INTRODUCTION OF IMPROVED XCSR ALGORITHM USING LIMITED TRAINING DATA: A CASE STUDY FOR FAULT DIAGNOSIS IN ANALOG CIRCUITS

Daily advancement of electronic science and analog/digital circuits has resulted in circuits with complicated tasks. Increased reliability of these systems, along with correct test, fault diagnosis andtroubleshoot of these circuits have become very important and critical issue. Step-by-step examination ofthe circuits during manufacturing and before its delivery to user is mentioned as a technique to enhance reliability of the circuits. Therefore, the best approach would be generation of a template of faults. Since extended classification system (XCS) is known as one of the most successful learning agents for problemsolving, XCS and other sample-based learning algorithms are utilized in this paper to diagnose the fault inanalog circuits. For example, an analog to digital converter (ADC) is used in this regard. Efficiency of these methods is also evaluated through comparison of the results (about the sample problem).

Keywords:

analog to digital converter , extended classification system (XCS) , neural network , reliability