From Drug Molecules to Thermoset Shape Memory Polymers: A Machine Learning Approach

Abstract

Ultraviolet (UV)-curable thermoset shape memory polymers (TSMPs) with high recovery stress but mild glass transition temperature (T g ) are highly desired for 3D/4D printing lightweight load-bearing structures and devices. However, a bottleneck is that high recovery stress usually means high T g . For a few TSMPs with high recovery stress, their T g values are close to the decomposition temperature, and thus, the shape memory effect cannot be triggered safely and effectively. While machine learning (ML) has served as a useful tool to discover new materials and drugs, the grand challenge of using ML to discover new TSMPs persists in the very limited data available. Here, we report an enhanced ML approach by combining the transfer learning− variational autoencoder with a weighted-vector combination method. By learning a large data set with drug molecules in a pretraining process, we were able to effectively map the TSMPs to a hidden space that is much closer to a Gaussian distribution. Through this approach, we created a large compositional space and were able to discover five new types of UV-curable TSMPs with desired properties, one of which was validated by the experiments. Our contribution includes (1) representing the features of TSMPs by drug molecules to overcome the barrier of a limited training data set and (2) developing a ML framework that is able to overcome the barrier of mapping the molar ratio information. It is shown that this approach can effectively learn TSMP features by utilizing the relatedness between the data-scarce (and biased) TSMP target and data-abundant drug source, and the result is much more accurate and more robust than the benchmark set by the support vector machine method using direct label encoding and Morgan encoding. Therefore, it is believed that this framework is a state-of-the-art study in the TSMP field. This study opens new opportunities for discovering not only new TSMPs but also other thermoset polymers.

Insight in thermomechanical constitutive modeling of shape memory polymers

“…In the future, it is expected that the three disadvantages can be resolved with a further understanding of physics. Additionally, it should be mentioned that recent machine learning models (Yan et al, 2021a;Yan et al, 2021b) indicate that the behaviors for SMP (such as recovery stress, moduli, etc.) can be completely determined once we know the chemical structures of the SMP monomers.…”

“…Frontiers in Mechanical Engineering frontiersin.org polymers (Wu et al, 2019). Yan et al (2021aYan et al ( , 2021b) also used machine learning and discovered new thermoset SMPs, which have been validated either by molecular dynamics simulation or by experimental testing (Wick et al, 2021). Shafe et al (2022) studied the effect of atomistic fingerprints on thermomechanical properties of epoxy-diamine thermoset shape memory polymers, which facilitates machine learning discovery of TSMPs.…”

Modeling glass transition temperatures of epoxy systems: a machine learning study

“…The results are promising, since the model's features were purely theoretical, solely based on the chemical structures of the molecules. The performance in terms of MAPE (9.38 %) was slightly better than relatively complex models published elsewhere [26,27].…”

“…The performance of their consensus model (homo-and heteropolymers) was good (R 2 = 0.848), while the thermoset-only model gave poorer results (R 2 = 0.687). Yan et al [26,27] developed the state-of-theart model currently available for predicting T g of thermoset shape memory polymers (TSMPs), whose mean absolute percentage error for the test set was 13.91 %. The model is based on a variational autoencoder trained with drug molecules and adapted to TSMPs with transfer learning, combined with a weighted vector combination method and a convolutional neural network as regressor.…”

A Thermoset Shape Memory Polymer-Based Syntactic Foam with Flame Retardancy and 3D Printability

“…The shape memory effect tests were performed following the procedure reported by our group. − In the test, the cylindrical sample with an initial height h 0 was placed in between the MTS clamps and heated to 275 °C for 30 min. Once the thermal equilibrium was achieved, the sample was compressed to a certain height ( h 1 ) at a compression rate of 0.5 mm/min, and this strain was maintained for 10 min; the sample was then rapidly cooled to room temperature by spraying water with a wash bottle.…”