Dissertation > Excellent graduate degree dissertation topics show

Research on Analog Circuit Fault Diagnosis Methods Based on SVDD and Parameter Identification

Author: LiChuanLiang
Tutor: WangYouRen
School: Nanjing University of Aeronautics and Astronautics
Course: Detection Technology and Automation
Keywords: Analog circuit Troubleshooting Fault classification Support Vector Machine Support vector domain description Genetic Algorithms Parameter Identification
CLC: TN710
Type: Master's thesis
Year: 2011
Downloads: 101
Quote: 0
Read: Download Dissertation

Abstract

Analog circuit fault mode classification is the key to intelligent fault diagnosis method, its research to improve the accuracy of diagnosis to protect the validity of the diagnosis has important significance. SVDD (support vector domain description, Support Vector Domain Description) not only has a few parameters, high precision, global optimization, etc., and high efficiency, scalability, there is the potential to solve problems online diagnosis in analog circuit fault diagnosis has broad application prospects. Traditional analog circuit fault diagnosis method of parameter identification and fault diagnosis method validation, although the performance is not very satisfactory, limited in scope, but there are still advantages of intelligent diagnosis method instead, they must be able to conduct in-depth research to promote analog circuit fault diagnosis rapid development. In this thesis, SVDD theory and genetic algorithm parameter identification based study of three analog circuit fault based SVDD classification method based on parameter identification, and two analog circuit fault diagnosis method: (1) based on the discrete degrees of support vector preselection SVDD Analog circuit fault classification method. SVDD-based analog circuit fault diagnosis method, the number of training samples increases with the failure mode increased significantly, while the large-scale data gathering SVDD applied to computing time and storage space encounter bottlenecks. To address this issue, this paper presents the concept of sample dispersion, dispersion theory explains the relationship and support vector and using Dispersion SVDD support vector constituted pre-selection training sample set reduction to reduce SVDD failure classifier training samples. Experimental results show that the dispersion of the sample accurately reflects the relationship between the vector and the support, with higher accuracy than the other metrics. Dispersion-based pre-selection of SVDD support vector classification method can guarantee fault classification accuracy of the premise, a substantial increase fault classifier training efficiency and reduce storage space requirements. (2) based on quadratic mapping SVDD analog circuit fault classification method. The second mapping method of mapping the use of secondary SVDD can be obtained a compact, better suitability of the sample distribution boundary line description. SVDD constructed using a quadratic mapping fault classifier can effectively reduce the failure modes of the overlap and improve fault classification accuracy. Troubleshooting experiments show that based on quadratic mapping SVDD analog circuit fault classification method is effective for improving the performance of SVDD fault classification, improved SVDD in analog circuit fault diagnosis applicability. (3) Based on the full sample SVDD analog circuit fault classification method. SVDD classification as a single value, ignoring the non-support vector sample information contained in the performance of its multi-class classification have a greater impact. In this paper, kernel density estimation and classification fuzzy combination of ideas, in the traditional SVDD fault classification decision rules in the integration of non-support vector contains sample information is proposed based on the full sample SVDD analog circuit fault classification method. Troubleshooting experiments show that with the traditional SVDD-based analog circuit fault diagnosis method, based on the full sample SVDD analog circuit fault classification method accuracy is improved greatly, and robustness, parameter selection convenient, reliable diagnostic results . (4) Based on Parameter Identification of analog circuit fault diagnosis methods, including those based on system parameter identification module level analog circuit fault diagnosis method based on parameter identification verification analog circuit fault diagnosis and French. Based on system parameter identification analog circuit module level fault diagnosis method based on system parameters and the correspondence between circuit modules using genetic algorithm for identification of system parameters on the resulting fault circuit module to locate and determine the degree of fault. Based on parameter identification verification analog circuit fault diagnosis method combined with a failure validation and parameter identification method has the advantage, the use of genetic algorithm parameter set identification verification failed component while faulty components obtained parameter values. The experimental results show that both diagnostic method is effective for analog circuit parameters fault diagnosis. This thesis work was supported by the National Natural Science Foundation of China (60871009), Aeronautical Science Foundation (2009ZD52045) and Jiangsu Province Graduate Research and Innovation Program (CX10B ₀ 98Z) funding.

Related Dissertations

CLC: > Industrial Technology > Radio electronics, telecommunications technology > Basic electronic circuits > Electronic circuit
© 2012 www.DissertationTopic.Net Mobile