Developing and Mining Materials Databases With Corrosion and Fatigue Data

Published

February 1, 2001

Author(s)

C P. Sturrock

Abstract

The development and mining of materials databases are described, with examples based on data from metals fatigue and corrosion experiments. Much of the research was conducted under the auspices of the NACE-NIST Corrosion Data Program, an industry-government partnership of materials engineers and scientists dedicated to extending the state-of-the-art in materials informatics and advancing the understanding of materials performance through the mining of corrosion databases. A database documenting fatigue crack propagation tests is described extensively, with screen shots illustrating the scope of the database, how the search criteria may be specified, and how the data are ultimately displayed. Results from a separate investigation of the mining of data classifying the presence/absence of stress corrosion cracking of austenitic stainless steels in various aqueous environments are presented. The evaluation of data-driven classifers in general and the sqecific methods investigated and applied in this research are briefly summarized, along with the results of the performance comparison of these methods. Of the methods investigated, decision trees were found to perform best overall, in terms of both accuracy and intelligibility.

Citation

Materials Science

Pub Type

Journals

Keywords

corrosion, data mining, fatigue, materials data, materials database

Citation

Sturrock, C. (2001), Developing and Mining Materials Databases With Corrosion and Fatigue Data, Materials Science (Accessed December 30, 2024)