The Application Of Neural Network Information Fusion Fault Diagnosis Methods In Oil Pipeline Leakage

The information fusion technology is a popular subject in recent years, it can comprehensive handle the information from multi-sensor, and make a correct judgment and estimate. So it has been widely applied in many fields by people. In the fault diagnosis of pipeline leakage, there is so much information to reflect system running status. Making full use of the useful information can improve the precision and accuracy of the fault diagnosis. On this basis, the paper presents a multiple sensor fault diagnosis integration system model, and it is applied to the pipeline leakage.This paper analyzes and researches the mechanism of the fault diagnosis of pipeline leakage, after that the experimental design is brief introduced. The experimental system can realize the acquisition of pipeline pressure data. On this basis, the wavelet packet transform is used to remove the noise of pressure signal and extract the fault feature vector. In addition, it builds the feature vector of pressure signal from energy angle. That all laying a foundation for further studying the method of neural network information fusion fault diagnosis.This paper introduces the structure of neural network and learning algorithm, and it is applied to the fault diagnosis. Through analyzing its advantages and disadvantages in the process of fault diagnosis, the paper puts forward and sets up a new fusion fault diagnosis model. It is the feature-level fusion diagnosis. On the basis of fuzzy technology combined with neural network and in order to improve the performance and speed of network, it puts forward an improved structure of fuzzy neural network and builds the corresponding model framework of fault diagnosis fusion. When it is applied to the complex systems fault diagnosis, the"dimension explosion"problem can be solved to final achieve the purpose of improve the diagnostic accuracy. Taking the method in the fault diagnosis of oil pipeline leakage, it respectively identifies three diagnosis modes about normal, leakage and regulating pump. The MATLAB simulation experiment is taken, using the pressure signals with fault information collecting by experiment devices. The experimental results show that it can greatly improve the diagnostic accuracy compared to the general neural network fault diagnosis methods and verify the feasibility and effectiveness this method.This paper is some of theoretical basic research work and algorithm analysis. If it is applied to the system development of actual fault diagnosis, still need to conduct further tests and perfected.