核电站混合式故障诊断系统的开发和评价

提出了一种基于规则和人工神经网络的混合式故障诊断方法,用于在线诊断核事故。基于规则的诊断方法用于事故类型的识别,人工神经网络用于准确判断事故的具体位置以及严重程度等基于规则难以诊断的信息。正常工况下的仿真试验验证了该方法的有效性。     

多重人工神经网络在10MW高温气冷堆故障诊断中的应用研究

对人工神经网络技术在10MW高温气冷堆故障诊断中的应用进行了可行性研究，并用事故工况下反应堆监控参数的实际值和趋势变化值分别对2个BP网络进行训练和检验，综合2个网络诊断结果得出最终诊断结果。经检验证明，采用神经网络具有较好的容错能力，对噪音信号有良好的鲁棒性。     

人工神经元网络在核电工程中的应用现状

本文综述了人工神经元网络（ＡＮＮ）在核电工程中的应用现状，包括，报警处理和故障诊断；热功率预测，传感信号的有效性确认；核电厂运行变量的动态辨识；堆芯管理等其它方面。并分析了ＡＮＮ应用于核电工程中的特点及前景，指出了今后有待解决的问题。     

A cascade intelligent fault diagnostic technique for nuclear power plants

Safe operation of nuclear power plant is one of the most important tasks in nuclear power development. This justifies the variety of methods that have been proposed to support the operators in the task of plant condition monitoring, fault detection, and diagnosis. A number of hybrid fault detection and diagnosis methods have also been proposed, with their attendant weaknesses. This work proposes the hybrid of principal component analysis (PCA), signed directed graph (SDG), and Elman Neural Network (ENN) for fault detection, fault isolation, and severity estimation, respectively. The proposed hybrid method is verified with the data derived from Personal Computer Transient Analyzer (PCTRAN) simulation. The verification result shows that the PCA-based fault detection methodology realized timely detection of anomaly in the simulated nuclear power plants system, the SDG-based fault recognition method was able to isolate the system abnormality and identify the root causes, and the ENN-based fault severity estimation method presents the failure fraction of fault, representing the severity. With this integrated hybrid method, more fault information is provided for the operators, which serves as a good foundation for further decision-making and interventions.     

Application of an Artificial Neural Network in Reactor Thermohydraulic Problem: Prediction of Critical Heat Flux

A new method for predicting Critical Heat Flux (CHF) with the Artificial Neural Network (ANN) method is presented in this paper. The ANNs were trained based on three conditions: type I (inlet or upstream conditions), II (local or CHF point conditions) and III (outlet or downstream conditions). The best condition for predicting CHF is type II, providing an accuracy of ±10%. The effects of main parameters such as pressure, mass flow rate, equilibrium quality and inlet subcooling on CHF were analyzed using the ANN. Critical heat flux under oscillation flow conditions was also predicted.     

用人工神经网络方法进行非能动安全壳冷却系统可靠性分析

主要介绍了利用人工神经网络方法对非能动安全壳冷却系统进行可靠性分析。采用人工神经网络方法从有限组数据样本中得到准确的关于安全壳内压力的极限表面(Limit Surface),然后用蒙特卡罗方法得到了系统在多种参数条件下的不可靠性。合理的结果证明了方法的有效性。     

基于BP神经网络的核探测器故障诊断方法研究

核探测器是核设施放射性监测的重要设备,为了保障该设备的持续稳定运行,本研究针对闪烁体探测器提出了一种基于BP神经网络的在线智能故障诊断方法。采用小波包变换将探测器输出信号从时域变换至频域后提取特征向量,将得到的特征向量作为BP神经网络故障诊断模型的输入,再通过误差梯度下降法对该模型的参数进行优化,最终利用最优的诊断模型完成故障类型的智能识别与分类,并将该方法与统计诊断方法和基于支持向量机的故障诊断方法进行横向的对比研究。研究结果表明,新方法的平均诊断准确率均优于上述两种方法。因此,该方法的应用能有效地提高核探测器的故障诊断准确率。     

Incipient fault detection of motor-operated valves using wavelet transform analysis

The necessity of improvements in monitoring and diagnosis methods started to be of extreme relevance in the predictive maintenance field, establishing the reliability and readiness of system components as an achievable goal. Taking into account these reasons, this paper presents an approach for incipient fault detection of motor-operated valves (MOVs) using wavelet transforms. The technique applied in this paper is the wavelet transform analysis using wavelet toolbox, where the main goal is to obtain more detailed information contained in the measured data, identifying and characterizing the transient phenomena in the time and frequency domains, correlating them to failure situations in the incipient stage. The wavelet analysis has provided good results establishing a new qualitative methodology for monitoring and diagnostics of motor-operated valves.     

流动不稳定性对CHF的影响

本文成功地训练了一个用于预测流动不稳定性条件下CHF的人工神经网络,利用所训练的人工神经网络,分析了流动不稳定性对无量纲因子F的影响规律.分析结果表明,系统压力对F的影响是非单值的;平均质量流速增大,总体上F也增大.还分析了流动不稳定性条件下系统主要参数对临界热流密度的影响.结果表明:当质量流速的振幅与平均质量流速的比增加时,CHF 减小;质量流速的振幅与平均质量流速的比不变时,随着周期增大,CHF减小.     

Iterative PET Image Reconstruction Using Translation Invariant Wavelet Transform

The present work describes a Bayesian maximum a posteriori (MAP) method using a statistical multiscale wavelet prior model. Rather than using the orthogonal discrete wavelet transform (DWT), this prior is built on the translation invariant wavelet transform (TIWT). The statistical modeling of wavelet coefficients relies on the generalized Gaussian distribution. Image reconstruction is performed in spatial domain with a fast block sequential iteration algorithm. We study theoretically the TIWT MAP method by analyzing the Hessian of the prior function to provide some insights on noise and resolution properties of image reconstruction. We adapt the key concept of local shift invariance and explore how the TIWT MAP algorithm behaves with different scales. It is also shown that larger support wavelet filters do not offer better performance in contrast recovery studies. These theoretical developments are confirmed through simulation studies. The results show that the proposed method is more attractive than other MAP methods using either the conventional Gibbs prior or the DWT-based wavelet prior.     

基于核磁共振的癌症非侵入式诊断理论与技术研究

分析了核磁共振技术在癌症临床诊断领域的应用现状、相关理论与技术的发展趋势,概述了对核磁共振设备产生的FID信号、MRI和MRS的数据分析技术,并以MRS为重点,研究了FID信号和波谱数据的特点,提出了利用小波变换和模式识别技术对FID信号和波谱数据进行分析的技术方案,并对信号的小波阈值去噪、小波基函数的选择、MRS特征识别与提取等问题进行了探索性研究,对核磁共振为基础的相关技术在癌症临床诊断领域内应用的进一步发展提示了方向。     

基于频域数据注意力机制的核电厂水泵故障模式识别模型研究

针对核电厂水泵共性的异常振动、转子部件摩擦与磨损等故障模式,利用水泵最容易获取的泵壳加速度信号的频域数据为输入,提出了一种结合卷积神经网络和注意力网络的频域数据注意力机制方法,并建立了核电厂水泵故障模式识别模型。研究结果表明:相对于传统方法,利用频域数据作为输入、基于频域数据注意力网络算法建立的水泵故障模式识别模型输入的数据长度更短,能够有效提升模型训练的效率,该故障模式识别模型在测试集上的故障模式识别准确率达到100%,优于其他基于深度学习算法建立的故障诊断模型,证明了本文提出方法的优势。      

Evaluation of applying axial variation of enrichment distribution method and radial variation of enrichment distribution in VVER/1000 reactor using a Hopfield neural network to optimize fuel management

In this present work the analysis technique was developed to find the optimum core configuration by applying neural network. This work investigates an appropriate way to solve the problem of optimizing fuel management in VVER/1000 reactor. To automate this procedure, a computer program has been developed.This program suggests an optimal core configuration which is determined to establish safety constraints. The suggested solution is based on the use of coupled programs, which one of them is the nuclear code, for making a database and modeling the core, and another one is Hopfield Neural Network Artificial (HNNA).The first stage of computational procedure consists of creating the cross section database and calculating neutronic parameters by using WIMSD4 and CITATION codes. The second one, consists of finding the optimum core loading pattern by applying the primary fuel assemblies of the VVER/1000 reactor core, using the HNNA method that based on minimizing power peaking factor (PPF) and maximizing the effective multiplication factor (keff). In the third second one, we apply a heuristic method to flat the flux core and decreasing the power peaking factor of the core. It consists of finding the best axial and radial variation of enrichment distribution to reach an optimum core loading pattern, by using HNNA and the cross section database.Finally, we compared obtained results of these methods to obtained results of the primary core, Suggested pattern of the Russian contractor.In total, the results show that applying the HNNA led us to the appropriate PPF and keff. Therefore, we achieved to a set of two basic parameters PPF and keff as effective factors on satisfying the safety constraints of VVER/1000 reactor core. It should be mentioned to say that the obtained results of HNNA suggested pattern is promising.Therefore, these methods ultimately eventuated to find the optimum configuration for VVER/1000 reactor core.     

基于模极大值法的γ能谱数据处理与分析

本文研究了Lipschitz指数与小波变换应用于γ能谱数据奇异点的使用方法,分析了模极大值去噪原理和Mallat提出的交替投影重构算法,建立了基于模极大值法γ能谱数据处理模型,基于MATLAB设计了小波变换模极大值去噪软件,通过实测γ能谱数据仿真分析基于模极大值法的γ能谱数据处理模型的效果。结果表明,利用小波变换模极大值法降噪,用交替投影法重构能谱数据,降噪效果较好。     

基于SWT和Hilbert的密封磁阻马达式CRDM振动信号去噪

针对控制棒驱动机构滚轮振动信号采集过程受到噪声干扰的问题,提出了小波半软阈值（SWT）和Hilbert变换法相结合的去噪方法。该方法首先利用小波半软阈值算法在时频域对滚轮振动信号进行降噪处理,然后进行Hilbert变换求出其包络谱,分析寿命试验与缺陷验证试验中不同的滚轮振动信号。试验结果表明,该方法可有效消除噪声对振动信号的干扰,证明了小波半软阈值与Hilbert变换相结合的方法在驱动机构滚轮状态识别和故障诊断中的有效性,为驱动机构的状态判别提供了理论支持。     

基于PSO优化神经网络响应面技术的非能动系统可靠性分析

在非能动可靠性分析数学模型的基础上，结合某型核动力装置非能动余热排出系统原理性试验装置和改进的热工水力程序的运行数据，识别了输入参数的不确定性，比较了不同神经网络响应面技术替代热工水力程序的精度和优度，分析了粒子群优化算法（PSO）优化神经网络响应面分类准确率。数值结果表明，该响应面具有较高的拟合优度，且能够较为准确的对非能动系统系统可靠性进行判定。      

Design and optimization of a multistage turbine for helium cooled reactor

This paper describes the aerodynamic design and explores the performance limits of a 600 MWt multistage helium turbine for a high temperature nuclear reactor closed cycle gas turbine. The design aims for maximum performance while limiting the number of stages for reasons of rotor dynamics and weight.A first part discusses the arguments that allow a preliminary selection of the overall dimensions by means of performance prediction correlations and simplified stress considerations. The rotational speed being fixed at 3000 rpm, the only degrees of freedom for the design are: the impeller diameter, number of stages and stage loading.The optimum load distribution of the different stages, the main flow parameters and the blade overall dimensions are defined by means of a 2D through-flow analysis method. The resulting absolute and relative flow angles and span-wise velocity variation are the input for a first detailed design by an inverse method. The latter defines the different 2D blade sections corresponding to prescribed optimum velocity distributions.The final 3D blade definition is made by means of a computer based 3D-DESIGN system developed at the von Karman Institute. This method combines a 3D Navier-Stokes (NS) solver, Database, Artificial Neural Network and Genetic Algorithm into a two level optimization technique for compressor and turbine stages. The use of 3D Navier-Stokes solvers allows full accounting of the secondary flow losses and optimization of the compound leaning of the stator vanes.The performance of the individual stages is used to define the multistage operating curves. The last part of the paper describes an evaluation of the cooling requirements of the first turbine rotor.     

基于SVM和DWT的辐射图像局部特征识别方法

针对辐射图像的特点,设计并开发了一种支持向量机（SVM）和基于离散小波变换（DWT）的局部特征识别方法。使用支持向量机解决了辐射图像局部特征提取的困难和分类器对样本数目的要求。而小波的应用使得该算法能够支持多分辨率的特征提取,并提高了总体识别效率。还对比了两种常见的核函数,实验结果表明高斯径向基函数能够取得比较好的分类效果。     

Applications of ANNs in flow and heat transfer problems in nuclear engineering: A review work

Artificial Neural Networks (ANNs) have been applied to deal with flow and heat transfer problems over the past two decades. In the present paper, recent work on the applications of ANNs for predicting the flow regime, pressure drop, void fraction, critical heat flux, onset of nucleate boiling, heat transfer coefficient and boiling curve has been reviewed, respectively. As can be noted in this review work, various types of ANNs can be employed as predictors with acceptable precisions. At the end of this review, methods to improve performance of ANNs and further applications of ANNs in flow and heat transfer problems were introduced.     

Water level in boiling water reactors — Measurement, modelling, diagnostic

The paper deals with the hydrostatic water level measurement in connection with the application of knowledge-based and model-based methods of signal processing using Fuzzy Set Theory, and under utilisation of internal gamma radiation as well as application of Artificial Neural Networks (ANN). The utilisation of Fuzzy-Set Theory and ANN's is explained. The principle of different measuring methods are described and placed underneath with applications, like Hybrid Observers and diverse measuring system for boiling water reactors.