CN114417723B - An application method of two-phase flow heat transfer model based on neural network - Google Patents
An application method of two-phase flow heat transfer model based on neural networkDownload PDFInfo
- Publication number
- CN114417723B CN114417723BCN202210068693.2ACN202210068693ACN114417723BCN 114417723 BCN114417723 BCN 114417723BCN 202210068693 ACN202210068693 ACN 202210068693ACN 114417723 BCN114417723 BCN 114417723B
- Authority
- CN
- China
- Prior art keywords
- heat transfer
- calculation
- neural network
- phase flow
- flow
- 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.)
- Expired - Fee Related
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E30/00—Energy generation of nuclear origin
- Y02E30/30—Nuclear fission reactors
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Monitoring And Testing Of Nuclear Reactors (AREA)
- Feedback Control In General (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本发明属于两相流动换热技术领域，具体涉及一种基于神经网络的两相流动换热模型的应用方法。The invention belongs to the technical field of two-phase flow heat transfer, and in particular relates to an application method of a neural network-based two-phase flow heat transfer model.
背景技术Background technique
两相流动换热机理复杂，受到压力、工质、管径、流速、含气率等许多因素影响，具有显著的非线性特性，尚未形成准确的理论模型；在核动力系统的设计分析中，沸水堆堆芯、蒸汽发生器二次侧、事故工况下压水堆堆芯的流体域计算都会涉及到两相流动换热工况，而核动力系统中热量的传输主要是靠流体与固体壁面的相对流动来完成，两相流动换热同时会影响燃料元件(固体域)的计算；The heat transfer mechanism of two-phase flow is complex, affected by many factors such as pressure, working fluid, pipe diameter, flow velocity, gas content, etc., and has significant nonlinear characteristics, and an accurate theoretical model has not yet been formed; in the design analysis of nuclear power systems, The calculation of the fluid domain of the boiling water reactor core, the secondary side of the steam generator, and the pressurized water reactor core under accident conditions will involve two-phase flow heat transfer conditions, and the heat transfer in the nuclear power system is mainly by fluid and solid The relative flow of the wall is completed, and the heat transfer of the two-phase flow will also affect the calculation of the fuel element (solid domain);
目前，两相流动换热计算模型为实验获得，开展不同压力、工质、管径、流速、含气率下的两相流动换热实验，获得换热系数、临界热流密度、流型等两相流动换热关键参数，将实验数据进行拟合，获得经验关系式；At present, the calculation model of two-phase flow heat transfer is obtained from experiments. Two-phase flow heat transfer experiments are carried out under different pressures, working fluids, pipe diameters, flow rates, and gas fractions, and two parameters such as heat transfer coefficient, critical heat flux density, and flow pattern are obtained. The key parameters of phase flow heat transfer are fitted by experimental data to obtain empirical relational expressions;
现行的热工水力程序中，通常使用上述方法获得的经验关系式进行计算，不同的经验关系式适用范围小并且具有一定的误差，因此会综合考虑多种经验关系式，使用最为保守的模型计算；但是过于保守的两相流动换热计算会高估燃料元件温度，降低经济性，若不考虑保守的计算又会降低反应堆的安全性，因此准确的两相流动换热计算能够在保证核动力系统安全性的同时提高经济性。In the current thermal-hydraulic program, the empirical relational formula obtained by the above method is usually used for calculation. Different empirical relational formulas have a small scope of application and have certain errors. Therefore, various empirical relational formulas will be considered comprehensively, and the most conservative model will be used for calculation. ; but too conservative two-phase flow heat transfer calculation will overestimate the temperature of fuel elements and reduce the economy, and if the conservative calculation is not considered, it will also reduce the safety of the reactor. System security while improving economy.
发明内容Contents of the invention
为了克服上述现有技术存在的问题，本发明的目的在于提供一种基于神经网络的两相流动换热模型的应用方法，通过建立神经网络模型预测两相流动换热关键参数并将其应用在热工水力程序中，克服了传统热工水力程序中两相流动换热模型适用范围窄、计算误差大的缺陷，扩大适用范围的同时能够提升计算精度；In order to overcome the problems existing in the above-mentioned prior art, the object of the present invention is to provide an application method of a two-phase flow heat transfer model based on a neural network, by establishing a neural network model to predict the key parameters of the two-phase flow heat transfer and applying it in In the thermal-hydraulic program, it overcomes the shortcomings of the narrow application range and large calculation error of the two-phase flow heat transfer model in the traditional thermal-hydraulic program, and can improve the calculation accuracy while expanding the application range;
为了实现上述目的，本发明采取以下技术方案予以实施：In order to achieve the above object, the present invention takes the following technical solutions to implement:
一种基于神经网络的两相流动换热模型的应用方法，通过建立神经网络模型预测两相流动换热关键参数并将其应用在热工水力程序中，克服了传统热工水力程序中两相流动换热模型适用范围窄、计算误差大的缺陷，扩大适用范围的同时能够提升计算精度；An application method of a two-phase flow heat transfer model based on a neural network. By establishing a neural network model to predict the key parameters of two-phase flow heat transfer and applying it to a thermal-hydraulic program, it overcomes the problem of two-phase flow in a traditional thermal-hydraulic program. The defects of the narrow application range and large calculation error of the flow heat transfer model can improve the calculation accuracy while expanding the application range;
该方法包含以下步骤：The method includes the following steps:
步骤1：收集两相流动换热实验数据，建立两相流动换热模型数据库；两相流动换热特性的关键参数为换热系数、临界热流密度和流型，因此需要收集两相流动换热的换热系数、临界热流密度和流型划分的实验数据，将不同来源的实验数据进行预处理，将单位统一为国际单位制并去除明显错误的数据点；Step 1: Collect experimental data of two-phase flow heat transfer and establish a database of two-phase flow heat transfer models; the key parameters of two-phase flow heat transfer characteristics are heat transfer coefficient, critical heat flux density and flow pattern, so it is necessary to collect two-phase flow heat transfer The experimental data of the heat transfer coefficient, critical heat flux and flow pattern division, preprocess the experimental data from different sources, unify the units into the international system of units and remove obviously wrong data points;
对于两相流动换热的换热系数、由于两相流动换热中核态沸腾为壁面上生成气泡，然而在壁面热流密度大的情况下壁面会瞬间形成气膜，因此核态沸腾换热系数和膜态沸腾换热系数会有差异，需要分别建立核态沸腾换热系数和膜态沸腾换热系数数据库；For the heat transfer coefficient of two-phase flow heat transfer, nucleate boiling generates bubbles on the wall surface in two-phase flow heat transfer, but when the wall heat flux density is high, the wall surface will instantly form a gas film, so the nucleate boiling heat transfer coefficient and The film boiling heat transfer coefficient will be different, and it is necessary to establish the nucleate boiling heat transfer coefficient and the film boiling heat transfer coefficient database respectively;
步骤2：确定预测两相流动换热关键参数所需的特征参数；根据两相流动换热特性，将两相流动的核态沸腾换热系数、膜态沸腾换热系数、临界热流密度和流型作为预测参数，分别选择对应的特征参数；Step 2: Determine the characteristic parameters required to predict the key parameters of the two-phase flow heat transfer; according to the heat transfer characteristics of the two-phase flow, the nucleate boiling heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flux The type is used as a prediction parameter, and the corresponding feature parameters are selected respectively;
对于两相流动换热的核态沸腾换热系数和膜态沸腾换热系数预测，由于两相流动换热实验中无法对液相流速和气相流速分别测量，因此核态沸腾换热系数和膜态沸腾换热系数的特征参数均为：压力、含气率、质量流密度、管内径和无量纲物性参数，无量纲物性参数能够扩大两相流动换热模型的适用范围，从单一工质扩展至多种工质，无量纲参数如式(1)所示：For the prediction of nucleate boiling heat transfer coefficient and film boiling heat transfer coefficient of two-phase flow heat transfer, since the liquid phase flow rate and gas phase flow rate cannot be measured separately in the two-phase flow heat transfer experiment, the nucleate boiling heat transfer coefficient and film boiling heat transfer coefficient can not be measured separately. The characteristic parameters of the state boiling heat transfer coefficient are: pressure, gas fraction, mass flow density, tube inner diameter, and dimensionless physical property parameters. Dimensionless physical property parameters can expand the scope of application of the two-phase flow heat transfer model, expanding to a variety of working fluids, the dimensionless parameters are shown in formula (1):
式(1)中：In formula (1):
cpl——液相定压比热/kJ·kg-1·K-1cpl ——Specific heat of liquid phase at constant pressure/kJ kg-1 K-1
μl——液相动力粘度/Pa·sμl ——liquid phase dynamic viscosity/Pa·s
kl——液相热导率/W·m-1·K-1kl ——thermal conductivity of liquid phase/W m-1 K-1
对于两相流动换热的临界热流密度预测，选择的特征参数为：压力、含气率和质量流密度；For the critical heat flux prediction of two-phase flow heat transfer, the selected characteristic parameters are: pressure, gas fraction and mass flow density;
对于两相流动换热的流型预测，选择的特征参数为：空泡份额、液相流速和气相流速；流型种类分为：泡状流、弹状流、过渡流、环状流和弥散流；For the flow pattern prediction of two-phase flow heat transfer, the selected characteristic parameters are: cavity fraction, liquid phase flow rate and gas phase flow rate; flow type types are divided into: bubbly flow, slug flow, transition flow, annular flow and dispersion flow;
步骤3：建立预测参数的神经网络模型并使用建立的数据库对其进行训练；Step 3: Establish a neural network model for predicting parameters and use the established database to train it;
建立预测核态换热系数、膜态沸腾换热系数、临界热流密度和流型的神经网络模型；Establish neural network models for predicting nucleation heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flow patterns;
以上四个神经网络模型的建立方法相同，需要确定输入层节点、输出层节点、隐含层节点，将步骤2中的特征参数作为神经网络的输入层节点，将待预测参数作为神经网络的输出层节点，为保证神经网络模型预测精度，隐含层层数选择为2层，隐含层神经元数量根据式(2)确定：The establishment methods of the above four neural network models are the same. It is necessary to determine the input layer nodes, output layer nodes, and hidden layer nodes. The characteristic parameters in step 2 are used as the input layer nodes of the neural network, and the parameters to be predicted are used as the output of the neural network. Layer nodes, in order to ensure the prediction accuracy of the neural network model, the number of hidden layers is selected as 2 layers, and the number of neurons in the hidden layer is determined according to formula (2):
式(2)中：In formula (2):
Nh——隐含层神经元数量；Nh - the number of neurons in the hidden layer;
Ni——输入层节点数量；Ni ——the number of input layer nodes;
No——输出层节点数量；No ——the number of output layer nodes;
α——调节常数，取1～10；α——Adjustment constant, ranging from 1 to 10;
从步骤1建立的数据库中选择的相应的实验数据并进行归一化处理，将处理后的数据作为神经网络模型的训练数据，将所有数据划分为训练集、验证集和测试集，对神经网络模型进行训练，使其能够在输入特征参数后输出相应的预测参数；Select the corresponding experimental data from the database established in step 1 and perform normalization processing, use the processed data as the training data of the neural network model, divide all the data into training set, verification set and test set, and perform neural network The model is trained so that it can output the corresponding prediction parameters after inputting the feature parameters;
基于上述方法建立了预测核态沸腾换热系数、膜态沸腾换热系数、临界热流密度和流型的神经网络模型，便能够开展基于神经网络的两相流动换热计算，基于神经网络的两相流动换热模型计算逻辑如下：首先进行临界热流密度的预测，若实际热流密度小于临界热流密度，认为当前状态为核态沸腾，采用核态沸腾换热系数神经网络模型进行计算；若实际热流密度大于临界热流密度，认为当前状态为膜态沸腾，采用膜态沸腾换热系数神经网络模型进行计算；计算完成后进行流型预测，若采用核态沸腾换热系数神经网络模型进行计算，流型必须为泡状流、弹状流和过渡流其中一种，若采用膜态沸腾换热系数神经网络模型进行计算，流型必须为过渡流、环状流和弥散流其中一种，否则重新进行临界热流密度的预测；Based on the above method, a neural network model for predicting nucleate boiling heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flow pattern can be established, and the calculation of two-phase flow heat transfer based on neural network can be carried out. The calculation logic of the phase flow heat transfer model is as follows: firstly, the critical heat flux is predicted. If the actual heat flux is less than the critical heat flux, the current state is considered to be nucleate boiling, and the nucleate boiling heat transfer coefficient neural network model is used for calculation; if the actual heat flux If the density is greater than the critical heat flux density, the current state is considered to be film boiling, and the film boiling heat transfer coefficient neural network model is used for calculation; after the calculation is completed, the flow pattern is predicted. If the nucleate boiling heat transfer coefficient neural network model is used for calculation, the flow The flow type must be one of bubbly flow, slug flow, and transitional flow. If the film boiling heat transfer coefficient neural network model is used for calculation, the flow type must be one of transitional flow, annular flow, and diffuse flow. Otherwise, restart Predict the critical heat flux;
步骤4：在热工水力程序中添加耦合接口，对神经网络模型进行调用；在热工水力程序的本构方程中增加基于神经网络的两相流动换热模型，在程序输入中选择是否调用神经网络模型计算；Step 4: Add a coupling interface to the thermal-hydraulic program to call the neural network model; add a neural network-based two-phase flow heat transfer model to the constitutive equation of the thermal-hydraulic program, and select whether to call the neural network model in the program input Network model calculation;
步骤5：修改热工水力程序计算逻辑，在热工水力程序计算时，首先判断是否为两相流动换热计算，若是两相流动换热计算，则在热工水力程序中向神经网络模型输入步骤2中的特征参数开展热工水力计算；若为单相流动换热计算，则采用热工水力程序原有的单相计算关系式；Step 5: Modify the calculation logic of the thermal-hydraulic program. When calculating the thermal-hydraulic program, first judge whether it is a two-phase flow heat transfer calculation. If it is a two-phase flow heat transfer calculation, input it to the neural network model in the thermal-hydraulic program. The characteristic parameters in step 2 are used for thermal-hydraulic calculation; if it is a single-phase flow heat transfer calculation, the original single-phase calculation relationship of the thermal-hydraulic program is used;
步骤6：判断计算结果是否收敛，若通过以上流动换热计算得到的壁面温度高于材料熔点或壁面温度低于流体温度，认为计算结果不收敛，则对输入参数迭代后重复步骤5，若通过以上流动换热计算得到的壁面温度低于材料熔点且不低于流体温度，认为计算结果收敛，则结束计算。Step 6: Determine whether the calculation result is convergent. If the wall temperature obtained through the above flow heat transfer calculation is higher than the melting point of the material or the wall temperature is lower than the fluid temperature, it is considered that the calculation result is not convergent, and then repeat step 5 after iterating the input parameters. The wall temperature obtained by the above flow heat transfer calculation is lower than the melting point of the material and not lower than the fluid temperature. If the calculation result is considered to be convergent, the calculation will end.
和现有技术相比较，本发明具备如下优点：Compared with the prior art, the present invention has the following advantages:
1.通过在训练数据库中补充新的实验数据的方式对神经网络模型进行优化，相较传统的经验关系式，过程更简单方便；1. The neural network model is optimized by supplementing new experimental data in the training database. Compared with the traditional empirical relational formula, the process is simpler and more convenient;
2.省去了传统热工水力程序中对于两相流动换热计算繁琐的公式选择，适用范围广、计算精度高；2. Eliminates the cumbersome formula selection for two-phase flow heat transfer calculations in traditional thermal hydraulic programs, with a wide range of applications and high calculation accuracy;
3.将神经网络模型耦合到热工水力程序后，后续对程序进行优化只需要替换神经网络模型，不需要再对程序进行修改，节省时间。3. After the neural network model is coupled to the thermal hydraulic program, the subsequent optimization of the program only needs to replace the neural network model, and does not need to modify the program, saving time.
附图说明Description of drawings
图1为本发明实施的流程框图。Fig. 1 is a flow chart of the implementation of the present invention.
图2为两相流动关键参数预测的神经网络结构。Figure 2 shows the neural network structure for the prediction of key parameters of two-phase flow.
图3为基于神经网络的两相流动换热模型计算逻辑。Fig. 3 is the calculation logic of the two-phase flow heat transfer model based on the neural network.
具体实施方式Detailed ways
下面结合实例、附图对本发明作进一步描述：Below in conjunction with example, accompanying drawing, the present invention will be further described:
如图1所示，以建立基于神经网络的两相流动换热模型并将其应用在热工水力程序中为例，该方法包含以下步骤：As shown in Figure 1, taking the establishment of a neural network-based two-phase flow heat transfer model and applying it to a thermal-hydraulic program as an example, the method includes the following steps:
步骤1：收集两相流动换热实验数据，建立两相流动换热模型数据库；两相流动换热特性的关键参数为换热系数、临界热流密度和流型，因此需要从公开文献、报告和已有的数据库中收集两相流动换热的换热系数、临界热流密度和流型划分的实验数据；换热系数和流型的数据来源为公开文献和报告，临界热流密度的数据主要为临界热流密度查询表，选择其他公开文献中的数据作为不足部分的补充；将不同来源的实验数据进行预处理，将单位统一为国际单位制，去除临界热流密度查询表中误差较大的数据和文献中错误的数据；Step 1: Collect experimental data of two-phase flow heat transfer and establish a database of two-phase flow heat transfer models; The experimental data of heat transfer coefficient, critical heat flux and flow pattern division of two-phase flow heat transfer are collected in the existing database; the data sources of heat transfer coefficient and flow pattern are public documents and reports, and the data of critical heat flux are mainly critical Heat flux look-up table, select data from other public documents as a supplement to the insufficient part; preprocess the experimental data from different sources, unify the unit into the international system of units, and remove the data and literature with large errors in the critical heat flux look-up table wrong data in
对于两相流动换热的换热系数、由于两相流动换热中核态沸腾为壁面上生成气泡，然而在壁面热流密度大的情况下壁面会瞬间形成气膜，因此核态沸腾换热系数和膜态沸腾换热系数会有差异，需要分别建立核态沸腾换热系数和膜态沸腾换热系数数据库；For the heat transfer coefficient of two-phase flow heat transfer, nucleate boiling generates bubbles on the wall surface in two-phase flow heat transfer, but when the wall heat flux density is high, the wall surface will instantly form a gas film, so the nucleate boiling heat transfer coefficient and The film boiling heat transfer coefficient will be different, and it is necessary to establish the nucleate boiling heat transfer coefficient and the film boiling heat transfer coefficient database respectively;
步骤2：确定预测两相流动换热关键参数所需的特征参数；根据两相流动换热特性，将两相流动的核态沸腾换热系数、膜态沸腾换热系数、临界热流密度和流型作为预测参数，分别选择对应的特征参数；Step 2: Determine the characteristic parameters required to predict the key parameters of the two-phase flow heat transfer; according to the heat transfer characteristics of the two-phase flow, the nucleate boiling heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flux The type is used as a prediction parameter, and the corresponding feature parameters are selected respectively;
对于两相流动换热的核态沸腾换热系数和膜态沸腾换热系数预测，由于两相流动换热实验中无法对液相流速和气相流速分别测量，因此核态沸腾换热系数和膜态沸腾换热系数的特征参数为：压力、含气率、质量流密度、管内径和无量纲物性参数，无量纲物性参数可以扩大模型两相流动换热的适用范围，从单一工质扩展至多种工质，无量纲参数如式(1)所示：For the prediction of nucleate boiling heat transfer coefficient and film boiling heat transfer coefficient of two-phase flow heat transfer, since the liquid phase flow rate and gas phase flow rate cannot be measured separately in the two-phase flow heat transfer experiment, the nucleate boiling heat transfer coefficient and film boiling heat transfer coefficient can not be measured separately. The characteristic parameters of the state boiling heat transfer coefficient are: pressure, gas fraction, mass flow density, tube inner diameter and dimensionless physical property parameters. The dimensionless physical property parameters can expand the applicable range of the two-phase flow heat transfer model from a single working fluid to multiple Working fluid, the dimensionless parameters are shown in formula (1):
式(1)中：In formula (1):
cpl——液相定压比热/kJ·kg-1·K-1cpl ——Specific heat of liquid phase at constant pressure/kJ kg-1 K-1
μl——液相动力粘度/Pa·sμl ——liquid phase dynamic viscosity/Pa·s
kl——液相热导率/W·m-1·K-1kl ——thermal conductivity of liquid phase/W m-1 K-1
对于两相流动换热的临界热流密度预测，选择的特征参数为：压力、含气率和质量流密度；For the critical heat flux prediction of two-phase flow heat transfer, the selected characteristic parameters are: pressure, gas fraction and mass flow density;
对于两相流动换热的流型预测，选择的特征参数为：空泡份额、液相流速和气相流速；流型种类分为：泡状流、弹状流、过渡流、环状流和弥散流；For the flow pattern prediction of two-phase flow heat transfer, the selected characteristic parameters are: cavity fraction, liquid phase flow rate and gas phase flow rate; flow type types are divided into: bubbly flow, slug flow, transition flow, annular flow and dispersion flow;
步骤3：建立预测参数的神经网络模型并使用建立的数据库对其进行训练；Step 3: Establish a neural network model for predicting parameters and use the established database to train it;
建立预测核态换热系数、膜态沸腾换热系数、临界热流密度和流型的神经网络模型；Establish neural network models for predicting nucleate heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flow patterns;
以上四个神经网络模型的建立方法相同，需要确定输入层节点、输出层节点、隐含层节点，将步骤2中的特征参数作为神经网络的输入层节点，将待预测参数作为神经网络的输出层节点，建立图2所示的四层神经网络，包括输入层、2层隐含层和输出层；从步骤1建立的数据库中选择的相应的实验数据并进行归一化处理，归一化处理公式如式(2)所示：The establishment methods of the above four neural network models are the same. It is necessary to determine the input layer nodes, output layer nodes, and hidden layer nodes. The characteristic parameters in step 2 are used as the input layer nodes of the neural network, and the parameters to be predicted are used as the output of the neural network. Layer node, set up the four-layer neural network shown in Figure 2, including the input layer, 2 hidden layers and the output layer; select the corresponding experimental data from the database established in step 1 and carry out normalization processing, normalization The processing formula is shown in formula (2):
式(2)中：In formula (2):
Xnorm(i)——归一化后的数据Xnorm (i) - normalized data
X(i)——原始数据X(i) - raw data
Xmin——原始数据中的最小值Xmin - the minimum value in the original data
Xmax——原始数据中的最大值Xmax - the maximum value in the original data
将处理后的数据作为神经网络模型的训练数据，将所有数据按照0.7：0.15：0.15的比例划分为训练集、验证集和测试集，对神经网络模型进行训练，使其能够在输入特征参数后输出相应的预测参数；The processed data is used as the training data of the neural network model, and all the data are divided into training set, verification set and test set according to the ratio of 0.7:0.15:0.15, and the neural network model is trained so that it can input the characteristic parameters. Output the corresponding prediction parameters;
基于上述方法建立了预测核态沸腾换热系数、膜态沸腾换热系数、临界热流密度和流型的神经网络模型，便能够开展基于神经网络的两相流动换热计算，基于神经网络的两相流动换热模型计算逻辑如图3所示，首先进行临界热流密度的预测，若实际热流密度小于临界热流密度，认为当前状态为核态沸腾，采用核态沸腾换热系数神经网络模型进行计算；若实际热流密度大于临界热流密度，认为当前状态为膜态沸腾，采用膜态沸腾换热系数神经网络模型进行计算；计算完成后进行流型预测，若采用核态沸腾换热系数神经网络模型进行计算，流型必须为泡状流、弹状流和过渡流其中一种，若采用膜态沸腾换热系数神经网络模型进行计算，流型必须为过渡流、环状流和弥散流其中一种，否则重新进行临界热流密度的预测；Based on the above method, a neural network model for predicting nucleate boiling heat transfer coefficient, film boiling heat transfer coefficient, critical heat flux and flow pattern can be established, and the calculation of two-phase flow heat transfer based on neural network can be carried out. The calculation logic of the phase flow heat transfer model is shown in Figure 3. First, the critical heat flux is predicted. If the actual heat flux is less than the critical heat flux, the current state is considered to be nucleate boiling, and the nucleate boiling heat transfer coefficient neural network model is used for calculation. ; If the actual heat flux is greater than the critical heat flux, the current state is considered to be film boiling, and the film boiling heat transfer coefficient neural network model is used for calculation; after the calculation is completed, the flow pattern is predicted. If the nucleate boiling heat transfer coefficient neural network model is used For calculation, the flow pattern must be one of bubbly flow, slug flow and transition flow. If the film boiling heat transfer coefficient neural network model is used for calculation, the flow pattern must be one of transition flow, annular flow and diffuse flow Otherwise, re-predict the critical heat flux;
步骤4：在热工水力程序中添加耦合接口，对神经网络模型进行调用；在热工水力程序的本构方程中增加基于神经网络的两相流动换热模型，在程序输入中选择是否调用神经网络模型计算；神经网络模型由Python编写生成，核工程领域主流热工水力程序为Fortran编写，采用在源码中增加对神经网络计算模型的调用的方法进行耦合；Step 4: Add a coupling interface to the thermal-hydraulic program to call the neural network model; add a neural network-based two-phase flow heat transfer model to the constitutive equation of the thermal-hydraulic program, and select whether to call the neural network model in the program input Network model calculation; the neural network model is written and generated by Python, and the mainstream thermal hydraulic program in the field of nuclear engineering is written in Fortran, and the method of adding calls to the neural network calculation model in the source code is used for coupling;
步骤5：修改热工水力程序计算逻辑，在热工水力程序计算时，首先判断是否为两相流动换热计算，若是两相流动换热计算，则在热工水力程序中向神经网络模型输入步骤2中的特征参数开展热工水力计算；若为单相流动换热计算，则采用热工水力程序原有的单相计算关系式；Step 5: Modify the calculation logic of the thermal-hydraulic program. When calculating the thermal-hydraulic program, first judge whether it is a two-phase flow heat transfer calculation. If it is a two-phase flow heat transfer calculation, input it to the neural network model in the thermal-hydraulic program. The characteristic parameters in step 2 are used for thermal-hydraulic calculation; if it is a single-phase flow heat transfer calculation, the original single-phase calculation relationship of the thermal-hydraulic program is used;
步骤6：判断计算结果是否收敛，若通过以上流动换热计算得到的壁面温度高于材料熔点或壁面温度低于流体温度，认为计算结果不收敛，则对输入参数迭代后重复步骤5，若通过以上流动换热计算得到的壁面温度低于材料熔点且不低于流体温度，认为计算结果收敛，则结束计算。Step 6: Determine whether the calculation result is convergent. If the wall temperature obtained through the above flow heat transfer calculation is higher than the melting point of the material or the wall temperature is lower than the fluid temperature, it is considered that the calculation result is not convergent, and then repeat step 5 after iterating the input parameters. The wall temperature obtained by the above flow heat transfer calculation is lower than the melting point of the material and not lower than the fluid temperature. If the calculation result is considered to be convergent, the calculation will end.
本发明未详细写明的内容均为本领域的公知常识。The contents not described in detail in the present invention are common knowledge in this field.
Claims (1)
Translated fromChinese
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210068693.2ACN114417723B (en) | 2022-01-20 | 2022-01-20 | An application method of two-phase flow heat transfer model based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210068693.2ACN114417723B (en) | 2022-01-20 | 2022-01-20 | An application method of two-phase flow heat transfer model based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114417723A CN114417723A (en) | 2022-04-29 |
| CN114417723Btrue CN114417723B (en) | 2023-05-30 |
Family
ID=81275167
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210068693.2AExpired - Fee RelatedCN114417723B (en) | 2022-01-20 | 2022-01-20 | An application method of two-phase flow heat transfer model based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114417723B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116384248B (en)* | 2023-04-14 | 2023-11-21 | 西安交通大学 | LBM two-phase flow and phase change numerical simulation method based on artificial neural network to predict physical properties |
| CN119150733A (en)* | 2024-08-27 | 2024-12-17 | 东南大学 | Pier scouring multiphase flow calculation method and system based on physical information neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5050114A (en)* | 1990-09-17 | 1991-09-17 | Motorola, Inc. | Simulation of two-phase liquid cooling for thermal prediction of direct liquid cooling schemes |
| CN1664555A (en)* | 2005-03-17 | 2005-09-07 | 上海交通大学 | Two-Phase Flow Recognition Method Based on Time Series and Neural Network Pattern Recognition |
| CN102419827A (en)* | 2011-11-02 | 2012-04-18 | 昆明理工大学 | Radial basis function (RBF) neural network-based boiling heat exchanging prediction method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8196395B2 (en)* | 2009-06-29 | 2012-06-12 | Lightsail Energy, Inc. | Compressed air energy storage system utilizing two-phase flow to facilitate heat exchange |
- 2022
- 2022-01-20CNCN202210068693.2Apatent/CN114417723B/ennot_activeExpired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5050114A (en)* | 1990-09-17 | 1991-09-17 | Motorola, Inc. | Simulation of two-phase liquid cooling for thermal prediction of direct liquid cooling schemes |
| CN1664555A (en)* | 2005-03-17 | 2005-09-07 | 上海交通大学 | Two-Phase Flow Recognition Method Based on Time Series and Neural Network Pattern Recognition |
| CN102419827A (en)* | 2011-11-02 | 2012-04-18 | 昆明理工大学 | Radial basis function (RBF) neural network-based boiling heat exchanging prediction method |
Non-Patent Citations (1)
| Title |
|---|
| 振荡热管研究进展及展望;崔晓钰;黄万鹏;翁建华;马虎根;;电子机械工程(01);全文* |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114417723A (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114417723B (en) | An application method of two-phase flow heat transfer model based on neural network | |
| CN108170924B (en) | Method for establishing nuclear power plant steam generator heat transfer pipe flow blockage working condition model | |
| CN111706785B (en) | Natural gas dendritic pipe network leakage pipe section identification method | |
| CN115859851B (en) | Calculation method for conjugate heat transfer of liquid metal coupling supercritical carbon dioxide | |
| CN112069692A (en) | Optimization solving method for natural gas pipe network transmission difference calculation | |
| CN106055520B (en) | System and method for predicting drum water level of circulating fluidized bed domestic garbage incineration boiler | |
| CN114491817B (en) | Main pipe simulation method and system in ship nuclear power pipe network simulation | |
| CN112231990A (en) | A method for calculating the coupling of primary and secondary loops in nuclear power systems | |
| CN115862779A (en) | Liquid metal conjugate heat transfer calculation method based on two-equation thermal turbulence model | |
| CN106777770A (en) | The analogy method of hole stream in aqueduct based on finite volume method | |
| Fang et al. | Data assimilation of steam flow through a control valve using ensemble Kalman filter | |
| Zhang et al. | Study of the heat transfer deterioration of supercritical CO2 in vertical pipes using a hybrid RANS/LES method | |
| Sharp et al. | Optimizing the ASME Venturi recovery cone angle to minimize head loss | |
| CN119004951B (en) | Pressurized water reactor nuclear fuel assembly reliability optimization method and system based on transfer learning | |
| Hao et al. | An inverse transient nonmetallic pipeline leakage diagnosis method based on Markov quantitative judgment | |
| CN112417680B (en) | Method and system for estimating mass gas holdup distribution of working fluid in secondary circuit of steam generator | |
| CN114925598A (en) | Tubular heat exchanger heat exchange error reliability analysis method based on Kriging method | |
| CN111753418B (en) | A method for analyzing the migration path of vapor bubbles generated by lead-based reactor accidents | |
| Yang et al. | Les of a Space-Developing Natural Convection Boundary Layer in Water Along a Vertical Wall with High Heat Flux | |
| CN116341399A (en) | A Prediction Method of Thermal-hydraulic Heat Transfer Coefficient Based on Physically Constrained Neural Network | |
| Li et al. | The Inlet Flow Blockage Accidents Analysis in the Rectangular flow channel of water cooled blanket | |
| CN1936907A (en) | Method for dynamically setting up mould of parallely operated machine group steam mather tube | |
| Duan et al. | Prediction method for burst pressure of oil and gas pipelines based on GA-BP algorithm | |
| Khalid et al. | Comparison of Standalone and Hybrid Machine Learning Models for Prediction of Critical Heat Flux in Vertical Tubes. Energies 2023, 16, 3182 | |
| Jiao et al. | A Flow State Estimation Framework Based on a Novel Natural Gas Pipeline Network Transient Simulation Method |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20230530 | |
| CF01 | Termination of patent right due to non-payment of annual fee |