基于改进HMM的模拟电路故障预测模型

摘要

针对现有模拟电路系统PHM技术故障预测部分准确度不高和时效性不强的问题,提出了基于MCGCPSO优化改进的HMM状态退化识别模型.将小波包故障特征提取法与LDA特征降维法结合,确保了对特征信息非线性部分的充分提取,同时又避免了维度过高的问题;利用MCGCPSO优化改进后的HMM模型,提升了状态退化识别模型的分类准确度.最后将MCGCPSO-HMM与改进的灰色模型组合为一个新的电路故障预测模型,克服了单个预测方法性能不稳定的缺陷.通过仿真实验验证了MCGCPSO-HMM与改进的灰色模型组合具有更高的预测准确度.%In view of the problem existing in analog circuit system PHM technology support part of prediction such as low accuracy and the immediate effect,this paper proposes a HMM degradation recognition model based on optimizing and improving MCGCPSO. Combined fault feature extraction method of wavelet packet with LDA feature dimension reduction method,this paper ensures the nonlinear part feature information is fully extracted and avoids the problem of high dimension;the optimized HMM model using MCGCPSO improves the classification accuracy degradation recognition model on the basis of the principle of classification interval maximum,and a simulation is carried out. Finally,as the new a circuit fault prediction model,the combination of MCGCPSO-HMM and the improved grey model can overcome the defect of single forecasting method performance is not stable. It is verified by simulation experiments that the combination of MCGCPSO-HMM and the improved grey model has higher predictive accuracy.