Predictive modelling of strength in recycled filler-modified glass/epoxy composites exposed to marine environments using MLP_ANN

Glass fiber-reinforced epoxy (GFRP) composites are extensively used in marine structures owing to their high specific strength and corrosion resistance; however, prolonged seawater exposure can adversely affect their long-term mechanical performance. This study investigates the influence of thermo-chemically recycled milled carbon fiber (rmCF), recycled milled Kevlar fiber (rmKF), and hybrid recycled fillers (rmHF) on the seawater ageing behaviour of GFRP laminates and develops a multilayer perceptron artificial neural network (MLP-ANN) model for long-term prediction of water absorption and flexural strength. Composite laminates were immersed in seawater for up to 90 days, and water absorption, flexural performance, and micro structural changes were evaluated at 15-day intervals using flexural testing, Fourier transform infrared spectroscopy (FTIR), scanning electron microscopy (SEM), and energy-dispersive spectroscopy (EDS). The results show that seawater ageing progressively reduced flexural strength due to moisture diffusion-induced matrix plasticization, hydrolytic degradation, fiber–matrix interfacial weakening, and micro crack initiation. In contrast, the incorporation of recycled milled fillers reduced seawater ingress by increasing the tortuosity of moisture diffusion pathways and enhancing filler–matrix interfacial integrity, thereby improving the retention of mechanical properties. The developed MLP-ANN model accurately predicted water absorption, ultimate load, and flexural strength, achieving correlation coefficients (R²) between 0.972 and 0.997 with prediction errors within 2–5% of the experimental values. The trained model further provided deterministic long-term extrapolations up to the estimated time required to reach complete moisture saturation, corresponding to approximately 409 days (Base_GE), 1099 days (rmCF_GE), 576 days (rmKF_GE), and 930 days (rmHF_GE). These predictions extend beyond the experimentally validated 90-day exposure period and should therefore be interpreted within the scope of the present study. The findings demonstrate that recycled milled fillers provide a sustainable approach for enhancing the seawater durability of GFRP composites, while the proposed MLP-ANN model offers an efficient data-driven tool for long-term durability assessment and service-life prediction of composite structures operating in marine environments.