基于经验模态分解的多尺度大坝变形预测方法研究

发布时间：2018-05-10 20:28

本文选题：大坝变形监测 + 小波去噪　；参考：《华北水利水电大学》2017年硕士论文

【摘要】：大坝的安全监测已成为大坝建设和运营必不可少的工作。变形监测、分析和预报是大坝安全监测最重要的一部分,掌握大坝的变形规律,实现可靠、准确的大坝变形预测能够提前化解潜在的危机,为大坝的设计提供科学上的依据,因此大坝的变形预测具有十分重要的意义。本文主要工作如下:(1)介绍了大坝变形预测方法的国内外研究现状,介绍了小波去噪、经验模态分解、果蝇优化算法及BP神经网络算法分别在大坝变形预测中的应用。(2)介绍了大坝变形监测技术和方法、大坝变形监测原则和要求及大坝变形预测模型,以及官地水电站外部变形监测技术方案。(3)介绍了小波去噪和经验模态分解的理论,采用改进的分层小波去噪,并将改进的小波去噪方法与经验模态分解方法融合,提出了IMF1-Wavelet去噪方法,并成功应用于大坝变形数据去噪处理。(4)介绍了果蝇优化算法和神经网络的理论知识。对果蝇优化算法进行了改进,并对其寻优能力进行了仿真实验,针对BP神经网络自身系统性缺陷,提出用果蝇优化算法优化神经网络的权值和阈值的方法。(5)结合官地大坝变形监测实例,采用IMF1-Wavelet去噪方法进行去噪处理,对去噪后的数据进行经验模态分解,针对分解得到的变形分量分别建立FOA-BP预测模型,将分量的预测结果相加作为最终的预测结果,实现了“隔离”预测。通过预测的结果对比分析,证明了基于小波去噪和经验模态分解的外部平台构建和果蝇算法内部优化BP神经网络的组合预测模型,相对传统模型具有更好的预测精度。
[Abstract]:Dam safety monitoring has become an indispensable work in dam construction and operation. Deformation monitoring, analysis and prediction are the most important part of dam safety monitoring. Mastering dam deformation law, realizing reliable and accurate dam deformation prediction can resolve potential crisis in advance and provide scientific basis for dam design. Therefore, the prediction of dam deformation is of great significance. The main work of this paper is as follows: (1) the research status of dam deformation prediction methods at home and abroad, wavelet denoising and empirical mode decomposition are introduced. The application of Drosophila optimization algorithm and BP neural network algorithm in dam deformation prediction is introduced. The dam deformation monitoring techniques and methods, dam deformation monitoring principles and requirements, and dam deformation prediction model are introduced. The theory of wavelet denoising and empirical mode decomposition is introduced, and the improved wavelet de-noising method is combined with the empirical mode decomposition method. The IMF1-Wavelet denoising method is proposed and successfully applied to the denoising treatment of dam deformation data. The theoretical knowledge of the optimization algorithm and neural network of Drosophila is introduced. The optimization algorithm of Drosophila has been improved, and its optimization ability has been simulated, aiming at the systemic defect of BP neural network. A method of optimizing weights and thresholds of neural network using Drosophila optimization algorithm is proposed. Combining with the deformation monitoring example of Guandi Dam, the denoising method of IMF1-Wavelet is adopted to decompose the de-noised data by empirical mode decomposition. The FOA-BP prediction model is established for the decomposed deformation components, and the prediction results of the components are added as the final prediction results, and the "isolated" prediction is realized. It is proved that the combined prediction model based on wavelet denoising and empirical mode decomposition and optimized BP neural network in Drosophila algorithm has better prediction accuracy than the traditional model.
【学位授予单位】：华北水利水电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TV698.11

【参考文献】