Design and analysis of computer experiments, structural equation modelling and wavelet analysis of daily flows

日常流量的计算机实验、结构方程建模和小波分析的设计和分析

• 批准号: 36329-2007
• 负责人: Lye, Leonard
• 金额: $ 1.46万
• 依托单位: Memorial University of Newfoundland
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2008
• 资助国家: 加拿大
• 起止时间: 2008-01-01 至 2009-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=390104
• 关键词: Design; analysis; computer; experiments; structural

This research proposal has four main objectives: 1) to investigate the use of recently developed design and analysis of computer experiments methodology for the surrogate modelling of complex environmental/civil engineering models. If successful, this surrogate model will reduce considerably the computing efforts for subsequent uncertainty, sensitivity, and Monte-Carlo analyses of complex computer models; 2) to investigate the possibility of using the response surface methodology techniques in extracting an explicit model from an artificial neural network. The absence of an explicit model associated with the ANNs makes it difficult for the networks to be reproduced and therefore not easily integrated to form coupled models for example. Explicit model extraction from various ANNs ranging in complexity will be investigated. The applicability and limitations of RSM will be comprehensively explored;(3) to investigate the application of structural equation modelling (SEM) for regional flow analysis and environmental modelling. The rapid progress made in terms of theory and software development warrants a new look at this method that has enjoyed exponential growth in usage in many fields of study. Regional flow analysis data and environmental monitoring data from Newfoundland will be used as case studies. The proposed method will also be compared to currently used methods; and 4) to investigate the application of wavelets analysis methods for daily river flow modelling. Wavelets has been found to be especially useful for analyzing time series with non-stationary or transitory characteristics such as drifts, trends, abrupt changes, self-similarity, and beginnings and ends of events, much like daily river flow series. This proposal will compare the proposed method with traditional methods. The results of this research will be especially useful to engineers involved in hydrological and environmental modelling. It will provide more efficient tools to tackle the increasing complex problems that engineers face.

这项研究建议有四个主要目标：1)调查最近开发的计算机实验设计和分析方法在复杂环境/土木工程模型代理建模中的应用。如果成功，这种替代模型将大大减少后续的不确定性、敏感性和复杂计算机模型的蒙特卡罗分析的计算工作量；2)研究使用响应面方法技术从人工神经网络中提取显式模型的可能性。例如，由于缺乏与ANN相关联的显式模型，使得网络很难被复制，因此不容易集成以形成耦合模型。从各种复杂程度不同的人工神经网络中提取显式模型将被研究。将全面探讨结构方程模型的适用性和局限性；(3)研究结构方程模型在区域水流分析和环境模拟中的应用。在理论和软件开发方面取得的快速进步需要对这种方法进行新的审视，这种方法在许多研究领域中的使用已经呈指数级增长。将使用纽芬兰的区域流量分析数据和环境监测数据作为案例研究。该方法还将与目前使用的方法进行比较；以及4)研究小波分析方法在日常河流径流模拟中的应用。小波被发现特别适用于分析具有非平稳或瞬时特征的时间序列，如漂移、趋势、突变、自相似性以及事件的开始和结束，非常类似于每日河流流量序列。该提案将把所提出的方法与传统方法进行比较。这项研究的结果将对从事水文和环境模拟的工程师特别有用。它将提供更有效的工具来解决工程师面临的日益复杂的问题。