Least square support vector machine alternative to artificial neural network for prediction of surface roughness and porosity of plasma sprayed copper substrates

Abstract

Plasma spraying technique has become a subject of intense research in many industrial structural/functional applications because its peculiarity surface properties. This investigation explains about plasma sprayed copper surface property. Here industrial waste and low grade ore (i.e. Flay-ash+ quartz+ illmenite), used as deposit material which is to be coated on copper substrates. In many applications, it is found that for structural modification, surface roughness & porosity parameters are very important. To decrease both surface roughness and coating porosity by optimizing other necessary properties, different soft computing methods like Artificial Neural Network (ANN) and Least Square support vector machine techniques used. The least square formulation of support vector machine (SVM) was recently proposed and rooted in the statistical learning theory. This technique potentially describes the approximation complexity of inter-relations between different parameters of atmospheric plasma spray process and helps in saving time & resources for experimental trials for which it is advantageous than all conventional methods. It is marked as a new development by learning from examples based on polynomial function, neural networks, radial basis function (RBF), splines or other function. From this above two methods (Multilayer Feed forward Neural Network& LS-SVM), we conclude that LS-SVM with RBF kernel gives better performance over ANN for prediction of surface roughness and coating porosity with minimum Mean Square Error. This methodology can provide clear understanding of various corelationships across multiple scales of length and time which could be essential for improvement of product performance and process. The results of this methodology give a good generalization capability to optimize the coating surface roughness& surface porosity.