Predictive modeling and mechanical characteristics optimization of sustainable hemp concrete using neural networks

Abstract

Numerical modeling, whether analytical or based on finite element methods, plays a fundamental role in the preliminary phases of scientific investigation, allowing researchers to eliminate costly and ineffective hypotheses. This study introduces the use of artificial intelligence, specifically artificial neural networks (ANNs), as a robust tool for addressing the increasing complexity of optimization tasks in material design. ANN models effectively capture nonlinear interactions among variables, offering significant advantages such as reduced experimentation time and cost, improved adaptability, and process flexibility. The proposed methodology focuses on predicting the thermal and mechanical behavior of hemp concrete under varying compositions. Using experimental design and regression analysis, the influence of input parameters, hydrated lime, Portland cement, water, and sodium silicate on thermal conductivity and compressive strength is modeled. Multi-criteria optimization is applied to identify configurations that meet distinct performance requirements, with a final solution presented that balances multiple objectives. This approach supports efficient material development with minimal reliance on extensive physical testing.