Keywords: classification, finite elements, artificial neural network, self-organizing map, learning vector quantization, XML schema.

Most of the finite element packages use different element types and formulations, by means of a variety of methodologies, tools and specifications. Consequently, sometimes it is necessary to use distinct finite element codes to run a given problem in order to complement or validate the analyses results obtained. However, this may not be easy to accomplish due to the fact that there is not a standard file format for the interchange of data amongst the variety of finite element programs currently in the market. Furthermore, there are not enough facilities or interface programs in these systems to allow for the data exchange between them.

On the other hand, it is possible to use a neutral format, recognisable by the different finite element programs, to share the program proprietary data files. This is so because, although the finite element programs have elements developed as mentioned above, the element definitions basically represent a group of rules to treat the basic characteristics of the elements. The standard format could improve the whole analysis process since a data input file could also be used as input for a different system, regardless of the source and the target, subsequently minimising errors while exporting data between different systems.

The growth and the extensive use of eXtensible Markup Language (XML) as the basis for different languages and in the creation of vocabularies for the exchange of information among several applications is already a consolidated fact. The XML proposal focuses not only on data structure, but also on its validation through the creation of vocabularies known as XML Schemas. These characteristics make it a natural candidate for the construction of neutral and expandable formats in order to facilitative in the data import and export process among the different applications. The finite element markup language (FEML) is a XML Schema proposed by Moita and Pinheiro [1] to describe a neutral and structured standard for the exchange of data in finite element analyses.

This work presents the basis of the FEML proposal and also an intelligent tool for the recognition and classification of finite elements families used in conjunction with this schema. The need for the classification becomes clear due to the necessity of a mapping between the several elements in a given code and the FEML format. A tree-type search could be used but it can lead to a lengthy if-then-else sequence to perform the matching. The use of a database-type classification could also be an option, however the search mechanism would still be required. Moreover, the use of an intelligent-type classification is also able to accommodate changes in the element families to be classified and new elements can be added with minimum effort.

Here, the so-called intelligent classification uses the technology of artificial neural networks, specifically through of application the self-organizing map networks and the learning vector quantization. To carry out this task, the network receives the attributes of the elements and makes its classification, in terms of equivalent (or similar, less specialised) elements. At the end, it allows for the interface programs to "translate" between a specific (proprietor) format and the neutral format, in an intelligent way, minimizing errors due to manual data transcription and eliminating the necessity of searching the entire tree of elements.

1

G.F. Moita, M.C. Pinheiro "Towards a standard for finite element data exchange using XML", CD-Rom Conference Proceedings of the 7th U.S. National Congress on Computational Mechanics, Albuquerque, USA, 2003.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)