中文摘要：

穿越地面沉降严重区域的高速铁路受工程沉降和区域沉降的耦合影响,准确预测高铁工后沉降的发展趋势对高铁安全运营有重要意义。通过分析影响工后沉降的因素,结合动态神经网络原理,以基准点、工作基点2个指标作为网络输入,以历史沉降数据作为延迟量反馈,用贝叶斯正则化算法训练网络,得到工后沉降的仿真非线性网络。应用此模型在沧州市沉降漏斗区进行沉降预测,以桥墩沉降量作为工后沉降的表征,和传统的双曲线法和灰色预测等模型对比。结果表明,动态神经网络考虑了区域沉降的影响,能更准确的预测工后沉降的发展趋势,具有很高的预测精度。

英文摘要：

As high-speed railway that went through areas with severe regional subsidence was subjected to, the coupling effects of the engineering and regional subsidence, accurate prediction of the development trend of post-construction subsidence of high speed railway has important significance to the safe operation of high-speed railway. Through analysis of the factors affecting the post-construction settlement and combination of dynamic neural network principle, based on Bayesian regularization algorithm, the simulated nonlinear network of the post-construction settlement was obtained, with the bench marks and working reference points as network input, and historical settlement data as delay feedbacks. The model was applied to the subsidence funnel area of Cangzhou city for settlement prediction. The post-construction settlement represented by pier subsidence was compared with the traditional forecast methods such as hyperbolic method and grey model. The results showed that dynamic neural network with consideration of regional subsidence can predict the developing trend of the post-construction settlement more accurately and has a high prediction accuracy.