ANN-Based Estimation of the Defect Severity in the Drilling of GFRP/Ti Multilayered Composite Structure

Abstract

The main purpose of this study was to develop a model for predicting the quality of holes drilled in the root part of the spar of helicopter main rotor blades made of glass fiber-reinforced plastic (GFRP)-Ti multilayer polymer composite. As the main quality criterion, delaminations at the entry and exit of the drill from the hole were taken. In the experimental study, a conventional drill and two modified geometry drills, a double-point angle drill and a dagger drill, were used. Preliminary experiments showed the best hole quality when using modified drills, which allowed further detailed study only with both modified drills at different drilling speeds and feed rates. Its results in the form of training sets were used to build artificial neural networks (ANNs) to predict delamination at the entry and exit of the drilled holes. An analysis of the fitted response functions presented as 3D surface plots and contour plots led to the selection of the best tool, a double-point angle drill, which demonstrated the lowest achievable delamination both at the entry and at the exit of the holes approximately 1.5 times less (0.45/0.48 mm) compared to dagger drills (0.68/0.7 mm) and determined the ~5 times larger optimal area for the drilling speed and feed rate. The results obtained confirm the possibility of effective prediction of the quality and productivity of mechanically processed composites of complex reinforcement using ANN to quantify the quality criteria and search for the optimal modes of such technologies.