Application of Neural Network Analysis to Correlate the Properties of Plasma Spray Coating

Abstract

Thermal spray coatings are more often being demanding process at the recent stages of industrial design processes to become fundamental element of the engineering system. The aim of the present paper is to develop a model-based estimation and control for regulating the coating adhesion strength, by using neural network. This proposed model permits cost reduction by the possibility of adjusting the parameter of the process for each of the desired properties, which can directly act as a quality control process. By this artificial neural network (ANN) technique, the possibility of adhesion strength of the mixture of fly-ash, quartz and illmenite (which is deposited on the mild steel and copper substrates) can be predicted as well as adhesion strength, merits of ease of fabrication, and high quality deposits, which optimize the amount of specified coating materials to achieve the desired properties. This technique involves database training to optimize the property parameter evolutions in processes having a large number of interdependent variables such as plasma current, voltage, powder feed rate and travel speed in this plasma spray coating deposition. By this neural network, it is observed that the coating adhesion strength largely depends on arc current, voltage, torch to base distance, powder size, etc.