CN-121981001-A - Rapid expansion cloud chamber environment simulation method and system based on artificial intelligence

CN121981001ACN 121981001 ACN121981001 ACN 121981001ACN-121981001-A

Abstract

The invention discloses a rapid expansion cloud room environment simulation method and system based on artificial intelligence, comprising the steps of obtaining experimental data and experimental conditions of an expansion cloud room to be simulated, carrying out structural classification according to expansion stages based on the experimental data according to the experimental conditions, obtaining feature vectors of each stage, carrying out element learning through a neural network according to the feature vectors, obtaining a cloud and fog simulation model and simulation errors, identifying a cloud room working condition interval based on the simulation errors and the experimental data, obtaining an output result of the cloud and fog simulation model in a high deviation interval according to the working condition interval, correcting interval results through a CFD equation based on the output result, and obtaining a cloud room environment simulation result. According to the method, CFD local correction is carried out on the simulated high deviation zone by utilizing the cloud expansion stage of splitting and the fog drop growth fitting process, so that the prediction error control precision is improved, the prediction and collaborative analysis efficiency of the fog forming process in a complex environment is improved, and meanwhile, the method has good interpretability.

Inventors

CHE YUNFEI
LIU WEI
LIU XIJING
FANG CHUNGANG
ZHANG JINMING
LIANG YUANXIN
SU ZHENGJUN
Dang juan
Xie Zhengshuai

Assignees

中国气象局人工影响天气中心

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (8)

1. The rapid expansion cloud chamber environment simulation method based on artificial intelligence is characterized by comprising the following steps of: Obtaining experimental data and experimental conditions of an expansion cloud chamber to be simulated, and preprocessing the experimental data, wherein the experimental data comprises a temperature field, a pressure field, a fog drop spectrum and a wall surface heat flux in a cabin; Based on the experimental data, carrying out structural classification according to the expansion stage according to experimental conditions to obtain a time sequence data set comprising a cooling stage, an expansion stage and a maintenance stage, and extracting feature vectors of each stage; constructing a physical information neural network through a dissipative Hamiltonian equation according to the feature vector, and performing element learning on the neural network to obtain a cloud and fog simulation model and a simulation error; identifying a cloud chamber working condition interval based on the simulation error and experimental data, and obtaining an output result of a cloud and fog simulation model in a high deviation interval according to the working condition interval, wherein the identification conditions comprise a strong dissipation interval and a phase change critical point of cloud and fog; And taking the output result as a boundary condition, correcting the interval result through a CFD equation according to the boundary condition, and splicing the low-deviation interval result and the interval correction result to obtain a cloud chamber environment simulation result.
2. The artificial intelligence based rapid expansion cloud chamber environment simulation method of claim 1, wherein the method of obtaining the experimental data and experimental conditions comprises: Acquiring temperature distribution data, pressure change data, fog drop spectrums and wall heat flux density in a cabin based on an expansion cloud chamber to be simulated according to an expansion period, unifying the temperature distribution data and the pressure change data into the same space-time coordinate system, performing wavelet filtering on the temperature distribution data and the pressure change data to obtain a temperature field and a pressure field, and converting the temperature field, the pressure field, the fog drop spectrums and the wall heat flux density into a structured time sequence format, wherein the wall heat flux density is the heat exchange rate of the cloud chamber wall and the cabin gas in a unit area; and taking the temperature, the humidity and the pressure of the corresponding cloud room control system as experimental conditions according to the structured time sequence data, and correlating the structured time sequence data with the experimental conditions to obtain experimental data.
3. The artificial intelligence based rapid expansion cloud chamber environment simulation method of claim 1, wherein the method of obtaining the time series data set comprises: Dividing the expansion period of the cloud chamber by a pressure change rate threshold method based on a pressure field, marking a stage with negative pressure change rate and gradually increased absolute value from initial pressure as a cooling stage, marking a stage with rapid decrease after the pressure change rate reaches a peak value as an expansion stage, and marking a stage with pressure change rate approaching zero and stable pressure as a maintenance stage; And carrying out structural classification on experimental data according to the cooling section, the expansion section and the maintaining section to obtain time sequence data sets of all stages, and extracting feature vectors of all stages based on the time sequence data sets.
4. The artificial intelligence based rapid expansion cloud chamber environment simulation method of claim 3, wherein the method for obtaining the feature vector comprises the following steps: Extracting an average temperature change rate, a wall heat flow density peak value, a pressure drop rate, a difference value between an initial temperature and an initial pressure based on a time sequence data set in the cooling section, calculating saturated water vapor pressure based on a temperature field of the time sequence data set in the expansion section through a Clausius keapertural equation, and taking the ratio of the actual water vapor pressure of a pressure field to the saturated water vapor pressure as a supersaturation degree; Obtaining an ordered list based on an expansion section according to a discrete measurement value of the average radius of mist drops changing along with time in a mist drop spectrum, carrying out center difference on middle points of the ordered list, respectively carrying out forward difference and backward difference on head and tail boundary points of the ordered list, calculating an experimental deduction value of a growth rate according to the difference result, minimizing residual errors of the predicted value and the experimental deduction value of the mist drop growth rate through an L-BFGS-B algorithm, obtaining an optimal interface energy coefficient according to the residual errors, and calculating a model value of the mist drop growth rate according to the supersaturation degree and the interface energy coefficient, wherein the calculation formula of the mist drop growth rate is as follows: ; Wherein the method comprises the steps of As the average growth rate of the mist droplets, The diffusion coefficient of water vapor in the air is fitted according to the Chapman Scott theory, The average thermal movement speed of the water vapor molecules in the temperature field is obtained according to experimental data, In order for the degree of supersaturation to be high, The average radius of the fog drops is set as, Is the interfacial energy coefficient of the fog drops, Is a boltzmann constant, Obtaining the instantaneous temperature in the cloud cabin according to a temperature field; and extracting a pressure change rate peak value, supersaturation spatial-temporal distribution, a fog drop growth rate model value and a fog drop number concentration peak value based on the time sequence data set of the expansion section, extracting an average pressure, a fog drop average radius steady-state value, a temperature field distribution variance and a wall surface heat flow density steady-state value based on the time sequence data set of the maintenance section, and obtaining feature vectors of each stage.
5. The artificial intelligence based rapid expansion cloud chamber environment simulation method according to claim 1, wherein the method for obtaining the cloud simulation model and simulation error comprises the following steps: dividing a diffusion period into a reversible process and an irreversible process based on feature vectors, generating a dissipative Hamiltonian equation according to the system Hamiltonian amount of the reversible process and the dissipative Hamiltonian amount of the irreversible process, taking the residual error of the dissipative Hamiltonian equation as a loss term of a physical information neural network to obtain a loss function, wherein the neural network structure is a deep fully-connected neural network, and the loss function is as follows: ; Wherein the method comprises the steps of As a loss function of the physical information neural network, As a total number of experimental data points, Is the neural network to the first Predicted values of the cloud state of each space-time point, Is the first Experimental observations of the state of the cloud of space-time points, Is the square of the L2 norm, As the weight coefficient of the light-emitting diode, For the purpose of the gradient operator, In order to be a hamiltonian amount, In the form of a poisson matrix, Dissipation factor for the dissipation process; Acquiring characteristic vectors of different batches of cloud chambers according to the experimental conditions, wherein the different batches comprise working conditions of different initial pressures, initial temperatures and initial aerosol concentrations, performing element learning on the neural network based on the characteristic vectors of the different batches of cloud chambers and a dissipation Hamiltonian equation, initializing the weight and bias of the neural network according to element learning, calculating a loss function of each batch of data, updating network parameters through back propagation, and completing pre-training until the continuous 100-round iteration variable quantity of the loss function is smaller than 10 -6 and the residual error of the dissipation Hamiltonian equation is smaller than 10 -4 ; And fine-tuning network parameters through characteristic vectors of a new working condition batch based on the weight of the pre-training model to obtain a cloud simulation model, outputting cloud states of the expansion cloud chamber at space-time points and simulation errors based on the cloud simulation model, wherein the cloud states comprise fog drop number concentration, supersaturation and temperature field distribution, the simulation errors comprise mean square errors of experimental values and predicted values and residual errors of a dissipation Hamiltonian equation, and fine-tuning rounds are 10-20.
6. The artificial intelligence based rapid expansion cloud chamber environment simulation method of claim 1, wherein the method for obtaining the output result comprises the following steps: Aligning cloud simulation model output, simulation errors and feature vectors according to time sequence of experimental data to obtain a working condition data set, identifying a working condition interval in a cloud chamber expansion period according to the working condition data set, wherein the working condition interval is a strong dissipation interval if a physical residual error of a dissipation Hamiltonian equation is larger than a preset threshold value, the working condition interval is a strong dissipation interval if a dissipation coefficient exceeds twice of an average dissipation coefficient of the cloud chamber, and the working condition interval is a strong dissipation interval if an experimental deduction value of a droplet growth rate and a model predicted value residual error are larger than 15%; Obtaining critical supersaturation according to the starting point of the droplet burst generation according to the droplet spectrum, identifying the critical point as a phase change critical point if the supersaturation reaches the critical supersaturation, and identifying the critical point as a phase change critical point of the cloud and mist system from unbalance to steady state transition if the second derivative of the pressure change rate has an extreme value; And screening candidate high deviation intervals based on the strong dissipation interval and the phase change critical point, if the mean square error of the experimental value and the predicted value in the candidate high deviation intervals is greater than 5%, identifying the candidate high deviation intervals as the high deviation intervals in the cloud chamber expansion period, and extracting output results in the intervals, wherein the output results comprise weight change trends in the cloud mist simulation model high deviation intervals.
7. The artificial intelligence based rapid expansion cloud chamber environment simulation method according to claim 1, wherein the method for obtaining the cloud chamber environment simulation result comprises the following steps: establishing a three-dimensional geometric model based on the diameter, height, air inlet position, observation window layout and sampling sensor position of a cabin of an expansion cloud chamber to be simulated, dividing a structured grid for the center of the cabin according to geometric features of the three-dimensional geometric model, dividing an unstructured grid for boundary layers, air inlets and strong dissipation regions of the wall surface of the cabin, and generating 5 to 10 layers of boundary layer grids at the wall surface of the cabin to obtain CFD calculation grids; Taking an output result as a boundary condition of a CFD calculation grid according to the high deviation interval, interpolating a speed field of the boundary condition to a time step of the CFD calculation grid based on the experimental data sampling time stamp, mapping a space coordinate of the boundary condition to a calculation grid node of the CFD, and interpolating the output result to the center of a grid unit through bilinear interpolation, wherein the CFD grid unit is associated with a corresponding high deviation interval label, a speed value and an error value; Calculating an interval correction result of the CFD grid unit through a Navigator equation, wherein the calculation formula of the interval correction result is as follows: ; Wherein the method comprises the steps of Is the velocity vector of the air-water vapor mixed flow in the cloud chamber, Is the velocity vector of the air-water vapor mixed flow in the cloud chamber, In the form of the outer product of the vectors, The kinematic viscosity of the air-water vapor mixed flow is obtained according to the supersaturation degree, In order for the laplace operator to be useful, The density of the air-water vapor mixed flow is calculated and obtained through an ideal gas state equation according to the actual water vapor partial pressure, the actual saturated water vapor pressure, the total cloud chamber pressure and the cloud chamber temperature corresponding to the fog drop spectrum, Is a pressure gradient vector, is obtained discretely through a central differential format according to a pressure field corresponding to the CFD grid, In order to relax the factor of the process, To indicate a function, the high deviation interval takes 1, the low deviation interval takes 0, For a neurophysically predicted velocity field, A velocity field initially calculated for the CFD; And splicing the cloud and fog simulation model output result of the low-deviation interval with the interval correction result according to the space-time coordinates based on the CFD grid unit to obtain a cloud chamber environment simulation result, and performing enhanced sampling on experimental data of the high-deviation interval according to error distribution based on the cloud chamber environment simulation result and supplementing the experimental data to a training set of a physical information neural network.
8. An artificial intelligence based rapid expansion cloud chamber environment simulation system for executing the artificial intelligence based rapid expansion cloud chamber environment simulation method according to any one of claims 1to 7, characterized in that the system comprises: the data acquisition module is used for acquiring experimental data and experimental conditions of the rapid expansion cloud chamber and preprocessing the experimental data, wherein the experimental data comprises a temperature field, a pressure field, a fog drop spectrum and a wall surface heat flux density in the cabin; The feature classification module is used for carrying out structural classification according to the expansion stage based on the experimental data according to experimental conditions, obtaining a time sequence data set comprising a cooling stage, an expansion stage and a maintenance stage, and extracting feature vectors of each stage; The data analysis module is used for constructing a physical information neural network through a dissipative Hamiltonian equation according to the feature vector, and performing element learning on the neural network to obtain a cloud and fog simulation model and a simulation error; the deviation interval identification module is used for identifying a cloud chamber working condition interval based on the simulation error and experimental data, and obtaining an output result of the cloud and fog simulation model in a high deviation interval according to the working condition interval, wherein the identification conditions comprise a strong dissipation interval and a phase change critical point of cloud and fog; And the environment simulation result module is used for taking the output result as a boundary condition, correcting the interval result through the CFD equation according to the boundary condition, and splicing the low-deviation interval result and the interval correction result to obtain the cloud chamber environment simulation result.

Description

Rapid expansion cloud chamber environment simulation method and system based on artificial intelligence Technical Field The invention relates to the field of expansion cloud chambers, in particular to a rapid expansion cloud chamber environment simulation method and system based on artificial intelligence. Background The rapid expansion cloud chamber is important equipment for simulating cloud-forming and rain-making environment in the atmosphere, researching the nucleation process of aerosol particles and artificially influencing the action mechanism of a weather catalyst, the expansion cloud chamber builds a water vapor supersaturated environment through the mutation of parameters such as accurate control pressure, temperature and the like, and analyzes the cloud and fog microcosmic physical process, so that important data references are provided for optimizing scenes such as artificial rain-making operation, a plurality of theories are formed in the warm cloud artificial rain-making mechanism research around the expansion cloud chamber, the cold cloud process and the ice nucleus research are mainly used, various isothermal cloud chambers and comprehensive cloud chambers are built, the experimental conditions of a cloud cabin are set according to required environmental data for simulating the cloud and fog physical environment, the cloud chamber is controlled through the experimental conditions, the experiment cannot completely cover all environmental scenes due to the complex change characteristics of the cloud physical environment, and along with the fusion of artificial intelligent technology in the fields such as fluid simulation, the self-adaptive optimization control of environmental parameters is gradually realized through the introduction of a machine learning algorithm, and the simulation precision and experimental efficiency are improved. However, the fusion application of the artificial intelligence and the rapid expansion cloud chamber is still in a preliminary stage, the application of multi-parameter post analysis is focused on, the application of multi-parameter dynamic association, boundary effect compensation and real-time regulation and control requirements is less, and the problem of parameter mutation, nonlinear response and the like easily occur in analysis results due to lack of the specific physical process of the cloud chamber is solved, so that the rapid expansion cloud chamber is subjected to environment simulation, the intelligent prediction of the fogging process in the complex environment and the multi-physical quantity collaborative analysis are improved, and the simulation precision and experimental efficiency are improved. Disclosure of Invention The invention aims to provide a rapid expansion cloud room environment simulation method and system based on artificial intelligence. In order to achieve the above purpose, the invention is implemented according to the following technical scheme: The first aspect of the invention provides a rapid expansion cloud chamber environment simulation method based on artificial intelligence, which comprises the following steps: Obtaining experimental data and experimental conditions of an expansion cloud chamber to be simulated, and preprocessing the experimental data, wherein the experimental data comprises a temperature field, a pressure field, a fog drop spectrum and a wall surface heat flux in a cabin; Based on the experimental data, carrying out structural classification according to the expansion stage according to experimental conditions to obtain a time sequence data set comprising a cooling stage, an expansion stage and a maintenance stage, and extracting feature vectors of each stage; constructing a physical information neural network through a dissipative Hamiltonian equation according to the feature vector, and performing element learning on the neural network to obtain a cloud and fog simulation model and a simulation error; identifying a cloud chamber working condition interval based on the simulation error and experimental data, and obtaining an output result of a cloud and fog simulation model in a high deviation interval according to the working condition interval, wherein the identification conditions comprise a strong dissipation interval and a phase change critical point of cloud and fog; And taking the output result as a boundary condition, correcting the interval result through a CFD equation according to the boundary condition, and splicing the low-deviation interval result and the interval correction result to obtain a cloud chamber environment simulation result. Further, a method of obtaining the experimental data and experimental conditions, comprising: Acquiring temperature distribution data, pressure change data, fog drop spectrums and wall heat flux density in a cabin based on an expansion cloud chamber to be simulated according to an expansion period, unifying the temperature distribution data and the pressure