【摘要】 提出一种利用神经网络结合自回归滑动平均时间序列分析来提取结构或系统模态参数的方法。这种方法首先运用神经网络计算出自回归滑动平均时间序列分析中的未知系数,然后提取出系统的模态参数。数值仿真及实验结果表明,该方法具有很高的识别精度、对噪声不敏感和具有很强的抗噪性的特性,这几点是传统时域辨识方法不具备的。文中提出的方法适合提取出实际系统的模态参数,对实际测量信号分析具有较好的参考价值。更多还原

【Abstract】 A novel artificial neural network(ANN)-based technique employing auto regression moving average(ARMA) time series method will be used to extract structural modal parameters.It takes advantage of the output-only vibration signals to obtain the modal parameters.Firstly,the method uses ANN to calculate the unknown coefficients in ARMA form,and then,it extracts the modal parameters from the calculated matrix before.Both numerical evaluations and experimental results demonstrate that the ANN-based method identifies structural modal parameter accurately and robustly.Furthermore,it has a more robust performance than the traditional time domain method.Thus the proposed method is very meaningful to the real measured vibration signals,and it is very suitable to extract the modal parameters from the real system.更多还原