ML Prediction of Nano-Porous Aerogels Material Properties

Due to the high prediction performance and low computational cost, machine learning (ML) based tools have been widely implemented in the field of material science for materials innovation, material design, deriving insights, and material properties prediction. In this context, the application of ML in the field of porous materials, particularly nanostructured aerogels, only rose to prominence recently. Despite many attempts to tailor and optimize the performance of nanostructured aerogels, the current methodologies mostly rely on experimental approaches, and thus further development of novel aerogels with optimum performance is greatly restricted. Considering the prolonged synthesis process, costly materials and equipment, and the necessity of analyzing microstructure and properties of aerogels, developing a ML platform can facilitate data-driven materials innovation and accelerate prediction of the material properties of nanostructure aerogels. Therefore, in this study, we report the application of artificial neural network (ANN) as an effective ML tool for properties prediction of nanostructured organic aerogels. Through optimizing the ANN network parameters, ANN architecture is constructed. Data mining is performed, and selected descriptors are chosen as the input for the model. Various properties, namely compressive modulus, density, and porosity of nanostructured aerogels, are predicted, and ANN performance is analyzed via evaluation metrics.