Artificial neural networks for surface modification of cobalt based tungsten cemented carbide deposits

Abstract

This paper presents the application of artificial neural networks (ANNs) with statistical experiments to model and characterise WC/Co deposits of the plasma sprayings. In this study, the eight control factors were designed in a L18 factorial orthogonal array, and the effects of process conditions on the surface morphology were critically reviewed in the experiments. The surface topography properties and microstructure were studied.

A gradient steepest descent algorithm in the trained ANN was used to explore the relationships between variables and responses. Artificial neural network modelling for WC/Co coatings estimation is compared by response surface methodology. The best values obtained were 2·164 and 2·871% of error percentage for the surface roughness by the best ANN and the response surface methodology model respectively. The experimental results indicate that using a statistical experiment coupled to an ANN strategy offers an effective, efficient and adaptive approach for developing a robust and highly efficient plasma sprayed process of high quality.