Development of a Digital Image Processing- and Machine Learning-Based Approach to Predict the Morphology and Thermal Properties of Polyurethane Foams

Polyurethane foams are frequently used to provide thermal insulation. Thanks to the blowing agents used during their synthesis, pores are created in the structure and thermal insulation is achieved through these pores. In this study, five different insulating polyurethane foam samples containing water and cyclohexane blowing agents were synthesized. Pore stabilities and their effects on pore neighboring were analyzed computationally (MP2/aug-cc-pVDZ). A digital image processing- and machine learning-based algorithm was developed to predict the mean neighboring effect distances of the produced foams. It was created using the Voronoi tessellation method used for the identification problems in industrial applications. This method showed that there was a close relationship between the calculated Voronoi neighboring effect distances of the samples and their thermal conductivity coefficients. Considering the Voronoi neighboring effect distances proposed in this study, the thermal conductivity coefficient of similar polyurethane foams could be predicted. This method required only a standard mobile phone to capture images of the samples and the algorithm developed using Python (version 3.13.2) programming language. In addition, when compared to the local surface imaging device SEM, it allowed the entire surface to be analyzed faster and at once, without any surface deterioration.