【作者】 许建福；

【导师】 黄真萍；

【作者基本信息】 福州大学 ， 环境工程， 2004， 硕士

【摘要】 福建省房地产业和高速公路建设正处在蓬勃发展阶段，但由于地处山区、河流极其发育，许多建设工程都在山地、江边等高洼不平的地带建设与施工。鉴于工程项目地基设计和施工的要求，在地基处理时，有些高凸地带需要挖平、低凹地带需要大量的填土和强夯加固处理。强夯加固是填土地基的一种地基处理手段，是使地基土体密实、承载力提高的一种地基加固方法。强夯效果的好坏直接影响上部建筑结构的稳定性。因此，该建筑地基的强夯加固效果的检测就显得非常重要。现有的地基加固效果的检测方法主要是采用钻孔和标准贯入、动力触探和载荷试验等原位测试方法，对于大型的（或大面积的）强夯地基来说，测点相对较少，不能全面的反映地基强夯处理的效果。因此，完善与发展一种快速的勘测与资料处理分析方法成为必要。本文对瞬态瑞雷面波勘测和人工神经网络多参数非线性预测理论进行了深入探讨，研究形成了一套对强夯地基的压实度、均匀性以及强夯地基承载力检测和预估的方法。在深入掌握瑞雷面波理论和人工神经网络理论及其工程应用现状的基础上，系统地推导了均匀半空间介质和层状介质的瑞雷波方程和人工神经网络BP法公式，完善了瞬态瑞雷面波正反演处理软件，编制了基于人工神经网络BP法非线性处理软件。通过实际强夯地基处理工程的大量瑞雷面波资料的采集、资料的处理，以及成果的分析，对强夯地基的压实度、均匀性进行研究和系统性探讨；收集了原位测试数据，并结合瑞雷面波波速参数，开展多参数拟合和神经网络预测地基承载力的研究。本论文的创新点：（1）在瞬态瑞雷面波资料采集前，通过试验重点讨论和选择了瞬态瑞雷面波采集的激发、接收方式和观测因素等参数；（2）多参数神经网络综合预估地基的承载力。研究表明：瞬态瑞雷波法是一种无损、快速、经济、连续大面积的评价强夯地基加固效果检测方法，成功地克服传统勘探方法中勘测点稀疏、耗时多、耗费大的不足，并具有无损检测的优点。BP神经网络算法通过多参数的统计和多参数神经网络的学习和识别过程模拟出场地地基承载力，是一种多种试验数据综合的非线性计算方法，克服单一测试误差的影响，具备多种测试方法综合预估的效果。研究结果有助于深入认识瑞雷波特性和人工神经网络BP算法的原理，对推动瑞雷波法和BP算法在地基处理加固效果检测上的应用具有重要的理论及实际指导意义。更多还原

【Abstract】 Nowadays the real estate and construction of highway are flourishing in Fujian Province. A lot of projects are built on a site of the mountain area or along the rivers. In order to the foundation design and construction of the project, some high protruding area are dug up, while low concave area filled with soil are treated by dynamic compaction . Dynamic compaction is a kind of foundation treatments, which can improve the bearing capacity and compactness of soil mass. The effect of the dynamic compaction directly influences the stability of the superstructure, it is important for the foundation design to test the effect of dynamic compaction. The detection of foundation improvement now mainly depends on drilling hole test and standard penetration test, dynamic penetration test and plate loading test. But in the large area dynamic foundation, the relatively less measure points can’t totally reflect the effect of dynamic compaction. It is necessary to develop a fast way of exploration and data processing. The thesis deals with transient R-surface wave processing and artificial neural networks and develops a way of detecting and estimating the degree of compaction, uniformity, the bearing capacity of the dynamic foundation.R-surface wave theory and its engineering application are discussed. The R-surface wave equation and artificial neural network BP algorithm are deduced. A direct and inverse calculation processing of transient R-surface wave is improved ,an artificial neural network processing software is designed. Using mounts of R-surface wave data sampled by practical engineering of dynamic compaction. The compaction degree and uniformity of dynamic foundation are systematically discussed in the thesis. The collected data of in situ testing ,combined with R-wave velocity parameter, was used to develop the way to estimate the bearing capacity by multi-parameter and neural network. The innovation points in this thesis are: (1) Before the collection of transient R-surface wave data, selected the parameter of impulse, receive, and observation in the experiment. (2) Estimated the bearing capacity of foundation with multi-parameter. The results show that the R-surface wave method is non-destructive, fast detecting method, and can be used in large area dynamic foundation compaction detecting. It gets advantage of non-destructive detection, successfully handles the problems in the early exploration method, such as sparseness of measure points, high cost of money and time. BP neural network<WP=7>algorithm contains kinds of non-linear calculation method. It avoids the interference of single testing and acquires the effect of the combined testing data. By the statistic of multi-parameter and the recognition of multi-parameter, the treated foundation bearing capacity can be stimulated. It is helpful for us to reveal the characteristic of R-surface wave and the principle of artificial neural networks, and improve the application of R-surface wave and BP algorithm in the theory and practice.更多还原