蒸汽发生器故障诊断和容错控制中强跟踪滤波的应用研究

基于考虑随机噪声的蒸汽发生器非线性模型,利用强跟踪滤波器( STF)对蒸汽发生器水位控制系统的传感器故障偏差进行估计,在线诊断故障并重构传感器测量值,实现容错控制.仿真结果表明,该方法可以诊断随机时变的故障偏差,具有很好的容错控制能力.     

基于PCA的核电站传感器状态监测方法研究

针对目前核电站中以物理冗余为主的传感器状态监测方法所存在的不足之处,提出了基于主元分析(PCA)的传感器状态监测方法,这种基于解析冗余的方法是对物理冗余方法的验证,解决了物理冗余方法不能实现传感器小漂移的监测,改善了冗余传感器组中多数传感器出现共模故障时,物理冗余监测方法可能给出错误融合结果的问题。使用核电站的真实传感器数据建立PCA监测模型,人为引入故障到测试数据中进行分析,仿真结果验证了文中提出的传感器状态监测模型的有效性。     

基于PCA的主泵传感器状态监测模型

核电厂主泵的主、辅系统中布置了大量的传感器,随着主泵的运行,传感器会出现不同程度的老化或故障。为了改善现有核电厂传感器周期性测试和校准方案的不足,提高运行的安全性与经济性,采用主成分分析(PCA)技术对主泵的传感器进行状态监测。使用某核电厂主泵的运行数据建立PCA监测模型,并利用该模型对传感器的小漂移故障和共模故障进行识别,仿真结果表明该模型对主泵传感器具有很好的监测效果。     

核电厂安全级仪表在线监测系统技术研究

目前国内核电厂主要采取定期校准的方式对安全级仪表漂移进行管理,但该方法过于保守且经济性差。基于此,本文对安全级仪表在线监测系统技术进行了研究,首先对安全级仪表实际漂移数据进行了分析,明确了核电厂安全级仪表漂移的主要类型,证明了对安全级仪表开展在线监测的可行性。其次,通过对相关法规及标准的分析和研究,明确了核电厂安全级仪表在线监测技术的基本要求。最后,开展了在线监测系统技术的数据分析研究,对冗余仪表提出了等价平均算法,对非冗余仪表算法进行了分析并对多元状态估计模型(MSET)方法开展了基于电厂实际数据的建模验证,证明了该方法在核电厂应用的可行性。     

凸度仪电离室探测器时间响应特性研究

本文利用脉冲X射线机发射宽度很短的脉冲射线,通过分析数字示波器记录的电离室负载电阻上的电压信号,测量电离室内的离子漂移时间。利用该方法对凸度仪电离室探测器的时间响应特性进行了研究,给出了不同内部结构、不同电压及不同气压的电离室离子漂移时间的测量结果,并分析了误差产生的原因。实验结果表明,凸度仪电离室的时间响应满足在线测量热轧钢板的要求。     

基于轮转电极的多电极电容式棒位测量传感器静态特性研究

电容式棒位测量传感器是200 MW低温核供热堆控制棒水压驱动系统的关键测量部件,其精度和可靠性直接关系到反应堆的安全性。本文基于有限元方法建立了电容式棒位测量传感器的理论模型,对两电极电容式棒位测量传感器检测场的灵敏度分布特性进行了分析。结果表明,在检测场内存在一个特定区域,被测杆向该区域偏移时偏心误差较小。基于该特性提出了多电极电容式棒位测量传感器的设计方案和轮转电极的电容检测方法,针对该传感器,利用有限元模型进行了传感器结构优化和被测杆偏心误差分析,完成了传感器静态特性实验并验证了模型计算结果。分析结果表明,多电极电容式棒位测量传感器与轮转电极的电容检测方法能有效利用检测场的灵敏度分布特性,优化后传感器的偏心误差能达到棒位测量不失步的要求。研究成果为电容式棒位测量传感器的设计和优化提供了新的方向。     

核电站保护系统通道间通信故障对热备冗余控制站切换的影响分析

核电站保护系统通常采用四通道冗余设计,单通道采用热备冗余结构的控制站,主要是利用其主从切换功能提高单通道的可靠性。通道间采用单向点对点通信,如果该通信发生故障,相应的热备冗余控制站进行主从切换固然能提高可靠性,但是,由于四个冗余通道间均有通信,通道间通信故障可能带来多通道同时主从切换动作造成保护系统扰动,所以通道间通信故障是否需要进行主从切换是一个两难的选择。本文结合和睦系统(FirmSys)热备冗余主控制站的特点,引入故障模式影响分析(FMEA)技术对通道间通信及相关设备的故障模式及影响进行详细的分析,通过对切换与否的影响进行对比分析,给出最优设计方案。目前基于该方案的热备冗余主控制站的主从切换机制已应用于阳江56号机组、红沿河56号机组等的保护系统。本文为FMEA技术在系统设计方案中的应用提供了良好的实践。     

HT-7中长时间低漂移模拟积分器的研制

电磁测量是托卡马克装置中最基本的诊断方法,通常利用磁感应原理来测量HT-7装置内部磁场和磁通,因此需要用积分器对磁测量的输出信号进行积分。讨论了单运放模拟积分器的特点,提出了利用实时积分漂移补偿的方法设计新型积分器,解决了积分过程中积分器本身的等效输入偏移量、温度漂移及地噪声等带来的积分误差。给出了系统原理和设计框图,并在HT-7超导托卡马克的实验中,验证了该积分器能够满足等离子体长时间放电对积分器在积分精度和噪声精度上的要求。     

基于数据融合的航空γ能谱特征峰信息提取技术研究

针对航空γ能谱测量仪中多路NaI(Tl)探测器中的不同噪声水平,采用加权最小二融合估算法对能谱信息进行提取。该算法通过对NaI(Tl)探测器测量的数据进行在线方差学习,及时调整各探测器的权系数。将该方法用于航空能谱仪在标准模型上采集的数据来进行分析,可以看出,其单秒测量得到的误差范围有所改善,测量结果的准确度有所提高。     

PGNAA方法测量元素的非线性规律研究

瞬发伽玛中子活化分析(prompt gamma neutron activation analysis,PGNAA)在线分析仪广泛应用于水泥、煤炭等工业控制领域,由于工业现场产量的变化,导致皮带上物料的厚度不恒定,当物料厚度变化时,在线分析仪的测量结果会出现偏差。为消除在线分析仪测量不同厚度物料的测量偏差,通过理论分析和蒙特卡罗模拟(MCNP)研究中子自屏蔽和伽玛自衰减影响,并寻求修正模型。根据探测器接收特征伽玛射线的计算公式进行理论分析,利用蒙特卡罗软件建立PGNAA在线分析仪装置模型,分别模拟Al_2O_3、SiO_2、CaO、Fe和水泥生料等物料随厚度变化的特征射线计数,以水泥在线分析仪为例,寻求修正模型。结果表明,特征伽玛射线强度不随物料厚度增加而线性增长,按照线性标定方法的在线分析仪测量误差增大,各元素平均相对误差大于10%;根据水泥生料各元素的特征射线计数与物料厚度之间的关系,建立适用于测量水泥生料的多项式厚度修正模型。与线性标定相比,修正后分析仪测量误差降低30%以上,可为煤质、烧结料、铝土矿等块状物料的PGNAA元素检测提供修正方法。     

Bubbles in the sensing line of nuclear power plant pressure transmitters: the shift of spectrum resonances

Noise analysis is a valuable tool when applied to predictive maintenance problems in nuclear power plants. When focused to the capacitive pressure sensors it can detect the presence of undesirable bubbles in the sensing line. From the noise signal power spectrum density a resonant peak shifting towards the lower frequency region, due to bubbles appearance, is observed. In this work an explanation for the peak displacement is given; besides, a method to estimate the void fraction from the shift is developed. Both results benefit the on-line maintenance of pressure sensors without disturbing plant operation.     

Prediction of the performance of an ion chamber amplifier under γ radiation

The ion chamber amplifier (ICA) plays a major role in the proper functioning of a nuclear reactor as it monitors the radiations from the nuclear reactor by measuring the ionic activity inside the ion chamber. The signal conditioning circuitry of the ICA detects and conditions the weak ionic currents coming from the ion chamber dome. Degradation in the performance of the semiconductor devices used in this part of the ICA, can lead to inaccurate monitoring of the reactor operation, resulting in a possible catastrophe due to malfunction. Further, the response of the ICA under irradiation also depends upon the strength of the input signal (ionic) current it is required to handle. The active devices used in the ICA under study are operational amplifiers (Op-Amps) such as DN8500A and OPA111, instrumentation amplifier INA101, transistor 2N2920A and a voltage reference device, AD584. Since these devices may be sensitive to radiation, one must know their radiation behaviour so that the performance of the ICA can be predicted. This paper examines the performance of the ICA by characterising the radiation profiles of its vital components, viz. the Op-Amps, instrumentation amplifiers, transistors, etc. by monitoring their parametric changes on-line, i.e. when the source is on, and the devices are biased. The simulation runs involve the simulation of the entire ICA circuitry using the changed values of the vital parameters such as input bias current and input offset voltage. The main advantage of this method is that it obviates irradiating the whole ICA circuit to study its irradiation performance, and simulates an environment of radiation leakage around the ICA. Based on this study, results are presented to predict the performance of the ICA.     

New methods to remove baseline drift in trapezoidal pulse shaping

Trapezoidal pulse shaping algorithm is widely applied to improve signal-to-noise ratio(SNR), throughput and energy resolution with the properties of noise suppression, pile-up pulse separation and ballistic deficit correction. The algorithm can be acquired by z transform method which is easier for derivation. However, the baseline drift of trapezoidal pulse appears because the noise superimposes on the input signal. In this paper,two new methods based on convergence analysis and noise suppression are proposed to remove the baseline drift resulting from trapezoidal pulse shaping. Simulations and experimental tests are carried out to verify the methods. The results demonstrate that the proposed methods can remove baseline drift in trapezoidal pulse shaping.     

Neural networks applied to pressure sensor monitoring

A Regulatory Commission Requisition compels the NPPs to a periodic inspection in order to know the pressure sensors response time. During the recharge period, a pressure signal is applied to the sensor, and the response time to this signal is observed. This method is certainly accurate, of course, but it is also very cumbersome, little versatile and expensive. So alternatives have arised, in the recent years, based in the Noise Analysis Technique. These alternatives provide satisfactory results, not only from the safety point of view but also from the economics, with a really smaller effort.

In this work Neural Networks are going to be used to obtain the pressure sensors response time. The purpose is to connect the noise correlation function to this time via a Neural Network. It will be plausible that an easy architecture is able to “learn” this problem, offering satisfactory results.     

Analysis of a BWR Stability Behavior using Generalized Discrete Shannon Functions. Application to Ringhals and Forsmark Benchmarks Data

In this paper we have introduced a new methodology to on-line signal conditioning and monitoring to determine the stability parameters of the BWR NPP, that is, the determination of the effective Decay Ratio DR and the frequency of the main oscillation causing instability events. This method is based on the generalized discrete Shannon function convolution, which removes the noise and filters the signal in a specified frequency band. We have focused our attention in noise signals, first on analytic ones to check how the algorithm works, and then we have tested it with some real neutron signals. The algorithm works very well with dirty real signals providing good results, even in the case of short time series. Main attemption has been focused on decomposing signals to detect when a global and/or a regional oscillations are taking place in a BWR. This methodology can be implemented in on-line monitors to determine the stability parameters of the BWR reactors, that is, the determination of the effective decay ratio (DR) and the frequency of the main oscillation causing instability events.     

A method for weak impact signal discrimination based on para-approximate entropy

Impact signal of the loose part in a nuclear power plant provides vital information about its mass and location. Hence detection of the impact signal is of great significance. In a practical NPP working environment, some weak impact signals may submerge in background noises. A signal detection and impact signal discrimination method is proposed in this paper. Teager energy operator is used to process a whitened signal using auto-regression model. It increases the signal to noise ratio (SNR) and makes the impact signal detection easy. Para-approximate entropy method is proposed to process the energy envelop for discriminating the loose part impact signal from other interferences. Simulated experiment has been carried out to verify the proposed method. Data analysis proves that it is capable to detect and discriminate a weak impact signal with low SNR.     

Development of Reactor Noise Monitor

A reactor noise monitor has been developed to provide a tool that serves the real time routine function of detecting anomalous states of operation in power reactors. In this monitor, RMS (root mean square) values and average frequencies of noise signals are calculated on a real time basis. The average frequency is defined as the frequency averaged by weighting with PSD (power spectral density) of the noise signal and proved to be equivalent to the ratio of an RMS value of the differentiated noise signal to an RMS value of the noise signal itself. Thus, the average frequency can be obtained on a real time basis by using a differentiator, an RMS calculator and a divider. The frequency components of 12 kinds of signals ranging of 0.01–36 Hz are divided into two or three frequency regions by band-pass filters and the RMS value and the average frequency of each frequency region are obtained by the digital calculation method. The performance of this monitor was tested by using a simulated anomalous signal with a peak on its PSD. From the test, it is proved that this monitor can detect more than changes of the RMS value and average frequency. The functions of this monitor were also confirmed at the actual BWR power plant.

In addition, a method to detect peak location of PSD is developed and the accuracy of the peak frequency measurement of the proposed method is investigated to be satisfactory.     

Structural Material Anomaly Detection System Using Water Chemistry Data, (VII)

Abstract

A method to detect small changes in water quality and diagnose their causes by analyzing on-line conductivity and pH data was proposed. Laboratory tests showed that effective noise reduction of measured on-line data could be got by using a median filter to detect small changes of conductivity; a relative change of 0.001 μS/cm was distinguishable. By simulating the changes of pH and conductivity in the reactor water against a small concentration change of sodium ion or sulfate ion in the feedwater, it was found that an adequate elapsed time for the diagnosis was 4 h from the start of the concentration change. A conductivity difference of 0.001 μS/cm in the reactor water made it theoretically possible to distinguish between a sodium ion concentration change of 4.6 ppt and a sulfate ion concentration change of 9.6 ppt in the feedwater.     

Drift-free magnetic equilibrium reconstruction using the response to plasma position modulation

Inductive magnetic sensors are widely used for plasma equilibrium reconstruction and control. However, their measurements involve electronic integration and can therefore experience drift leading to inaccurate plasma positioning. For this reason, we have studied an original drift-free approach to estimate the plasma equilibrium. The principle of this correction is based on modulation of the plasma position and current at three independent frequencies and analysing the modulated magnetic signals to provide additional estimation on the equilibrium. Using a plasma model based on current wires, the accuracy of such a method is assessed for Tore Supra in terms of the signal to noise ratio. The plasma position is recovered within a precision of 5 mm for a signal to noise ratio better than 80 dB. Applying our approach to dedicated experiments performed on the Tore Supra tokamak, we confirm the quality of the result and find that we can estimate the radial and vertical positions of the plasma to 1 mm with a one standard deviation confidence interval.     

一种测量高压气体中离子漂移速度的方法

利用脉冲X射线机作为离子触发源,用高频数字示波器记录负载电阻上的电压信号,通过曲线拟合得出离子在高压气体中的漂移速度。用该方法实测氙离子在1.5 MPa氙气中的漂移速度,结果表明,该方法可简便有效地测量高压气体中的离子漂移速度。