论文部分内容阅读
为了分析温度对大型桥梁模态频率的影响程度及规律,在桥梁的长期监测中,寻找一种直观、准确、具有可操作性的方法,以预测、滤除温度对模态频率的影响。以桥梁结构的温度及温差分布作为输入矢量,以模态频率作为输出矢量,建立了基于单截面温度分布和多截面温度分布的两个BP神经网络模型,进行拟合及预测效果的对比分析。结果发现,两模型均对模态频率进行了较好地预测、拟合,具有较强的泛化能力。基于多测面温度分布的神经网络模型,预测效果更好,平均相对偏差仅略大于千分之一。因此,温度分布、变化对桥梁模态频率有显著影响,BP神经网络模型能较好地拟合、预测频率随温度的变化,温度沿桥梁纵向分布的差异对模态频率的影响不可忽略。
In order to analyze the influence degree and rule of temperature on the modal frequencies of large bridges, in the long-term monitoring of bridges, an intuitive, accurate and feasible method is proposed to predict and filter out the effect of temperature on modal frequencies. Taking the temperature and temperature distribution of the bridge structure as input vector and two moduli as the output vector, two BP neural network models based on single-section temperature distribution and multi-section temperature distribution are established, and the fitting and prediction results are compared and analyzed. The results show that both models predict and fit the modal frequencies well and have strong generalization ability. Based on the multi-surface temperature distribution neural network model, the prediction effect is better, the average relative deviation is only slightly larger than one per thousand. Therefore, the temperature distribution and the variation have a significant impact on the modal frequency of the bridge. The BP neural network model can well fit and predict the variation of the frequency with the temperature. The influence of the temperature distribution along the bridge on the modal frequency can not be neglected.