Font Size: a A A

Study On The Intelligence Technology Of Nuclear Power Plant Fault Diagnosis

Posted on:2007-06-14Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y K LiuFull Text:PDF
GTID:1102360215959698Subject:Nuclear science and engineering
Abstract/Summary:PDF Full Text Request
Condition monitoring and fault diagnosis of nuclear power plant (NPP) plays a vital role in guaranteeing the safety and the reliability of NPP. At present threshold monitoring method is the main method employed in our country. This method is adequate to indicate the deviation of the important parameters of NPP, however it can not find the root cause or the deterioration trend. Applying intelligent technologies such as artificial neural networks, the expert system (rule inference), fuzzy logic and so on, to the early fault detection of NPP will improve the safety reliability and the overall efficiency of NPP.This thesis is sponsored by the×××advanced research project. The research is carried out on a NPP simulator with an emphasis on the basic structure, the technology and the methodology of designing the NPP fault diagnosis systemThe research work and the major achievements can be summarized as follows:1. An advanced diagnostic algorithm is developed. The advanced diagnosis algorithm is the key to the real-time fault diagnosis. A hybrid diagnosis method was put forward in the dissertation, which combines the diagnosis subsystems based on the neural network (in the subsystem including FNN network diagnosis subsystem, BP network data fusion diagnosis subsystem and RBF network diagnosis subsystem) and the inference-based rule subsystem. Since neural network has a fast diagnosis speed and a low diagnosis missing rate, it is used to carry out the diagnosis in the first stage. If the diagnosis reliability is not enough, then the inference-based rule diagnosis subsystem is used to carry on the confirmation and the explanation;2. A diagnostic system for NPP has been successfully developed using VB 6.0, where the basic structure of NPP real-time fault diagnosis system is based on the intelligent technology;3. A variety of auxiliary function systems and the friendly human-machine interface (HMI) have been designed, which improve the expandability and operation of the system. The knowledge base (rule base) of the inference-based rule diagnosis subsystem and the neural networks subsystem can be modified or expanded easily by users through the human-machine interaction. It also improves the system's ability on the deeper understanding of the NPP safety characteristics and the collection of the accident data;4. The diagnostic system successfully realized the automatic diagnosis and alarming function. With the NPP running, the real-time fault diagnosis system is operating in the background, which carries on the real-time monitoring on various parameters. If the fault occurred, the system will automatically diagnose to the faults and inform the operator to take the corresponding actions;5. Many experimental study were carried out on the simulator of NPP, the test result confirmed that the diagnosis system is suitable for the NPP real-time fault diagnosis. It also confirmed the feasibility and the validity of fault diagnosis according to the fault premonition. The accuracy of the designed condition monitor and fault diagnosis system has also been verified.
Keywords/Search Tags:Nuclear power plant, Fault diagnosis, Rough set, Artificial neural networks, Inference-based rule
PDF Full Text Request
Related items