Prediction and optimization of mechanical and moisture properties of jute–coir hybrid composites using a DBN–KNN–GWO model

Composite materials reinforced with natural fibers such as jute and coir have been attracting increased attention due to their environmental sustainability, affordability, and good mechanical properties. However, most current research has been limited to laboratory experimentation and traditional modeling methods, with little consideration for predictive models that can accommodate nonlinear interactions and optimize multiple objectives. The main objective of this study is to fabricate and characterize the jute–coir composite and precisely determine its mechanical behavior and moisture absorption (MA) properties. These composites were prepared with varying stacking patterns and then evaluated for tensile strength (TS), flexural strength (FS), impact strength (IS), and MA. Response surface methodology (RSM) is used to study the influence of factors such as fiber composition, fiber length, and alkali treatment concentration. The proposed hybrid model, DBN–KNN–GWO, uses Deep Belief Network (DBN) for feature extraction, k-Nearest Neighbor (KNN) for prediction, and Grey Wolf Optimizer (GWO) for hyperparameter optimization. Experimental analysis revealed that the laminate with jute dominance had the maximum TS (68 MPa) and FS (110 MPa). In contrast, the laminate with coir dominance had the maximum IS (13.5 kJ/m²) and the highest MA (11.2%). Multiobjective optimization revealed optimum parameters resulting in a desirability of 0.937. As evidenced by its performance relative to other models, the hybrid modeling approach developed here proved an effective means for designing and optimizing natural fiber composites. The study presents a new approach that combines deep learning, metaheuristics, and statistical design to predict multiple responses in composite materials.