Multi-response optimization of drilling temperature and delamination in CNT-enhanced E-glass/epoxy laminates

In this study, multi-criteria optimization strategy is proposed to mitigate drilling-induced thermal loading and delamination damage in carbon nanotube (CNT) reinforced E-glass/epoxy laminated composites. The experiments were designed using an L25 orthogonal array, where composite material type (CM), drill geometry (D), cutting speed (Vc), and feed rate (f) were selected as control factors at five levels. Drill-point temperature (T) was monitored using an infrared thermal camera, while the delamination factor (DF) was quantified through two-dimensional area analysis of digital microscope images. Optimal parameters for each quality response were initially identified using the Taguchi approach, and a combined performance evaluation of T and DF was carried out via Gray Relational Analysis (GRA) to determine the overall best parameter setting. The influence of each factor was further assessed by ANOVA, revealing that CM and D were the dominant contributors to T, whereas D and f predominantly affected DF. Additionally, response surface-based mathematical models demonstrated high prediction accuracy for both responses (R² > 97%), and their reliability was confirmed through validation trials. SEM examinations provided microstructural insights into surface integrity and damage mechanisms, supporting the statistical findings. Overall, the results establish an effective parameter-selection and multi-objective optimization framework for improved damage control during the drilling of CNT-enhanced laminated composites. CNT reinforcement led to increased drill-point temperature during drilling. Drill geometry was identified as the most influential factor governing delamination damage. The lowest combined temperature and delamination level were achieved using the D1 tool geometry under a low feed rate. Gray Relational Analysis (GRA) revealed drill geometry as the dominant control factor for overall performance. Response Surface Methodology (RSM) models exhibited high prediction accuracy for both responses (R² > 97%).