【作者】 冯俊婷；

【导师】 徐銤； 王桂增； 刘国发；

【作者基本信息】 中国原子能科学研究院 ， 核能科学与工程， 2003， 博士

【摘要】 由于核电厂设备故障可能对核电厂安全运行带来严重的影响，因此，故障诊断技术在核电领域受到高度重视。随着故障诊断技术的发展，以检测、识别、预测和干预为核心的先进的故障诊断系统在核工业中得到了广泛的应用，系统能够及时、准确地诊断出故障，并采取相应的对策，保证和提高核反应堆运行的安全性和可靠性。 本课题对中国实验快堆钠泵的故障进行了研究，把先进的诊断方法应用到正在制造中的钠泵上，设计和开发了不同方法的钠泵的故障诊断系统；同时也应用主元分析方法，建立了某核电厂的主冷却剂泵的故障诊断系统，仿真实验证明了以上三个系统的可行性。 论文主要内容包括： 1．对小波包分析应用于故障诊断进行了研究，利用仿真出来的故障数据，将快堆钠泵非平稳时变振动信号分解到不同层次和不同频带上，有效地提取出反映钠泵不同故障(状态)的特征向量，在识别中得到了满意的结果。 2．开发了小波包与专家系统结合的故障分析软件，利用小波分析方法进行快堆钠泵的状态监测和分析，通过故障诊断专家系统进行反向推理来进一步验证并给出解释。这样就把小波分析的快速诊断能力和故障诊断专家系统的解释能力有机地结合起来，充分发挥了两者的长处，提高了诊断的快速性和准确性。 3．研究了基于自适应径向基函数(RBF)网络的故障诊断方法。在RBF网络的基础上，对自适应RBF网络做了一定的改进，使故障诊断的准确率大大提高。经过快堆钠泵的故障仿真，证明该方法能够较迅速地、自适应地检测出系统的故障。 4．应用主元分析方法将高维数据转换到低维数据空间，这使得过程监测可以在低维的空间内进行。不断观察模型的平方预测误差(SPE)，一旦发现异常，利用主元分析方法，求出当前数据的特征方向，并与故障特征方向库进行比较，从而诊断出故障来。本文将此方法应用于某核电厂主冷却剂泵，诊断结果比较理想，证明此方法在核电厂的故障诊断中是可行的，对快堆钠泵的诊断也是可取的。更多还原

【Abstract】 The serious influences may be caused by equipment fault in a nuclear power station to operation safety. Hence, fault diagnosis technologies are met with altitudinal recognition in nuclear field. Along with fault diagnosis technologies used and developed, advanced fault diagnosis systems based on checking, identification, pre-measuring and intervention are applied widely and deeply in nuclear industry, which is capable of diagnosis to fault in time and truly. At the same time corresponding countermeasures adopted ensure and enhance operation safety and reliability in a nuclear power station.Based on adequate survey to faults of sodium pump for The China Experiment Fast Reactor, in order that the advanced diagnosis methods are used in the sodium pump which is under fabrication, different methods of fault diagnosis system of sodium pump are designed and programmed; and the principal component analysis (PCA)is used to fault diagnosis system of main coolant pump in a nuclear power station. The feasibility of fault diagnosis systems developed have been proved by simulation results.The main works of this thesis include:1. The wavelet packet analysis applied in fault diagnosis system is studied in this thesis. Using simulation fault data, uncalm and time difference vibration signal of sodium pump are decomposed on different levels and frequencies. To picked-up the eigenvectors reflecting different fault of sodium pump with availability, simulation results have shown its satisfaction in the identification.2, Fault diagnosis software which combines wavelet packet and expert system is developed, based on wavelet packet analysis and reverse reasoning of fault diagnosis expert system, it has diagnosis and exploration. In this way it. combines fast diagnosis ability of wavelet packet analysis with exploration ability of expert system with a sufficient, exertion of both advantages,increasing rapidity and veracity of fault diagnosis.3. A self-adapting fault diagnosis method based on radial basis function (RBF) networks is studied in the thesis. Based on the basic RBF networks, adaptive RBF networks is improved, it brings to increase precision greatly. The method is applied to sodium pump, Simulation results have shown’the better capability for detecting faults.4. High dimension was changed into low dimension by using principal component analysis method, process detecting could be carried out in the low dimension space. Squared Prediction Error(SPE) of model is detected continuously, in case abnormity is found, currently data direction is calculated by PCA method, and compared with fault character direction storeroom, thereby fault is diagnosed. This method is used in a nuclear power station, diagnosis results is better ideality, and it has been shown that it is feasible for fault diagnosis of nuclear power station and Fast Reactor sodium pump.更多还原