摘要
本文根据模糊识别原理，引入集合分类概念，提出了一种用于在线监测发动机缸内部件故障的无监督竞争学习模糊神经网络。该网络仅需正常状态下的振动信号样本集及少量的故障状态样本进行学习，并且它可根据噪声及运行工况的变化，自适应地调整代表正常状态下的网络权值；采用对比增强及加权方法，抑制各样本中的噪声影响。用本文方法对ＥＱ６１００汽油机和１９０Ａ柴油机人工设置的缸套活塞磨损故障进行诊断，取得了理想的效果。本文为以理论指导为主、少量实验为辅的在线诊断内燃机缸内部件故障，探索了一条有效、便捷的途径。
According to the principle of pattern fuzzy recognition,the unsupervised competing ANN for on line monitoring the inner parts of I.C.engine is presented by means of the concept of sets classification.The ANN only needs the samples of vibration signals under the normal condition and a few abnormal samples to learn,and the weight values stands for normal condition can be adjusted by resort to operating condition of engine to inhibit the effects of noise in the samples.The faults previously designed about piston liner worn out of the EQ6100 model gasoline engine and the 190A diesel engine are diagnosed by using the method of this paper,and the satisfactory results are derived.The purpose of the paper tries to find out an efficient way to on line diagnose faults of the inner parts of cylinder of I.C.engine under relying on more direction of theory as well as less experiments.
出处
《内燃机学报》
EI
CAS
CSCD
北大核心
1998年第1期88-93,共6页
Transactions of Csice
关键词
模糊分类
无监督
神经网络
内燃机
故障监测
Fuzzy classification, Sets classification, Unsupervised learning ANN, On line monitoring
