Multilayer Perceptrons and Radial Basis Function Networks for Corrosion Monitoring

Developing a model of aqueous corrosion of metal has proven to be a complex and intractable problem. Although the electro-chemistry of the exchange of electrons is well documented, the influence of other factors such as changes in water temperature, oxygen levels, and the levels of pH and alkalinity on the corrosion process are less well understood. As yet there is no model which adequately explains this interaction, because of the extreme non-linearity of the problem. One method of achieving this is through the use of artificial neural networks which are well established as a means of mapping complex non-linear relationships onto a desired output. However the best architecture for the extrapolation of data is very problem dependant. Because of the high dimensionality of the data sets, we have compared and contrasted two methods for eliminating highly correlated data sets. Our claim is that accurate regression modeling on such a complex problem can best be achieved using radial basis function networks, which have demonstrated a superiority over multi layer perceptrons for modeling highly non-linear problem surfaces, combined with genetic algorithms as an initial pre-processing step.

@InProceedings{AWM01, 
 	 author =  {Addison, J. F. Dale and Wermter, Stefan and MacIntyre, J.},  
 	 title = {Multilayer Perceptrons and Radial Basis Function Networks for Corrosion Monitoring}, 
 	 booktitle = {Proceedings of the International Conference on Artificial Intelligence and Applications},
 	 editors = {},
 	 number = {},
 	 volume = {},
 	 pages = {77--81},
 	 year = {2001},
 	 month = {Sep},
 	 publisher = {},
 	 doi = {}, 
 }