中文摘要：

针对模拟电路的健康性能退化状况，提出一种特征选择与降维提取法（B＆B＋LDA）和隐马尔可夫模型（HMM）相结合、以KL距离为衡量标准的状态监测和健康评估方法．首先设置元件的参数提取幅频特征；其次针对特征存在的冗余性及高维性，采用B8LB＋LDA对原始特征进行提取，从而获得有效的特征集；再根据获得的特征监测出早期故障类型；最后利用正常态下的特征来训练HMM，并用其计算各状态对应的KL距离，得出故障程度，即实现电路健康退化的评估．将该方法应用于某模拟电路中，通过实验验证了其具有良好的模拟电路早期故障监测性能，与B＆B，LDA，PCA及原始特征相比具有最好的状态监测与健康评估能力．

英文摘要：

Aiming at monitoring analog circuit with health performance degradation, a new method on state monitoring and health evaluation is proposed. By combining B＆B algorithm to realize feature selection and LDA to reduce feature dimension, then HMM is used to identify fault category and calculate KL distance to evaluate its health performance. Firstly, the corresponding frequency features are extracted from analog circuit with its components change gradually. Secondly, due to redundancy and high domain of original features, the superior features are obtained by using B＆B and LDA. Then, HMM initialized by K-means clustering is trained by the superior features to identify incipient fault. Finally, the corresponding KL distance under different state is calculated by the trained HMM to evaluate the health degradation. Applying this new method to an analog circuit, the experimental results show that the new method has excellent capability of monitoring incipient fault. In comparison with B＆B, LDA, PCA and original features, our method owns the best ability to monitor analog circuit state and evaluate its health.