【作者】 陈冲；

【导师】 尼古拉斯； 周庆国； 周剑；

【作者基本信息】 兰州大学 ， 计算机技术， 2012， 硕士

【摘要】 计算机建模已经被应用到包括自然科学、农业科学、医药科学等各个科学领域。它最初被作为其他的方面研究的补充,但当人们发现它的重要性之后,它便作为一门单独的课题被使用得相当广泛。至今,已经发展成为进行科学研究的一个重要方法。不同研究领域的专家们已经开发了许多针对自己所关心问题的模型。但是,人们发现,仅仅将所关心的某一个或几个物理过程进行建模分析并不能很好地来模拟实际情况,往往会有很大的误差。因此,我们需要将更多的物理过程加入模型中。随着人们对软件需求的不断增加,软件的功能、可操作性、智能化程度也迅速发展,从而使软件变得更加庞大、更加复杂,开发难度也越来越大,开发周期越来越长、参与开发人数越来越多。为了解决这种软件危机,减少重复劳动,降低被开发软件出错的概率,我们需要对已经存在的一些模型进行复用。本研究设计出一个基于组件的集成建模环境,建模环境由数据输入／输出,组件／模型库,控制组件,模型运行,模型分析五部分组成。模型库中保存组件或者模型供建模使用,用户也可以开发新的组件或者模型加入到模型库当中。建模时,将组件从组件／模型库中取出,组装,生成模型工程文件；然后加入所需输入数据、参数以及情景条件,模型就可以运行并生成模拟结果。对建模框架内模型组件的自解释功能(即不需要阅读组件的源码或者专门的组件说明文档,而通过简单的关键字或者信息便可以了解组件的数据需求、组件功能等建模过程需要的信息)进行了较大改进。OMS最初设计是用XML来实现组件的自解释功能的,而本研究是通过在组件开发时加入annotation实现的。使用annotation+POJO来替换OMS中的API,这样开发者不需要对框架本身的技术细节有深入了解,只需要具备很少的必要的编程知识就可进行模型开发。本建模环境开发中加入了WOFOST模型、HYDRUS-1D模型和MODFLOW模型来进行试验,以验证本建模环境的设计以及数据传递机制的正确性。单模型模拟验证中,实验数据用的是临泽站1991年全年的观测数据,WOFOST使用Boogaard等于1998年开发的WOFOST1.51版本；HYDRUS-1D模型试验使用J.Simunek等开发的HYDRUS-1D1.14版本；MODFLOW使用由Mc Donald等于2004年开发的MODFLOW-20001.15.00版本。试验的结果证明了本建模环境软件模型运行、数据传递机制的正确性。更多还原

【Abstract】 Modeling is already applied to many fields include natural science, agricultural science and medical science. It was originally a complement of other researches, but after its importance to studies was discovered, it was used quite widely. So far, computer modeling has become an important method of scientific research.Different researchers have developed a number of models about their own concerns. However, it is found that only concern one or several physical processes is not good enough to simulate the real world. Therefore, we need to add more physical processes into the model to meet the increasing demands for software. With the software functionality, operability, intellectualization developing faster and faster the software has become much larger and more complex. The difficulty of developing models is becoming more and more, the time spend on the developing is becoming longer and longer, the number of people engage on developing is becoming more and more. In order to solve the software crisis, and reduce duplication of effort, reduce the probability of error when developing software, we need to reuse some models already exist.This study designs a component-based IME. This IME composes of five parts, data I/O, component/model library, control components, model operation and model analysis. Model library is set up to organize components or models. When modeling, components are retrieved from the model/component library, and then resembled to build a model which can be run with inputs of forcing data and necessary parameters. The self-explanatory of components/models in the framework (which means there is enough information attached with the components for the IME framework to manipulate them) is improved a lot in this study. Initially, OMS achieve this function by adding XML files to each component/model, but here we find another solution which is annotation. By replacing OMS API with annotation and POJO, the modelers need not to know much about the framework in detail, they can develop their scientific models with little essential programming knowledge instead of learning a lot of sophisticated skills for software development.Three models, WOFOST, HYDRUS-1D and MODFLOW, are used to evaluate this IME. The data from Linze station in1991is used as experimental data. For WOFOST, we use WOFOST1.51version developed by Boogaard et al in1998; The version of HYDRUS-1D is HYDRUS-1D1.14, which was developed by Simunek et al; We use the version MODFLOW-2000that developed by Mc Donald in2004. The designed test shows that the IME is able to run as expected and is capable to assist in modeling practices by providing a powerful and flexible software environment.更多还原