直升机神经网络模型辨识研究

Neural Network for Model Identification of Helicopter

研究直升机飞控系统的控制问题,直升机是强耦合的非线性系统,直升机的模型的建立是研究整个飞控系统的基础。为改善飞行姿态误差,提高稳定性,建立其较为精确的动力学模型对于设计、验证飞行控制系统具有重要意义。根据直升机动力学方程的形式,建立了一种基于BP神经网络的非线性辨识模型,利用典型状态下的试飞数据作为样本,采用LM优化学习算法对神经网络进行训练,并用其他飞行状态的数据对模型进行验证。计算结果表明,证明所建立的神经网络模型具有较好的拟合精度与泛化能力,能够较好地反映直升机的动态特性,并为设计提供依据。

Helicopter is a strongly coupled and nonlinear system,The model of Helicopter is the foundation of the whole flight control system,so it＇s very important to build an exact model for designing and validating the flight control system.According to the dynamic equation of helicopter,a BP neural network model is established for describing the characteristic.The flying data in given condition is used as the sample,the LM algorithm is used for training the model,and the data in other conditions is used for validating the model.The computation result shows the model is conformable and flexible with the data,and is competent for describing the helicopter＇s behavior.