摘要
针对建立发动机动力学模型过程中,试车台机架结构边界环境的不确定状况,对神经网络在边界刚度识别中的应用进行了研究。以结构模态频率为网络输入,边界X、Y、Z方向的刚度为输出,通过一种增加训练样本的方法大大提高了网络的映射性能,最终的识别结果达到了预期目标,满足工程需要。
Because of the boundary condition uncertainty for the frame structure of rocket engine in establishing the dynamic model of engine, the characteristics of structural stiffness boundary will be identified with neural network in this paper. It would take the modal frequencies of engine as inputs, and the X, Y, Z-axis of stiffness boundary as outputs. Then the way of adding training samples improve the learning speed and mapping capability. Finally, the results reach the prospective goal, and the FEM （Finite Element Method） calculations can meet the engineering requirements.
出处
《火箭推进》
CAS
2009年第4期30-33,共4页
Journal of Rocket Propulsion
关键词
发动机机架
边界刚度
参数识别
BP神经网络
frame of rocket motor
stiffness boundary
characteristics identification
BP neural network
