Applying machine learning for predicting thermal conductivity coefficient of polymeric aerogels

Owing to nano-porous morphology, aerogels are classified as low thermal conductive materials, suitable for thermal insulation applications. Therefore, aerogels became an interesting topic for many studies, in order to predict and achieve desirable thermal properties. In the current study, three supervised machine learning algorithms were developed for the prediction of thermal properties. A reference learning dataset is provided by collecting information from previous experimental investigations on various aerogels. In this study, K-nearest neighbor (KNN), radius nearest neighbor (RNN), and Gaussian processes (GP) are applied as machine learning regression methods. For validating proposed machine learning-based models, thermal conductivities of polyurethane aerogel and silica–resorcinol formaldehyde aerogel are predicted and then compared to the real conductivity values. The output data showed acceptable results for predicting thermal conduction coefficient values and optimum parameters for thermal insulation applications.