【作者】 司春棣；

【导师】 李书全；

【作者基本信息】 河北农业大学 ， 水利水电工程， 2004， 硕士

【摘要】 就我国水库大坝工程监测及自动化管理的现状而言，土石坝工程是一个薄弱环节，不少大坝在带病运行，工程隐患较多，因此及时准确地对土石坝安全监测资料进行分析并建立监测模型，以识别土石坝的运行状况，监控其安全，是十分必要的。而遗传算法、神经网络等人工智能技术的成熟及迅猛发展为上石坝安全监测资料分析提供了新的理论和技术上的支持，依赖于遗传算法基于“优胜劣汰”机制的全局概率搜索特性及神经网络强大的非线性映射能力，本文做了以下主要工作： 首先，应用分层遗传算法对土石坝变形监测中的漏测沉降进行优化计算，并将漏测沉降作为沉降曲线回归模型中的回归因子，克服了常规方法对实际沉降过程描述差而造成漏测沉降计算误差较大的缺点，研究结果表明计算出的漏测沉降更接近于事实。 另外，在土石坝测压管滞后时间难以确定问题上，根据库水位与测压管水位扣除滞后时间后基本满足线性关系原理，提出了基于遗传算法思想的滞后时间优化计算方法，其中库水位与测压管水位过程线的拟合由径向基函数神经网络来实现。不仅实现了电算化，而且实例表明该方法是合理有效、简单可行的。 针对以往监测模型预测精度难以保证的的问题，本文在分析比较了几种土石坝安全监测预报模型的基础上，提出了一种新的安全监测预报模型——遗传回归模型，该模型通过遗传染色体对建模因子进行优选，并在适应度函数中综合考虑模型拟合精度与预测精度平衡。通过实例对模型进行验证，证明该模型是高效可行的，在保证一定拟合精度的基础上，实现了模型预测精度的提高。 最后，总结了本文的研究工作，对今后的研究进行了展望。更多还原

【Abstract】 As to the actuality of safety monitoring and modern management of dam in our country, earth-rock dam is the feeble section. Many of them have more hidden problems, function in gear, so it is necessary to analyze the prototype observed data and to establish the monitoring model accurately and timely for discerning the operation conditions and controlling its safety. And analysis of safety monitoring data were offered new theoretical and technological supports because of Genetic Algorithm and Artificial Neural Network, based on the development of artificial intelligence. Herein lies the overall probability searching characteristic on Genetic Algorithm which is based upon superior selected and inferior eliminated mechanism, and the powerful ability in nonlinear mapping of Artificial Neural Network, the main work is as follows:In this paper, omission settlement of earth-rock dam is optimized with Hierarchic Genetic Algorithm by setting the omission settlement as a new regression gene, the result has testified that the value attained from the HGA model is better than the one from early model which is far away from the real settlement process.Then, concerning the difficulty of ascertaining hysteresis time of piezometric tube, Genetic Algorithm-Radial Basis Function Neural Network model is developed according as the method that reservior level is rectilinear correlation with piezometric tube level, which not only could be calculated by computer but also is showed logical and feasible.Mainly, in order to resolve the problem of prediction accuracy, after analyzing and comparing several safe monitoring and forecasting models, a newmodel--Genetic Regression model is suggested, which the genes used to erectthe earth-rock dam safety monitoring and forecasting model are selected properly through genetic chromosome and the balance between simulation and prediction accuracy is considered in fitness function, finally the insurance of the model’s simulation precision and the improvement of the model’s forecasting precision is demonstrated by an example.The last section draws a conclusion from the whole research and take a look into the forth-coming work on the analyzing of earth-rock dam safe monitoring data.更多还原