Generalized probabilistic models for performance assessment of bio-based insulation materials in sustainable constructions

The widespread adoption of sustainable, bio-based insulation is hindered by the inherent variability of its properties, making reliable performance prediction a significant challenge for engineers. To address this problem, this study develops a robust probabilistic framework to predict performance while explicitly accounting for this uncertainty. The methodology is based on a comprehensive literature review of 266 studies to establish a performance database for thermal conductivity, Noise Reduction Coefficient, and carbon footprint. Bayesian statistical methods were then applied to derive generalized predictive models from this data. The findings reveal that thermal conductivity correlates non-linearly with mass density, while Noise Reduction Coefficien is a nonlinear function of both density and thickness. The carbon footprint is characterized using material-specific uniform and normal distributions, confirming the net carbon sequestration potential of key bio-based options. The main conclusion is that performance can be reliably predicted when variability is statistically quantified. The key contribution of this work is a practical framework that enables designers to make informed material selections, accounting for performance variability and environmental impact. The novelty lies in the creation of generalized, data-driven models applicable across a wide range of bio-based materials, providing a valuable tool to streamline the design of sustainable buildings and support the transition to a low-carbon built environment.