Modeling of the APS plasma spray process using artificial neural networks: basis, requirements and an example

Thermal spraying is a versatile technique to manufacture coatings which offers a large choice of processes (i.e., plasma spraying, flame spraying, electric arc spraying, etc.) and materials (i.e., metallic, ceramic, polymer and composite materials). To obtain functional coatings exhibiting selected in-service properties, combinations of processing parameters have to be planned. These combinations differ by their cost and by their influence on the coating properties and characteristics. In order to control the manufacturing process, one of the challenges nowadays is to recognize parameter interdependencies, correlations and individual effects on coating properties and characteristics and influences on the in-service properties. This is why a robust methodology is needed to study theses interrelated effects. A statistical method, responding to the previous constrains, was implemented to correlate the atmospheric plasma spray processing parameters to the coating properties. This methodology is based on artificial neural networks which is a technique based on database training to predict property-parameter evolutions. This introductory work points out the implementation protocol, the database construction, the optimization process and an example of predicted results related to the deposition yield (i.e., deposited thickness per pass).