摘要
针对信息不足、噪声会导致模拟电路故障诊断效率降低问题,提出基于小波分解、主成分分析和神经网络的信息融合故障诊断方法。为了减少噪声影响和减低故障特征维数,采用该方法对电路测试信号进行小波多尺度分解、主成分分析和归一化预处理。根据不同测试激励源,分别构造独立神经网络完成故障初级定位,进而运用D-S证据融合初级诊断结果实现故障最后定位。研究结果表明:所提方法能充分利用不同信息源对容差下模拟电路故障进行诊断,且定位准确率高。
Considering the problem of the low location ratio of analog circuit faults due to the lack of diagnosis information and the noisy interference, a novel fusion method was proposed to diagnose the analog circuit, and analyze the principal component and evident fusion theory based on wavelet decomposition. To reduce the noisy interferences and decrease the dimension of faulty features, the signals of circuit under test were processed with the wavelet decomposition, principal component analysis and normalization. The faults were primarily located by the independent neural networks reconstructed by different testing sources. Then, the final location of faults was implemented by the D-S fusion method with the preliminary results. The results show that the proposed method has the capability to use different information for fault diagnosis of analog circuits under the tolerance and higher accuracy can be obtained.
出处
《中南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2015年第1期127-134,共8页
Journal of Central South University:Science and Technology
基金
国家自然科学基金资助项目(51477040
51377044)
教育部博士点基金资助项目(20121317110008)~~
关键词
故障诊断
神经网络
小波分解
D-S证据推理
fault diagnosis
neural network
wavelet decomposition
D-S evident inference
