Prediction of Water Uptake Percentage of Nanoclay-Modified Glass Fiber/Epoxy Composites Using Artificial Neural Network Modelling

This research explores the water uptake behavior of glass fiber/epoxy composites filled with nanoclay and establishes an Artificial Neural Network (ANN) to predict water uptake percentage from experimental parameters. Composite laminates are fabricated with varying glass fiber $(40-60\text{wt}.\mathrm{\%})$ and nanoclay $(0-4\text{wt}.\mathrm{\%})$ contents. Water absorption is evaluated for 70 days of immersion following ASTM D570-98 standards. The inclusion of nanoclay reduces water uptake by creating a tortuous path for moisture diffusion due to its high aspect ratio and platelet morphology, thereby enhancing the composite’s barrier properties. The ANN model is developed with a 3–4–1 feedforward structure and learned through the Levenberg–Marquardt algorithm with soaking time (7 to 70 days), fiber content $(40,50,\text{and}60\text{wt}.\mathrm{\%})$ and nanoclay content $(0,2,\text{and}4\text{wt}.\mathrm{\%})$ as input parameters. The model’s output is the water uptake percentage. The model has high prediction efficiency, with a correlation coefficient $(R)$ of $0.998$ and a mean squared error of $1.38\times {10}^{-4}$. Experimental and predicted values are in excellent agreement, ensuring the reliability of the ANN for the simulation of nonlinear water absorption behavior. The results identify the synergistic capability of nanoclay and fiber concentration to reduce water absorption and prove the feasibility of ANN as a substitute for time-consuming testing in composite durability estimation.