Knowledge acquisition and Support for Finite Element Modeling based on Case Database

作为著名的有限元分析程序，有几种通用程序（MARC，ABAQUS，NASTRAN，SAP等）。实现各种多功能性。为了使初学者能有效地编辑这些通用代码的输入数据集，一种解决方法是利用抽象的IF/THEN规则构成的基于咨询系统的经验知识支持。另一方面，输入案例数据的惯用模板被用作简单的学习或指导方法。由于有限元建模涉及的属性种类繁多，而且对物理条件的概括或理想化也需要先进的习惯知识，本文提出了一种基于汉明距离的相似性度量模型，并定义了向量化的属性编码。在案例检索系统的交互过程中，由于查询菜单的设计对于初学者来说是非常重要的， ...更多信息 应该研究查询过程的TIC性能。设计并实现了一个用HTML和CGI描述的查询菜单页面。并对实际的检索过程进行了实验研究，重点研究了查询关键字的选择过程和知识库的排序过程。提出了一种基于关系数据模型的错误案例检索模型。通过一个通用有限元程序的输出结果实例，验证了由错误状态符号、解释内容、参考语法信息组成的关系表的相关性。此外，由于一个原型错误诊断系统已经开发与perl和shell脚本，先进的可扩展性和易于翻译。针对案例的知识获取（学习模型），提出了一种基于层次神经网络的检索菜单智能自动生成方法。以几个实例为教学模式进行了不同的处理，展示了检索系统用户界面（菜单模式）自动设计的可能性，并进行了验证。少

As well-known FEA (finite element analysis) programs, there are several general purpose codes (MARC,ABAQUS,NASTRAN,SAP etc.) implementing various multifunctionalities. In order to edit efficiently for beginners input data set for these general purpose codes, empirical knowledge support based on consulation systems, constituted with abstracted IF/THEN rules, is a method of solution, And there are recently weveral developments for expert system or design system based on the hypermedia. On the other hand, idiomatic templates of input case data are utilized as easy learning or instruction methods. Since there are various attributes related to FEA modeling, and also advanced customary knowledge is needed to sumplify or idealize physical conditions, we habe proposed a metric model of similarity based on the hamming distance, defined with vectorized attribute codes. In the interactive process of this case retrieval system, since the design of query menus is important for beginners, the synthe … More tic performance of the query process shoule be investigated. In this work, a query menu page described with the HTML and the CGI was designed and implemented. And actual retrieval, focused on selecting procedures of query keys and sorting of knowledge base, have been investigated experimentally. An retrieving model of error case was proposed based on the relation data model. The pertinence of the proposed relation table composed of the error status symbol, the contents of explanation, the reference grammar information was recognized experimentally through examples of output results for a general purpose finite element program. Moreover, since a proto-type error diagnosis system have been developed with the perl and the born-shell scripting, the advanced extendibility and the easy translatability could be performed. For knowledg aqcuisition of cases (learning model), a method of intelligent automatic generation of the retrieval menu based on the hierarchical neural network was proposed. Several examples were processed variously as the teaching patterns, and the possibility of automatic design for the user interface of retrieval system (menu pattern) was shown, and its verification was carried out. Less