基于监测资料的海堤非稳定渗流反演研究

发布时间：2018-06-22 06:50

本文选题：海堤渗压 + 非稳定渗流　；参考：《合肥工业大学》2015年硕士论文

【摘要】：海堤工程对维护沿海地区的安全,保障堤内人民的生产、生活,推动堤内的经济发展起到积极作用。随着海堤工程管理工作的推进,加强海堤安全监控,研究海堤工作运行状态,成为保障海堤安全的重要工作内容。渗压的分布状况对海堤的安全运行有重要影响,本文以实际监测资料为依据,研究了变化潮位下的海堤渗压分布状况。受快速变化潮位的影响,海堤的渗压也处于快速变化之下。鉴于这种快速变化的渗流场的影响,对海堤进行非稳定渗流分析显得尤为重要。本文从水流连续性原理与达西定律出发,研究了适用于海堤非稳定渗流的基本微分控制方程,并依据变分原理分析了海堤非稳定渗流分析的有限元分析方法。研究中以变化的潮位作为外荷载,并基于ANSYS有限元软件热分析与渗流分析的相似性,通过ANSYS软件进行迭代计算,最终获得各个时刻下的海堤渗压分布,为合理分析海堤的渗压状况奠定了基础。传统的渗透系数获得方法主要有野外试验法、室内试验法等,而随着监测技术及计算分析方法的进步,反演分析广泛应用于渗透系数的获取。本文分别采用数值优化反演方法和神经网络参数辨识方法反演海堤的渗透系数,数值优化反演方法以实际监测的渗压时序与模拟渗压时序构建关于海堤渗透系数的最小二乘目标函数,再采用逐步扫描法逐渐缩小渗透系数的范围,最后再利用共轭梯度算法迭代优化,获得最优的海堤渗透系数。神经网络可以有效的建立海堤测点渗压时序与渗透系数的映射关系,本文以数值模拟技术获得网络的样本,再通过改进的BP神经网络进行训练,构建海堤渗压时序与渗透系数的参数辨识模型,最后利用实测渗压时序代入训练好的神经网络中可计算出海堤的渗透系数。将两种方法获得的渗透系数分别进行数值模拟计算并提取相应测点的渗压时序与实测渗压时序比较发现,测点的模拟渗压时序与实测渗压时序的平均相对误差较小,曲线振幅及变化规律拟合非常好,说明采用数值优化反演方法和神经网络参数辨识方法均可有效地获得海堤的渗透系数,为揭示海堤渗压分布状况提供了重要依据。通过对两种反演方法的研究发现,数值优化反演方法围绕有限元分析程序展开,其思路清晰、简洁可操作性强。而神经网络参数辨识方法将渗压与渗透系数的联系信息构建于网络的连结权值中,在获得一定的样本数量下即可有效反演出渗透系数,其反演结果理想、稳定,具有良好的应用前景。
[Abstract]:Seawall engineering plays a positive role in maintaining the safety of coastal areas, ensuring the production and life of the people in the embankment, and promoting the economic development in the embankment. With the advance of seawall engineering management, strengthening the safety monitoring of seawall and studying the working state of seawall become the important work content of ensuring the safety of seawall. The distribution of seepage pressure has an important effect on the safe operation of the seawall. Based on the actual monitoring data, the distribution of seepage pressure in the sea embankment under the changing tide level is studied in this paper. Under the influence of rapidly changing tide level, the seepage pressure of seawall is also changing rapidly. In view of the influence of the rapidly changing seepage field, it is very important to analyze the unsteady seepage of the seawall. Based on the principle of continuity of water flow and Darcy's law, this paper studies the basic differential governing equation for unsteady seepage of seawall, and analyzes the finite element analysis method of unsteady seepage of seawall according to variational principle. Based on the similarity of thermal analysis and seepage analysis of ANSYS finite element software, the seepage pressure distribution of seawall is obtained by iterative calculation with ANSYS software. It lays a foundation for reasonable analysis of seepage pressure of seawall. The traditional methods of obtaining permeability coefficient mainly include field test method and indoor test method. With the development of monitoring technology and calculation and analysis method, inversion analysis is widely used to obtain permeability coefficient. In this paper, numerical optimization inversion method and neural network parameter identification method are used to inverse the permeability coefficient of seawall, respectively. Numerical optimization inversion method is used to construct the least-square objective function about seawall permeability coefficient by using actual monitored seepage time series and simulated seepage pressure time series, and then gradually reducing the scope of permeability coefficient by means of stepwise scanning method. Finally, the optimal seawall permeability coefficient is obtained by iterative optimization using conjugate gradient algorithm. Neural network can effectively establish the mapping relationship between seepage pressure time series and permeability coefficient of seawall points. In this paper, the network samples are obtained by numerical simulation technique, and then trained by improved BP neural network. The parameter identification model of seepage time series and permeability coefficient of seawall is constructed. Finally, the permeability coefficient of seawall can be calculated by using the measured seepage pressure time series in the trained neural network. The osmotic coefficients obtained by the two methods are numerically simulated and compared with the measured osmotic time series. It is found that the average relative error between the simulated osmotic pressure timing and the measured osmotic pressure time series is smaller. The curve amplitude and variation law are well fitted, which shows that the seepage coefficient of seawall can be obtained effectively by using numerical optimization inversion method and neural network parameter identification method, which provides an important basis for revealing the distribution of seepage pressure of seawall. Through the study of the two inversion methods, it is found that the numerical optimization inversion method is developed around the finite element analysis program, and its thinking is clear and simple and operable. The neural network parameter identification method constructs the connection information between osmotic pressure and permeability coefficient in the connection weight value of the network and can effectively reverse the permeability coefficient when a certain number of samples are obtained. The inversion result is ideal and stable. It has good application prospect.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U656.314

【相似文献】