基于BP神经网络的非平稳激励机械系统响应预估

利用ＢＰ人工神经网络进行非平稳激励条件下机械系统动态响应评判指标值与系统设计参数之间函数的逼近，采用收敛效率较高的学习率自适应调整算法，利用少数几次的方程求解结果进行网络训练，便可得到高效的响应评估映射，较多地减少了动态优化中评判值的计算规模。

BP Networks Response Evaluation Method for Nonstationary Stochastic Excited Mechanical System

Four layer BP networks is constructed to evaluate the response of mechanical system excited by nonstationary stochastic loads. By use of adapting modulation learning algorithm and only a few results to train the networks, response evaluation mapping can be obtained in the solution of system dynamical equation. The comparison results prove that this method is applicable and high efficiency.