论文部分内容阅读
为实现航空发动机在全包线的解耦控制,在飞行包线内选择了若干点,使用遗传算法对单神经元自适应解耦控制器的比例系数进行了离线优化。以优化得到的若干组参数为训练样本,离线训练径向基函数(RBF)神经网络,训练后的网络可映射高度、马赫数与比例系数之间的非线性关系,飞行包线内任意点的解耦控制器比例系数即可由该网络得到。仿真表明:在设计点和非设计点,系统均具有良好的动态特性和解耦特性。该方法结构简单、易于实现,具有实用价值。
In order to realize decoupling control of aeroengine in all-enveloping line, several points were selected in flight envelope, and the proportion coefficient of single-neuron adaptive decoupling controller was optimized by genetic algorithm. Based on the training set of several optimized parameters, the radial basis function (RBF) neural network for offline training, the network mappable height after training, the non-linear relationship between Mach number and proportional coefficient, The decoupling controller scaling factor can be obtained from this network. The simulation shows that the system has good dynamic characteristics and decoupling characteristics both at the design point and the non-design point. The method has the advantages of simple structure, easy implementation and practical value.