基于遗传编程的发动机滑油系统故障诊断  

Fault Diagnosis of Aeroengine Lubricating System Based on Genetic Programming

作　　者：侯胜利[1] 王威[1] 胡金海[2] 周根娜[1] 

机构地区：[1]徐州空军学院航材管理系,徐州221002 [2]空军工程大学工程学院,西安710038

出　　处：《振动．测试与诊断》2008年第4期400-403,共4页Journal of Vibration,Measurement & Diagnosis

摘　　要：为了提高航空发动机滑油系统故障诊断的有效性,提出了一种基于遗传编程的故障特征提取模型。该模型首先利用遗传编程从原始特征集中提取更能反映故障本质的复合特征,然后通过Fisher判别分析进行二次特征提取,得到对分类识别最有效、数目最少的特征。在神经网络的分类试验中,经过遗传编程和Fisher判别分析提取的特征使样本集的可分性增大,分类正确率从80%左右提高到了97%以上,并且对分类器具有较强的鲁棒性,表明该模型提取的特征对滑油系统的几种典型故障具有更好的识别能力。In order to improve the validity of fault diagnosis of the lubricating oil system of an aeroengine, a new feature extraction model based on genetic programming in fault diagnosis is proposed. In this model, genetic programming constructs compound features from original feature set. And Fisher discriminant analysis is followed and employed second feature extraction in the transformed space, which can get rid of the correlation among features and reduce their dimensions. Thus a more effective and smaller subset of features for classification can be gained. In experiments of neural network classification, the rate of correctness increases from 80% around up to 97% above, while the extracted features are robust for classifiers. Practical results show that the model shows better ability in fault recognition, and opens a new way for engine condition detection and fault diagnosis.

关 键 词：滑油系统 故障诊断 特征提取 遗传编程 FISHER判别分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]