Analog-circuit fault diagnosis method based on DAGSVC

Abstract

The invention discloses an analog-circuit fault diagnosis method based on DAGSVC. The method comprises the following steps: applying certain test excitation to a analog circuit to be tested; acquiring circuit output response signals to be tested in a testable node of the circuit; performing multi-order fractional order wavelet packet (FRWPT) transformation on the acquired circuit output response signals and extracting fault characteristics so as to form a test sample; and performing calculation by utilizing fault-dictionary prestored information together with the DAGSVC to classify and position faults. The method has the advantages of global training optimality, fewer needed training samples, high fault-information resolution and the like, and can improve the precision and efficiency of diagnosing the faults of analog circuits.

Description

technical field [0001] The invention relates to a DAGSVC-based analog circuit fault diagnosis method, which belongs to the field of circuit network testing and signal processing Background technique [0002] Circuit network diagnosis technology has become another research focus after circuit network theory analysis and synthesis. For digital circuits, the diagnosis method is relatively mature, but there is no new breakthrough in the fault diagnosis method for analog circuits, and it is still hesitating. [0003] At present, the diagnostic methods for analog circuits mainly include: modeling and pattern recognition methods. Among them, the modeling method is mainly suitable for the fault diagnosis of linear or weak nonlinear circuits, and the analog fault diagnosis technology based on intelligent pattern recognition methods such as neural network and fuzzy mathematics can diagnose both linear circuits and nonlinear circuits, and is becoming the current research hotspots. A...

AI Technical Summary

Problems solved by technology

However, conventional binary support vector machine classifiers (mainly "1-v-r" and "1-v-1" two types of structures) have problems such as low diagnostic accuracy or slow diagnosis speed; and DDAG classification with a similar structure The structure of the device is not unique, and different structures may have inconsistent diagnostic accuracy

In addition, in the fault diagnosis process, the fault sample extraction technology is also very critical. The conventional fault feature extraction technology mainly uses signal analysis methods such as amplitude-frequency characteristic analysis, Fourier analysis, and wavelet packet transform, and these methods only obtain the signal after signal transformation. Feature information, but does not reflect the change of feature information in the process of signal transformation, resulting in the lack of fault information, the reduction of fault resolution and other problems

Images