This study was developed to provide an innovative methodology for designing steel alloys. Information on the general corrosion of a wide number of steel alloys in various electrolytes and environments was obtained from the National Institute of Standards and Technology (NIST). Parameters such as pH and conductivity used in each experiment (alloy in contact with an environment), along with alloy composition (from UNS number), and corrosion rates, were all collected in a single data row, or vector. To cluster by similarities, a web-based, publicly available Kohonen mapping software was used to perform the clustering analysis; Kohonen maps work by clustering together similar vectors and separating those vectors that differ. A vector was formed for each experiment for which corrosion rates were recorded; 1521 experiments were performed and each of those vectors was used to train the Kohonen Map. Once the Kohonen map is trained, each one of the cells forming the two-dimension Kohonen map will form clusters of vectors. Vectors containing similar information will be clustered together while dissimilar vectors will be clustered separately on the Kohonen map. The cells of the Kohonen map will adopt a “prototype” vector to be the representative of that cell; the prototype vector adopts the average values of all stored vectors in that cell. After the Kohonen map is trained, new vectors containing fabricated metal alloy composition (steels) and environment information can be input into the map. These new vectors, even though they do not contain corrosion rates, can be classified by the Kohonen map and entered into a cluster on the map. This methodology can be use to explore “if-then” scenarios of a new alloy in a different environment as well as obtain an expected corrosion rate of that particular alloy in that particular environment. Preliminary results of the trained Kohonen map are shown and discussed. The map results are used to explore the effects of the experiment environments and alloy composition on the general corrosion of the stainless steels.
- Pressure Vessels and Piping Division
Data Mining of General Corrosion of Most Commonly Used Iron-Based Alloys Using Kohonen Mapping
- Views Icon Views
- Share Icon Share
- Search Site
Kirsch, KL, & Urquidi-Macdonald, M. "Data Mining of General Corrosion of Most Commonly Used Iron-Based Alloys Using Kohonen Mapping." Proceedings of the ASME 2010 Pressure Vessels and Piping Division/K-PVP Conference. ASME 2010 Pressure Vessels and Piping Conference: Volume 5. Bellevue, Washington, USA. July 18–22, 2010. pp. 447-453. ASME. https://doi.org/10.1115/PVP2010-25509
Download citation file: