Prediction and Optimization of Machining Responses in Graphene Nanoplatelet-Reinforced Jute–Basalt Composites Using the SURF-XNet

Abstract

Jute–basalt hybrid composites reinforced with graphene nanoplatelets (GNPs) were created, and their machining properties were evaluated using explainable artificial intelligence techniques. The jute–basalt laminates were produced via hand lay-up and compression molding with a GNP loading of 1.5 % by weight. The laminates were evaluated for mechanical (tensile, flexural, and impact) and thermal (thermogravimetric) properties, as well as by using scanning electron microscopy (SEM). Drilling was selected as the machining process for the composites. Various factors were analyzed to determine their influence on the drilling process, including tool edge geometry, tool coating, fiber orientation angle, and moisture content. The experimental design and optimization were conducted using response surface methodology. The optimized solution resulted in a surface finish of 1.91 µm, a delamination factor of 1.36, an average fiber pull-out length of 27.61 µm, and a thermal damage zone of 51.11 µm, yielding a desirability of 0.90 and R2 values greater than 0.97. A new approach, called SURF-XNet, has been developed to increase the trustworthiness and interpretability of AI model predictions. This approach employs an ensemble of deep neural trees. It combines SHAP and layer-wise relevance propagation to determine how each machining parameter affects surface integrity responses during machining. This demonstrates a successful partnership between sustainable raw materials and explainable AI techniques for optimizing the machining of composites.