Predicting Young’s Modulus of Linear Polyurethane and Polyurethane–Polyurea Elastomers: Bridging Length Scales with Physicochemical Modeling and Machine Learning

Abstract

Predicting the properties of complex polymeric materials based on monomer chemistry requires modeling physical interactions that bridge molecular, interchain, microstructure, and bulk length scales. For polyurethanes, a polymer class with global commercial and industrial significance, these multiscale challenges are intrinsic due to the thermodynamic incompatibility of the urethane and polyol-rich domains, resulting in heterogeneities from molecular to microstructural length scales. Machine learning can model patterns in data to establish a relationship between the monomer chemistry and bulk material properties, but this is made difficult by small data sets and a diverse set of monomers. Using a data set of 63 industrially relevant and complex elastomers, we demonstrate that accurate machine learning predictions are possible when monomer chemistry is used to estimate interactions at interchain length scales. Here, these features were used to accurately (r 2 = 0.91) predict the Young’s modulus of polyurethane and polyurethane–urea elastomers. Furthermore, by a query of the trained model for compositions that yield a target modulus within the range of accessible values, the capabilities of using this methodology as a design tool are demonstrated. The presented methodology could become increasingly useful in building models for materials with small data sets and may guide the interpretation of the underlying physicochemical forces.

Machine Learning-Assisted Identification of Copolymer Microstructures Based on Microscopic Images

“…The RF model can handle high-dimensional data and is insensitive to overfitting, while the low accuracy of the linear-regression method (Figure d) suggested that T g width and the extracted features do not follow a linear relationship. However, related studies , have shown that linear models are expected to use a data-driven approach to obtain semiempirical formulas based on weak equivalents, so we independently perform a nonlinear treatment for each feature (Figure e) and finally get the semiempirical formula: The formula corresponds to a substantial improvement in model predictions compared to direct linear predictions (Figure h). Notably, from the above formula, it can be unexpectedly found that the blur degree shows a weak negative correlation with the T g width, which is contrary to common sense.…”

“…Conceptually, this model pretraining method involves the transfer of parameters, also called transfer learning or fidelity learning . Previous research works have proved that the transfer learning method is feasible to be applied in the field of materials. − For the explanatory problem, nonlinear variation treatments can be used to obtain an analytic solution formula in traditional machine learning. , However, it is difficult to get an explicit analytical expression for the more black-box neural networks. Here, the feature visualization method will be used to investigate the contribution of each pixel in the image to the recognition results and present them intuitively.…”

“…ing. 39,40 However, it is difficult to get an explicit analytical expression for the more black-box neural networks. Here, the feature visualization method will be used to investigate the contribution of each pixel in the image to the recognition results and present them intuitively.…”

Catalyzed and Non‐catalyzed Synthetic Approaches to Obtain Isocyanate‐free Polyurethanes

“…Based on both experimental and computational results, multiscale forces were studied by high predictive-strength models and HML for PU and four types of properties chosen as the subject of modelling, such as thermal, rheological, mechanical and failure. [185] Chain junctions via covalent bonds developed by multifunctional monomers fix their network at the length of scale shorter than trivial entanglement or dispersion of hard segment to impose further restrictions on mobility of chain. [186] The model procedure assisted as the role of viscoelasticity for the normalization parameter in network fracture which occurred to be the focal point of any other models, Mooney-Rivlin model, its adaptations to describe hyper-viscoelastic responses with constitutive models.…”

Extrapolation performance improvement by quantum chemical calculations for machine-learning-based predictions of flow-synthesized binary copolymers

“…Materials exploration using ML has been investigated for a wide range of purposes, including searching for polymer compositions that exhibit required properties, [8][9][10] classifying polymers by their crystalline phase and microstructure, 11,12 and predicting the physical properties of polymers. 13 As summarized by Hu et al, the number of different algorithms and learning models is increasing at an accelerating rate and, with it, the scope of exploration. 14 ML can be conducted with a small amount of data; however, many data are preferable considering the scope of the material search and the desired prediction accuracy.…”