摘要
研究复杂发动机的故障检测,提高检测的快速性和准确性。传统故障检测方式必须通过计算发动机不同部位实时的故障系数,并且与正常状态下的发动机系数标准逐个进行对比,从而判断发动机部件是否存在故障,当发动机结构复杂,部件较多的情况下,会造成检测方法计算量加大,检测耗时,结果滞后。为了克服上述问题,提出一种粒子群神经网络的复杂发动机故障检测技术,通过自适应粒子神经网络进行迭代计算,对发动机可能出现故障的部位进行相关参数统计,从而提前进行预判,减少大量由计算带来的滞后性影响。实验证明,改进方法能够大幅减低检测耗时,弥补滞后误差,取得了令人满意的效果。
Traditional way fault detection must first calculate engine fault coefficient,and then put it in the database and compared them with normal engine coefficients one by one to judge whether there are engine components faults,which means certain hysteresis.In order to avoid the above problems,the paper proposed a prediction engine technology based on the complex fault detection.This fault detection method,does not overly depend on fault parameters.Through the stereo space function neural network algorithm,the motivation of the malfunction sites can be predicted,thereby reduce engine temporary running stop.Experiments show that the prediction based on the complex engine fault detection technology can predict fualts starting point,and obtain satisfactory results.
出处
《计算机仿真》
CSCD
北大核心
2012年第4期272-275,共4页
Computer Simulation
基金
黑龙江省教育厅科研项目"基于NiO/Si纳米线微纳甲醛气体传感器的研制"(11541398)
关键词
故障检测
立体函数
神经网络
Fault detection
Stereo space functions
Neural network
