publications (2)

Publication Number Publication Date
CN114117954A CN114117954A (en) 2022-03-01
CN114117954B true CN114117954B (en) 2023-04-18

Family

ID=80377586

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111316547.9A Active CN114117954B (en) 2021-11-08 2021-11-08 Dynamic real-time visualization method for three-dimensional reaction field in reactor

Country Status (1)

Country Link
CN (1) CN114117954B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358152B (en) * 2022-08-26 2023-04-07 绍兴市再生能源发展有限公司 Garbage incineration gas control and feedback regulation system and method
CN116478778A (en) * 2023-04-18 2023-07-25 泸州老窖股份有限公司 Space samming type temperature control fermentation system
CN116486404B (en) * 2023-06-25 2023-09-26 苏州创腾软件有限公司 Needle coke microscopic image detection method and device based on convolutional neural network

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005249418A (en) * 2004-03-01 2005-09-15 Mazda Motor Corp Prediction analysis method of engine performance, prediction analysis system and its control program
CN111339716A (en) * 2020-02-19 2020-06-26 浙江大学 Boiler high-temperature flue gas flow field online proxy model construction method
CN112036100A (en) * 2020-09-24 2020-12-04 哈尔滨锅炉厂有限责任公司 Method for predicting hearth oxygen concentration by using regression algorithm based on numerical simulation
CN112163380A (en) * 2020-09-24 2021-01-01 哈尔滨锅炉厂有限责任公司 System and method for predicting furnace oxygen concentration based on numerical simulation neural network
CN112216354A (en) * 2020-09-17 2021-01-12 江苏集萃工业过程模拟与优化研究所有限公司 Intelligent dosing system and method based on CFD numerical simulation and machine learning
CN112528569A (en) * 2020-10-17 2021-03-19 中国石油化工股份有限公司 Industrial heating furnace temperature field prediction method based on digital twinning
CN113505440A (en) * 2021-07-28 2021-10-15 大连理工大学 Automobile pneumatic performance parameter real-time prediction method based on three-dimensional deep learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005249418A (en) * 2004-03-01 2005-09-15 Mazda Motor Corp Prediction analysis method of engine performance, prediction analysis system and its control program
CN111339716A (en) * 2020-02-19 2020-06-26 浙江大学 Boiler high-temperature flue gas flow field online proxy model construction method
CN112216354A (en) * 2020-09-17 2021-01-12 江苏集萃工业过程模拟与优化研究所有限公司 Intelligent dosing system and method based on CFD numerical simulation and machine learning
CN112036100A (en) * 2020-09-24 2020-12-04 哈尔滨锅炉厂有限责任公司 Method for predicting hearth oxygen concentration by using regression algorithm based on numerical simulation
CN112163380A (en) * 2020-09-24 2021-01-01 哈尔滨锅炉厂有限责任公司 System and method for predicting furnace oxygen concentration based on numerical simulation neural network
CN112528569A (en) * 2020-10-17 2021-03-19 中国石油化工股份有限公司 Industrial heating furnace temperature field prediction method based on digital twinning
CN113505440A (en) * 2021-07-28 2021-10-15 大连理工大学 Automobile pneumatic performance parameter real-time prediction method based on three-dimensional deep learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
*** ; .机车散热器空气流动的数值模拟分析.内燃机车.2010,(第09期),全文. *
童宝宏 ; 尹军 ; .流场分析与智能建模在机油泵CAD中的联合应用.农业工程学报.2011,(第08期),全文. *

Also Published As

Publication number Publication date
CN114117954A (en) 2022-03-01

Similar Documents

Publication Publication Date Title
CN114117954B (en) Dynamic real-time visualization method for three-dimensional reaction field in reactor
Wang et al. Hybrid physics-based and data-driven models for smart manufacturing: Modelling, simulation, and explainability
Hummel et al. Comparative visual analysis of Lagrangian transport in CFD ensembles
US11373026B2 (en) Deep learning surrogate for turbulent flow
CN114757070A (en) New WENO format construction method under trigonometric function framework for numerical simulation
Nnabuife et al. Development of Gas–Liquid Flow Regimes Identification Using a Noninvasive Ultrasonic Sensor, Belt-Shape Features, and Convolutional Neural Network in an S-Shaped Riser
CN112948932A (en) Surrounding rock grade prediction method based on TSP forecast data and XGboost algorithm
Benamara et al. Adaptive infill sampling criterion for multi-fidelity optimization based on Gappy-POD: Application to the flight domain study of a transonic airfoil
CN113570165B (en) Intelligent prediction method for permeability of coal reservoir based on particle swarm optimization
Wu et al. Sparse local fisher discriminant analysis for gas-water two-phase flow status monitoring with multisensor signals
CN114912364A (en) Natural gas well flow prediction method, device, equipment and computer readable medium
Yan et al. Exploring hidden flow structures from sparse data through deep-learning-strengthened proper orthogonal decomposition
Zhu et al. New insights into experimental stratified flows obtained through physics-informed neural networks
CN109960146A (en) The method for improving soft measuring instrument model prediction accuracy
CN113378492A (en) Low-magnetic-Reynolds-number magnetic fluid turbulence numerical calculation method based on electromagnetic correction
Freulon et al. CytOpT: Optimal transport with domain adaptation for interpreting flow cytometry data
Chen et al. An airfoil mesh quality criterion using deep neural networks
Nielsen et al. Novel strategies for predictive particle monitoring and control using advanced image analysis
Huang et al. Data-driven-based forecasting of two-phase flow parameters in rectangular channel
CN114565209A (en) Process industry energy consumption state evaluation method based on clustering
Mallia et al. Intelligent Approaches for Anomaly Detection in Compressed Air Systems: A Systematic Review
Ries et al. Utilizing distributional measurements of material characteristics from sem images for inverse prediction
Beykal et al. Data-driven Stochastic Optimization of Numerically Infeasible Differential Algebraic Equations: An Application to the Steam Cracking Process
Yu et al. Predicting unavailable parameters from existing velocity fields of turbulent flows using a GAN-based model
Li et al. Pressure prediction for air cyclone centrifugal classifier based on CNN-LSTM enhanced by attention mechanism

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