Analysis of Profile and Surface Roughness of Holes Drilled in Basalt Fiber Reinforced Polymers Laminates: Statistical Analysis and Predictive Approach Based on Machine Learning

Fiber-reinforced polymers such as basalt fiber-reinforced polymers (BFRP) can be used in structural parts, which often require assembly operations. Thus, the surface quality after drilling operations is especially important. BFRP laminates have been drilled with three different tools, and their profile roughness and surface roughness have been evaluated by analyzing the following variables: average roughness (Ra), maximum height of profile (Rz), arithmetic mean height (Sa) and maximum height (Sz), by means of an optical system. The optical measurement of surface roughness has been hampered by fiber breakage. A statistical analysis has allowed for developing a general linear model that predicts the values of variables. The fitted model for Ra and Rz has a variation coefficient of 97.00% and 95.58% respectively, while that 91.74% and 68.02% for Sa at the inlet hole and outlet hole respectively; and 86.08% and 82.22% for Sz at the inlet hole and outlet hole respectively. Additionally, different machine learning regression algorithms have been applied using different configurations to establish prediction models of the main rugosity parameters. In this way, linear methods, Gaussian regression methods, Support Vector Machines, and fine trees have been applied using the rotation speed, feed rates, and tool as features. Also, a neural network has been optimized and applied for the same goal. The methods yielded satisfactory prediction results within the tested experimental domain for some roughness parameters. Although the behavior of all variables is similar across all drill bit types, drill bits with a point angle of 120° provided better results.