Title :
Method of analog circuits fusion diagnosis based on BP network and DS theory
Author :
Lin Zhigui ; Feng Zhihong ; Xiao Zhitao ; Zhong Qingqing
Author_Institution :
Sch. of Inf. & Commun. Eng., Tianjin Polytech. Univ., Tianjin, China
Abstract :
To overcome disadvantages of analog circuits fault diagnosis based on single information, we introduce multi-information fusion technology to fuse multiform circuit responses information in order to diagnose analog circuits fault. By three kinds of circuit responses information, that is accessible node voltages under determinate test frequency, components temperature and circuit output voltage under variable test frequencies, a method of analog circuits fault diagnosis combining Back-Propagation(BP) network and improved Dempster-Shafer(DS) theory is proposed. Firstly, Based on BP network, each kind of information is preprocessed respectively. Secondly, the preprocessing results are fused by improved DS theory. Lastly based on the fusion result, we diagnose analog circuits fault. Experimental results show that the method can overcome disadvantages of those methods based on single information and effectively deal with conflict evidences, and that distinguish ability of fusion results on the method is improved.
Keywords :
analogue circuits; backpropagation; electronic engineering computing; fault diagnosis; inference mechanisms; neural nets; BP network; DS theory; Dempster-Shafer theory; analog circuit fusion diagnosis; multiform circuit; multiinformation fusion technology; node voltage; Analog circuits; Circuit faults; Fault diagnosis; Frequency measurement; Fuses; Temperature measurement; Voltage measurement; BP network; DS evidence theory; analog circuit fault diagnosis; information fusion;
Conference_Titel :
Natural Computation (ICNC), 2010 Sixth International Conference on
Conference_Location :
Yantai, Shandong
Print_ISBN :
978-1-4244-5958-2
DOI :
10.1109/ICNC.2010.5582904