Application of artificial neural networks in analysis of CHF experimental data in round tubes

Artificial neural networks (ANNs) are applied successfully to analyze the critical heat flux (CHF) experimental data from some round tubes in this paper. A set of software adopting artificial neural network method for predicting CHF in round tube and a set of CHF database are gotten. Comparing with common CHF correlations and CHF look-up table, ANN method has stronger ability of allow-wrong and nice robustness. The CHF predicting software adopting artificial neural network technology can improve the predicting accuracy in a wider parameter range, and is easier to update and to use. The artificial neural nefwork method used in this paper can be applied to some similar physical problems.     

基于高斯过程回归的临界热流密度预测

准确地预测临界热流密度（CHF）对于反应堆的安全和运行十分重要。针对现有人工神经网络（ANNs）预测方法所存在的缺点，提出一种基于高斯过程回归（GPR）的CHF预测方法。首先对获取的当地条件下CHF数据进行预处理，将数据划分为训练集和测试集；然后，利用训练数据对GPR模型进行训练，并得到最优超参数；再利用训练好的GPR模型对CHF进行预测，并将结果与径向基神经网络（RBFNN）进行比较，同时分析了重要参数对CHF的影响趋势。结果表明，与RBFNN相比，GPR模型的预测结果具有更高的预测精度和更小的误差，且与对应的实验值吻合较好，其参数趋势符合通用的趋势变化规律。      

Critical heat flux prediction by using radial basis function and multilayer perceptron neural networks: A comparison study

Critical heat flux (CHF) is an important parameter for the design of nuclear reactors. Although many experimental and theoretical researches have been performed, there is not a single correlation to predict CHF because it is influenced by many parameters. These parameters are based on fixed inlet, local and fixed outlet conditions. Artificial neural networks (ANNs) have been applied to a wide variety of different areas such as prediction, approximation, modeling and classification. In this study, two types of neural networks, radial basis function (RBF) and multilayer perceptron (MLP), are trained with the experimental CHF data and their performances are compared. RBF predicts CHF with root mean square (RMS) errors of 0.24%, 7.9%, 0.16% and MLP predicts CHF with RMS errors of 1.29%, 8.31% and 2.71%, in fixed inlet conditions, local conditions and fixed outlet conditions, respectively. The results show that neural networks with RBF structure have superior performance in CHF data prediction over MLP neural networks. The parametric trends of CHF obtained by the trained ANNs are also evaluated and results reported.     

Prediction of critical heat flux for water flow in vertical round tubes using support vector regression model

This paper deals with ν-support vector regression (ν-SVR) based prediction model of critical heat flux (CHF) for water flow in vertical round tubes. The dataset used in this paper is obtained from available published literature. The dataset is partitioned into two independent sets, a training data set and a test data set, to avoid overfitting problem. To train the ν-SVR models with more informative data, the training data is selected using a subtractive clustering (SC) scheme, and then the remaining data is used as test data to evaluate the performance of the ν-SVR models. Next, the parametric trends of CHF are investigated using the ν-SVR models. The results obtained from the ν-SVR models are compared with those obtained from the radial basis function (RBF) network, which is a kind of artificial neural networks (ANNs). It is found that the results of the ν-SVR models are not only in better agreement with the experimental data than those of the RBF network, but also follow the general understanding. The analysis results indicate that the ν-SVR models can be successfully applied to CHF prediction.     

Prediction of residual stress in the welding zone of dissimilar metals using data-based models and uncertainty analysis

Since welding residual stress is one of the major factors in the generation of primary water stress-corrosion cracking (PWSCC), it is essential to examine the welding residual stress to prevent PWSCC. Therefore, several artificial intelligence methods have been developed and studied to predict these residual stresses. In this study, three data-based models, support vector regression (SVR), fuzzy neural network (FNN), and their combined (FNN + SVR) models were used to predict the residual stress for dissimilar metal welding under a variety of welding conditions. By using a subtractive clustering (SC) method, informative data that demonstrate the characteristic behavior of the system were selected to train the models from the numerical data obtained from finite element analysis under a range of welding conditions. The FNN model was optimized using a genetic algorithm. The statistical and analytical uncertainty analysis methods of the models were applied, and their uncertainties were evaluated using 60 sampled training and optimization data sets, as well as a fixed test data set.     

Parametric trends analysis of the critical heat flux based on artificial neural networks

Parametric trends of the critical heat flux (CHF) are analyzed by applying artificial neural networks (ANNs) to a CHF data base for upward flow of water in uniformly heated vertical round tubes. The analyses are performed from three viewpoints, i.e., for fixed inlet conditions, for fixed exit conditions, and based on local conditions hypothesis. Katto's and Groeneveld et al. dimensionless parameters are used to train the ANNs with the experimental CHF data. The trained ANNs predict the CHF better than any other conventional correlations, showing RMS errors of 8.9%, 13.1% and 19.3% for fixed inlet conditions, for fixed exit conditions, and for local conditions hypothesis, respectively. The parametric trends of the CHF obtained from those trained ANNs show a general agreement with previous understanding. In addition, this study provides more comprehensive information and indicates interesting points for the effects of the tube diameter, the heated length, and the mass flux. It is expected that better understanding of the parametric trends is feasible with an extended data base.     

人工神经网络在棒束临界热流密度预测中的应用

基于已有的棒束临界热流密度数据库，采用COBRA-Ⅳ程序计算得到子通道局部临界热流密度数据库。用人工神经网络（ANN）理论对数据库进行训练，得到基于ANN理论的棒束临界热流密度预测模型。预测模型的预测精度显著高于常用经验关系式的预测精度，其预测值的均方差为5.63%。     

临界热流密度的人工神经网络预测法

本文成功地训练了3种用于预测临界热流密度(CHF)的人工神经网络,其输入参数分别是系统压力、质量流速、平衡含汽量;其输出参数是CHF.通过人工神经网络,分析了压力、流量、热平衡含汽量和进口过冷度对CHF的影响,且成功地将人工神经网络应用于CHF的预测中,预测结果与实验值符合很好.分析结果表明:人工神经网络训练的3种类型中,类型Ⅱ的预测精度最高,可达±10%.     

内螺纹管临界热流密度预测与数值计算

基于壁面汽泡壅塞理论，针对近临界压力区两相流动沸腾的偏离泡核沸腾(DNB)现象，对垂直上升内螺纹管的DNB型临界热流密度(CHF)进行了数值计算研究。以内螺纹管为分析对象改进已有的汽泡壅塞模型，计算了汽泡层区与主流区的极限传递质量流量、湍流速度分布、汽泡层区临界截面含气率等本构关系，汽泡脱离直径的计算考虑了汽泡接触角的影响。本文模型还根据大量CHF实验数据拟合得到了新的α_b关联式。最后，基于Fortran语言编制了CHF的理论预测数值模型程序，研究分析了压力、质量流速、热平衡干度及进口欠焓对CHF的影响，并根据CHF查表值对本文模型进行评估，同时将实验得到的内螺纹管CHF数据与采用Bowring模型、Katto模型、Shah模型和本文模型计算的CHF进行比较，发现本文模型的误差最小，与实验值吻合结果较好，说明本文模型能较好地对垂直上升内螺纹管DNB型CHF进行预测。     

基于神经网络的人工放射性气溶胶中氡子体扣除算法

介绍和分析了人工放射性气溶胶在线监测仪氡子体扣除算法中比例系数扣除法,现有算法存在分类粗糙、扣除准确度不高以及适应性不强等不足。为进一步提高扣除的准确度,降低检测限,提出了利用聚类分析先对谱线进行分类,然后在每个类中利用神经网络进行计算,最后进行扣除的方法。测试结果证明了聚类分析和神经网络扣除方法均能明显降低人工放射性气溶胶在线监测仪的检测限。     

Critical heat flux for finned and unfinned geometries under low flow and low pressure conditions

This paper presents the experiment and analysis for the critical heat flux (CHF) in a vertical annulus with finned and unfinned geometries under low flow and low pressure conditions. To consider the fin effect on CHF, the tests were performed on both finned heater and unfinned heater having same dimension as finned heater without fins. An analytical model was applied to estimate the heat flux and temperature distributions along the periphery of the finned geometry. The physical phenomena observed during the experiments are discussed and the parametric trends of the obtained data are examined to investigate the CHF characteristics for the finned geometry. A new correlation is proposed to predict the CHF for both finned and unfinned geometries at low flow and low pressure conditions. The developed correlation predicts the experimental data with an RMS error of 13.7%.     

人工神经网络及其在地学中的应用综述

人工神经网络(ANN)是一种模仿人脑神经网络结构和功能的信息处理系统,是一种分布式并行处理信息的抽象数学模型,现已在许多科学领域得以成功应用。在地球科学领域,人工神经网络最早应用于地球物理反演问题,随后逐渐扩展至其他领域。通过简要介绍人工神经网络的发展历程、基本特征及其模型,对地学领域中常用的人工神经网络模型进行了简单对比,并论述了其在地学领域中的应用特点,总结了近年来人工神经网络在地学领域中的主要应用,着重从判别分类、模式识别、预测评价以及信息数据处理等方面的应用进行了详细阐述。同时,结合地学领域的实际特点和人工智能领域中大量出现的优化理论和技术,分析认为人工神经网络在地学领域中的应用将逐渐呈现多种技术和深度学习的相融合态势,且在地学领域中应用效果会日益显著。这些探讨和分析对推动地学工作数字化、智能化具有参考意义。     

神经网络在CHF预测中的应用

利用人工神经网络理论对均匀加热垂直上升圆管内的临界热流密度（ＣＨＦ）进行预测和参数趋势分析。本研究采用局部条件假设,并选用Ｃｒｏｅｎｅｖｌｄ的ＣＨＦ查询表数据为本文神经网络训练的样本,采用训练成功的网络预测ＣＨＦ值可以得到比常规方法更好的效果,其均方差为１４．９％。     

基于v-支持向量机的事故工况下反应堆功率预测

针对事故工况下堆芯功率变化的特点和神经网络（ANNs）易陷极小值、收敛速度慢等问题，提出一种基于ν-SVR）的事故工况下堆芯功率预测方法。该方法运用k重交叉验证（k-CV）完成对ν-SVR预测器并将其用于弹棒事故（REA）和落棒事故（RDA）工况下的堆芯功率预测。研究表明，与ANNs相比，该方法具有更高的预测精度和更短的响应时间。      

A critical heat flux approach for square rod bundles using the 1995 Groeneveld CHF table and bundle data of heat transfer research facility

The critical heat flux (CHF) approach using CHF look-up tables has become a widely accepted CHF prediction technique. In these approaches, the CHF tables are developed based mostly on the data bank for flow in circular tubes. A set of correction factors was proposed by Groeneveld et al. [Groeneveld, D.C., Cheng, S.C., Doan, T., 1986. 1986 AECL-UO Critical Heat Flux lookup table. Heat Transf. Eng. 7(1–2), 46] to extend the application of the CHF table to other flow situations including flow in rod bundles. The proposed correction factors are based on a limited amount of data not specified in the original paper. The CHF approach of Groeneveld and co-workers is extensively used in the thermal hydraulic analysis of nuclear reactors. In 1996, Groeneveld et al. proposed a new CHF table to predict CHF in circular tubes [Groeneveld, D.C., et al., 1996. The 1995 look-up table for Critical Heat Flux. Nucl. Eng. Des. 163(1), 23]. In the present study, a set of correction factors is developed to extend the applicability of the new CHF table to flow in rod bundles of square array. The correction factors are developed by minimizing the statistical parameters of the ratio of the measured and predicted bundle CHF data from the Heat Transfer Research Facility. The proposed correction factors include: the hydraulic diameter factor (K_hy), the bundle factor (K_bf), the heated length factor (K_hl), the grid spacer factor (K_sp), the axial flux distribution factors (K_nu), the cold wall factor (K_cw) and the radial power distribution factor (K_rp). The value of constants in these correction factors is different when the heat balance method (HBM) and direct substitution method (DSM) are adopted to predict the experimental results of HTRF. With the 1995 Groeneveld CHF Table and the proposed correction factors, the average relative error is 0.1 and 0.0% for HBM and DSM, respectively, and the root mean square (RMS) error is 31.7% in DSM and 17.7% in HBM for 9852 square array data points of HTRF.     

基于ARIMA和LSTM组合模型的核电厂主泵状态预测

为了对核电厂主泵的运行过程进行监测和追踪,进而提高主泵的预警能力,提出了基于差分自回归移动平均（ARIMA）和长短期记忆（LSTM）神经网络组合模型的主泵状态预测方法,并用该方法对某核电厂主泵止推轴承温度和可控泄漏流量进行单步和多步预测,以根均方误差（RMSE）为指标对预测精度进行评估。结果表明,所建立的ARIMA和LSTM神经网络组合模型能够对主泵的状态进行准确的预测和追踪,并且组合模型的预测精度要优于ARIMA和LSTM单一模型,尤其在多步预测中,组合模型的优势更加明显。      

波形板壁面液膜的神经网络及混沌特性分析

波形板干燥器是船用核动力系统中重要的汽水分离设备，其壁面上自由下降液膜的流动特性对干燥器的汽水分离效率及船用核动力装置的安全性指标有着较大的影响。基于平面激光诱导荧光技术（PLIF）对不同雷诺数下的壁面薄层液膜厚度进行测量。通过小数据量法计算不同工况下的液膜厚度时间序列的最大Lyapunov指数，分析壁面液膜的混沌特性并进行相空间重构。利用反向传播（BP）神经网络解决非线性问题的优势对液膜厚度进行预测，完成了单隐层BP神经网络预测模型的建立并实现了自由液膜厚度的非线性特征分析。结果显示:最大Lyapunov指数与液膜雷诺数呈正相关关系；在大雷诺数区生成的孤立峰同重力及液膜间的叠加作用相互耦合，使液膜混沌特性变得更加明显。      

棒束子通道CHF机理模型开发及初步验证

目前棒束通道中临界热流密度的预测多基于实验关系式,受限于特定的适用范围,无法有效外推或外推后预测精度下降。为满足不同轻水堆中临界热流密度的预测要求,有必要开发适用于不同几何尺寸及热工边界的宽范围临界热流密度预测方式。本文以子通道分析方法为基础,考虑偏离泡核沸腾和干涸两类临界现象,通过耦合子通道分析程序与临界热流密度机理模型,实现对棒束通道中临界热流密度的计算。通过与临界热流密度实验数据的对比,初步证明了耦合程序对棒束通道中临界热流密度具有较好的预测精度。     

CHF prediction for square arrays by the bundle enthalpy imbalance method

For both economic and safety reasons, the reactor designer should be able to predict accurately the conditions under which critical heat flux (CHF) is reached. Consequently, extensive experimental and analytical studies have been carried out in many nations. The laboratory work discussed in this paper has involved tests on round tubes, annuli, rod bundles having from three to at least 49 rods, and other geometries, and very accurate data have been obtained. This paper evaluates the McPherson method for CHF prediction for rods in square channels or arrays.     

Application of neural networks to signal prediction in nuclearpower plant

The feasibility of using an artificial neural network for signal prediction is studied. The purpose of signal prediction is to estimate the value of the undetected next-time-step signal. In the prediction method, which is based on the idea of autoregression, a few previous signals are input to the artificial neural network, and the signal value of next time step is estimated from the outputs of the network. The artificial neural network can be applied to a nonlinear system and has fast response. The training algorithm is a modified backpropagation model, which can effectively reduce the training time. The target signal of the simulation is the steam generator water level in a nuclear power plant. The simulation result shows that the predicted value follows the real trend well