Machine learning for mechanics prediction of 2D MXene-based aerogels

Abstract

Hybrid aerogels of two-dimensional (2D) transition metal carbide (MXene) and nanocellulose show huge potential in a wide range of applications owing to their unique compressive mechanical properties. However, the compressive mechanical properties of hybrid aerogels are sensitive to the physical parameters of its building blocks, which are difficult to be optimized by high throughput experiments. Considering the inherent complex variables of MXene/nanocellulose aerogels, this work realizes the prediction of their mechanical properties by machine learning (ML). Based on the reported 34 sets of data on Ti3C2 MXene, we trained three ML algorithms: artificial neural network (ANN), support vector machine (SVM) and random forest (RF). Results indicate that the ANN outperforms other algorithms as it fits various nonlinear input features well. The relative content of Ti3C2 is the most effective factor in the compressive strength of hybrid aerogel. The mechanical properties of the 540 input possibilities are predicted by the outperforming ANN model, and quantitative structural adjustment is obtained for a maximum compression modulus of 29 kPa. This work provides guideline for the mechanical property prediction of composite materials using ML.