基于ART的智能BIT测试技术研究

Intelligent BIT measurement technique based on ART

根据电子设备测试及维修的需要,同时为解决传统BIT虚警率高等缺陷,提出智能BIT测试方法。在介绍智能BIT的基本实现手段的基础上,重点分析ART网络的网络模型及工作原理,ART网络根据数据集自适应聚类实现模式识别,相对于传统人工神经网络而言,其收敛速度快,精度较高,自适应诊断能力强,解决了采用传统神经网络测试分类误差大等问题,最后提出了一种基于ART的BIT系统无监督故障诊断方法,使被测系统具有更高的故障诊断能力。

In order to meet the test and maintenance requirement for electronic devices, and to reduce the false alarm rate of the traditional BIT, an intelligent BIT test approach is presented. Based on introduction to basic implementation means of the intelligent BIT, the ART network model and its working principle are analyzed in detail. The ART net realizes the pattern recognition according to the data cluster self-adaptive cluster- ing. Compared with traditional artificial Neural Network （NN）, it has the advantages of faster convergence speed,higher precision and better adaptive diagnosis capability ,and the problem of high categorizing error in the traditional NN is solved. Finally, a free-of-surveillance fault diagnosis approach is presented based on ART, thus the tested system may have higher fault diagnosis capability.