摘要
针对汽油机轻微漏气故障会被闭环反馈控制所掩盖,提出了一种基于在线希尔伯特变换(Hilbert-Huang Transform,HHT)和支持向量机(Support vector machine,SVM)的两阶段微小故障识别方法(Online HHT-SVM,OHS).第一阶段在嵌入式两滑动时间窗内通过HHT对发动机空燃比数据流进行在线时频分析,以实时获取空燃比发生异常的时刻;第二阶段通过SVM对异常时刻的数据流故障模式进行离线识别.根据氧传感器信号特征,对经验模态分解(Empirical mode decomposition,EMD)算法进行了改进,并从理论上进行了证明.基于两款发动机的实际运行数据验证了该方法的有效性.
The incipient intake leak fault will be covered up by closed loop feedback control system. Aiming at the problem, a new fault recognition method with two stage based on online HHT and SVM is proposed. In the first stage, the time-frequency distribution of engine air-fuel ratio stream will be analyzed in real time via HHT in embedded two sliding windows, which can get the abnormal time of air-fuel ratio in real time; In the second stage, do fault pattern recognition in off-line way by SVM at the fault time. According to the characteristics of oxygen sensor signal, the paper improve the process of Emd, and the proof is given in theory. The experiment results based on operating data from two varied engines validate the proposed approach.
出处
《数学的实践与认识》
北大核心
2017年第9期102-114,共13页
Mathematics in Practice and Theory
基金
国家自然科学基金项目(61305134)
博士点基金(20133219120035)