This study focuses on the development, characterization, and numerical simulation of novel composite materials based on natural vegetable fibers for applications in civil engineering. Three different bio-based composites were formulated using Alfa plant fibers, wood waste, and an equal mixture of both (50% Alfa, 50% wood), with polyvinyl acetate (PVAc), a non-polluting polymer matrix, as the binder. The performance of these composites is strongly influenced by the fiber morphology, structural characteristics, and the nature of the matrix. Thus, understanding and optimizing these parameters is crucial for tailoring materials to meet specific design requirements. The experimental approach began with the morphological and structural characterization of the raw and treated fibers, followed by the evaluation of the thermal a properties of the resulting composites. Furthermore, thermal conductivity simulations were performed using COMSOL Multiphysics to validate the experimental results and gain deeper insights into heat transfer behavior within the composites. A comparative analysis with conventional synthetic insulation materials revealed that the developed bio-composites offer competitive thermal performance while being more environmentally sustainable. These findings highlight the potential of Alfa and wood waste fibers as effective, eco-friendly alternatives for thermal insulation in building applications.

Optimizing concrete production requires balancing ingredient ratios and using local resources to produce an economical material with the desired consistency, strength, and durability. Compressive strength is crucial for structural design, yet predicting it accurately is challenging due to the complex interplay of various factors, including component types, water–cement ratio, and curing time. This study employs a Multi-layer Perceptron Neural Network (ANN_MLP) to model the relationship between input variables and the compressive strength of normal and high-performance concrete. A dataset of 1030 samples from the literature was used for training and evaluation. The optimized ANN_MLP configuration included 16 neurons in a single hidden layer, with the ‘tanh’ activation function and ‘sgd’ solver. It achieved an R² of 0.892, an MAE of 3.648 MPa, and an RMSE of 5.13 MPa. The model was optimized using a univariate sensitivity analysis to measure the impact of each hyperparameter on performance and select optimal values to maximize the accuracy and robustness.