摘 要
运行中的混凝土坝，本质上是一个复杂的非线性动力系统，目前的监测模型尚不能反映混凝土坝的非线性动力成分。混凝土坝变形监测资料蕴藏着大坝系统的本质特征，包括其混沌特性。通过对大坝变形监测资料中的混沌特性进行深入研究，建立相应的混沌分析和预测模型，对大坝安全监控具有十分重要的理论意义和应用价值。
本文以混凝土坝变形监测资料序列为研究对象，以混沌理论、相空间重构技术和人工神经网络为研究手段，重点研究了混凝土坝变形中的混沌成分，建立了与统计模型互补的变形混沌模型。
本文的主要工作有：
（1）研究了混沌特征量的提取方法，相空间重构参数的确定方法，通过对几种算法的比较分析，最终选取了适合数据长度较短且含有噪声的时间序列混沌分析算法。
（2）分析了混沌时间序列相空间的预测方法，提出了在统计模型基础上，分别结合自适应预测法和径向基函数神经网络的混沌预测模型。
（3）利用紧水滩大坝变形监测资料，通过建立统计模型提取残差序列，然后对残差序列进行混沌分析，重构残差序列的相空间，应用两类预测模型对其中的混沌成分进行预测，得到可以和统计模型相互补充的、有效的变形混沌预测模型。
目前，监测混沌模型的研究还刚刚起步，还有许多问题有待深入研究。如混沌模型的可预测尺度的提高问题，监测数据的降噪问题，与其他非线性理论联合进行预测等。
关键词：监测模型 混沌理论 相空间重构 变形预测 RBF神经网络
Abstract
The operating concrete the current monitoring model can not reflect the nonlinear dynamic components. The dam is a complex nonlinear dynamic system in essence, deformation monitoring data contains the essential characters of the concrete dam, including its chaotic characteristics. An in-depth study of chaotic characteristics is made by establishing the corresponding chaotic analysis and prediction models, which has great theoretical significance and utility value for the dam safety monitoring.
On basis of the deformation monitoring data sequence of concrete dam, and taking the chaos theory, phase space reconstruction technology and artificial neural networks as research means, to focus on the study of its chaotic component, a deformation chaotic-prediction model is established which complements with the statistical model. The main work is:
(1) Study on the extraction method of chaotic characters and definite method of parameters for reconstructing phase space. By comparing several of the algorithms, it selected some logical algorithms for time series chaotic analysis, whose data is in shorter length and contains noise.
(2) Analyses the phase-space forecasting method of chaotic time series; put forward to two chaotic prediction models, which combines Volterra prediction method and radial basis function neural network respectively basing on statistical model.
(3) Using the monitoring data of JinShuitan dam, taking the residual sequence into Chaotic-analyzing and phase-space reconstruction after establishing the statistical model, and then predicting the chaotic components of which by utilizing the two model upper, in order to get the effective deformation chaos-prediction model.
At present, the study on chaotic monitoring model has just started that there are still many issues should to be studying in-depth, such as the predictable size of prediction model, noise reduction of observation data, the problem of combining other nonlinear theory into predicting, and so on.
Key words: Monitoring model，Chaos theory，Phase-space reconstruction，
Deformation Prediction，RBF neural network.
文章仅提供中英文摘要！需要的可以联系作者本站。
