摘要
以某型涡轴发动机为研究对象,提出了基于卡尔曼滤波器的方法进行发动机气路部件性能诊断。结合发动机实际运行情况,讨论了参数的合理选择以及非线性模型线性化程度对卡尔曼滤波结果的影响,考虑到非线性模型与真实发动机之间存在的建模误差会导致卡尔曼滤波器诊断失效,本文采用加入神经网络模块来补偿建模误差的方法,使模型能够真实反映发动机工作状况。实验表明,此方法可以提高故障诊断系统的置信度。
A method of a turbo-shaft engine gas path perfotnalanoe diagnosis based on Kalman filter is proposed. Selection of sensors and the estimation influence of linearization of nonlinear model are discussed in this paper. Considering the diagnosis fault of Kalman filter caused by the error between non-linear model and real engine, a method of utilizing neural net- wort to compensate the modeling error is proposed. It allowed the model to match the working condition of real engine. The result of simulation indicates the oonfidenco level can be improved by this method.
出处
《长春理工大学学报（自然科学版）》
2010年第3期33-36,58,共5页
Journal of Changchun University of Science and Technology(Natural Science Edition)
关键词
涡轴发动机
状态变量模型
性能估计
卡尔曼滤波
神经网络
turbo-shaft engine
state variable model
Kalman filter
health parmneter estimating
neural network
