Prediction of surface finish in extrusion honing process by regression analysis and artificial neural networks

The current work explores the influence of process parameters such as mesh size and volume fraction of abrasives with number of passes, on the interior surface quality of a pre machined component by extrusion honing process. The finishing process is highly flexible and unconventional while modifying the surfaces in case of miniature components involving complex profiles. The method is extensively used to deburr, polish, edge contour and removing recast layers by producing compressive stresses. By, the pressurized flow of semi viscous abrasive laden across the surface to be processed. The experimental study has been carried out on Inconel-625 alloy by one way EH process, with the carrier medium silicone polymer blended with SiC as abrasives. Experiments are planned by constructing L27 orthogonal array for the factors such as mesh number 36, 46, 54 and volume fraction 40, 50, 60 % of abrasives followed by number of passes 5, 10 and 15. Also, the study focuses in developing a regression model, training neural network and comparison of experimental R_a with both regression and ANN model. The prediction of R_a is accomplished by developing a linear regression model and a feed forward back propagation neural network model. Both the developed models are able to predict the output response with an error of 5 to 12%.