Simultaneous prediction of delamination and surface roughness in drilling GFRP composite using ANN

Abstract

Delamination in the drilling of polyester composite reinforced with chopped fiberglass is a problematic phenomenon. The material’s structural integrity is reduced by delamination, which results in poor tolerance during assembly and is a primary reason for decreased performance. Surface roughness is another important factor to consider when drilling fiber-reinforced plastics, as surface roughness causes failures by inducing high stresses in rivets and screws. Due to the random orientation of fiberglass and the non-homogenous, anisotropic properties of this material, an exact mathematical model has not been developed yet. Instead, modelling by artificial neural networks (ANNs) is adopted. In the present work, a multilayer perception ANN architecture has been developed with a feed-forward back-propagation algorithm. The algorithm uses material thickness, drill diameter, spindle speed, and feed rate as input parameters and delamination factor (Fd) at the entrance of the drilled hole, average surface roughness (Ra), and root mean square surface roughness (Rq) as the output parameters. The ANN model is then used to develop response surfaces to examine the influence of various input parameters on different response parameters. The developed model predicts that surface roughness increases with increases in feed rate and that a smaller-diameter drill will be advantageous in reducing surface roughness. A reduced feed rate will minimize delamination as well.