The Research On Soft-fault Diagnosis Of Large-scale Analoge Circuits

Fault diagnosis in analog circuit have been a major field since1960s.For severaldecades,a series of diagnosis theory and methods have been raised by the researchers.However，because of the widely used of large-scale analog circuit and the existence ofcomponents tolerance, non-linear and the diversity of faults, traditional diagnosistheory and methods have been difficult to meet the needs of practical engineering.Th-erefor,some others diagnosis theory and methods shoud be put forward.In recentyears,the artificial neural network theory have developed rapidly.It is a new andeffective approach for fault diagonsis of analog circuits,and it also caused theextensive interest of the scholars.This paper researches the fault diagnosis approach in large-scale analog circuitson the aspects as follows:Firstly, the application of artificial neural network in fault diagonsis of analogcircuits have been intensive studied,and the general steps of this method have beenanalyzed and stated. Although artificial neural network approach could find out thefault components accurate and efficient,its efficiency still have been difficult to adoptto the needs of practical engineering in fault diagnosis of large-scale analogcircuits.This paper presented a approach for fault diagnosis of large-scale analogcircuits to combine the hierachical search approach with interval diagnosis approach.Hierachical searching the fault subnetwork by interval diagnosis approach,and findingout the fault components from fault subnetwork by artificial neural network approach.Secondly, fault dictionary approach is the most practical in fault diagnosis,andmany methods are evolved from it.It is the most important to establish the faultdictionary. Hard fault dictionary is relatively simple to set up,but soft fault number isuncountable,so the soft fault dictionary can not establish completely.This paperpresented tearing node interval voltage which get from parameter sweep of PSPICE.Soft fault state can be described completely by those interval voltage,and soft faultdictionary can be also established.Thirdly,it is difficult to process interval data by artificial neural network,butcombine fuzzy logic with neural network is very useful.So,this paper use fuzzy neuralnetwork to locate fault components in analog circuits.