Machine Learning-Assisted Design of Advanced Polymeric Materials

Accounts of Materials Research, Год журнала: 2024, Номер 5(5), С. 571 - 584

Опубликована: Апрель 16, 2024

ConspectusPolymeric material research is encountering a new paradigm driven by machine learning (ML) and big data. The ML-assisted design has proven to be successful approach for designing novel high-performance polymeric materials. This goal mainly achieved through the following procedure: structure representation database construction, establishment of ML-based property prediction model, virtual high-throughput screening. key this lies in training ML models that delineate structure–property relationships based on available polymer data (e.g., structure, component, data), enabling screening promising polymers satisfy targeted requirements. However, relative scarcity high-quality complex multiscale pose challenges method, such as modeling challenges.In Account, we summarize state-of-the-art advancements concerning Regarding digital representations are predominant methods cheminformatics along with some newly developed integrate characteristics. When establishing choosing optimizing attain high-precision predictions across vast chemical space. Advanced algorithms, transfer multitask learning, have been utilized address challenges. During process, defining combining genes, candidates generated, subsequently, their properties predicted screened using models. Finally, identified verified computer simulations experiments.We provide an overview our recent efforts toward developing approaches discovering advanced materials emphasize intricate nature structural design. To well describe structures polymers, methods, fingerprint cross-linking descriptors, were developed. Moreover, multifidelity method was proposed leverage multisource isomerous from experiments simulations. Additionally, graph neural networks Bayesian optimization applied predicting compositions.Finally, identify current point out development directions emerging field. It highly desirable establish materials, particularly when constructing large language. Through seek stimulate further interest foster active collaborations realizing innovation

Язык: Английский

---

Machine Learning Approaches in Polymer Science: Progress and Fundamental for a New Paradigm

SmartMat, Год журнала: 2025, Номер 6(1)

Опубликована: Янв. 9, 2025

ABSTRACT Machine learning (ML), material genome, and big data approaches are highly overlapped in their strategies, algorithms, models. They can target various definitions, distributions, correlations of concerned physical parameters given polymer systems, have expanding applications as a new paradigm indispensable to conventional ones. Their inherent advantages building quantitative multivariate largely enhanced the capability scientific understanding discoveries, thus facilitating mechanism exploration, prediction, high‐throughput screening, optimization, rational inverse designs. This article summarizes representative progress recent two decades focusing on design, preparation, application, sustainable development materials based exploration key composition–process–structure–property–performance relationship. The integration both data‐driven insights through ML deepen fundamental discover novel is categorically presented. Despite construction application robust models, strategies algorithms deal with variant tasks science still rapid growth. challenges prospects then We believe that innovation will thrive along approaches, from efficient design applications.

Язык: Английский

---

Polymer Informatics at Scale with Multitask Graph Neural Networks

Chemistry of Materials, Год журнала: 2023, Номер 35(4), С. 1560 - 1567

Опубликована: Фев. 15, 2023

Artificial intelligence-based methods are becoming increasingly effective at screening libraries of polymers down to a selection that is manageable for experimental inquiry. The vast majority presently adopted approaches polymer rely on handcrafted chemostructural features extracted from repeat units-a burdensome task as libraries, which approximate the chemical search space, progressively grow over time. Here, we demonstrate directly "machine learning" important unit cheap and viable alternative extracting expensive by hand. Our approach-based graph neural networks, multitask learning, other advanced deep learning techniques-speeds up feature extraction 1-2 orders magnitude relative without compromising model accuracy variety property prediction tasks. We anticipate our approach, unlocks truly massive scale, will enable more sophisticated large scale technologies in field informatics.

Язык: Английский

---

Estimation and Prediction of the Polymers’ Physical Characteristics Using the Machine Learning Models

Polymers, Год журнала: 2023, Номер 16(1), С. 115 - 115

Опубликована: Дек. 29, 2023

This article investigates the utility of machine learning (ML) methods for predicting and analyzing diverse physical characteristics polymers. Leveraging a rich dataset polymers' characteristics, study encompasses an extensive range polymer properties, spanning compressive tensile strength to thermal electrical behaviors. Using various regression like Ensemble, Tree-based, Regularization, Distance-based, research undergoes thorough evaluation using most common quality metrics. As result series experimental studies on selection effective model parameters, those that provide high-quality solution stated problem were found. The best results achieved by Random Forest with highest R2 scores 0.71, 0.73, 0.88 glass transition, decomposition, melting temperatures, respectively. outcomes are intricately compared, providing valuable insights into efficiency distinct ML approaches in properties. Unknown values each characteristic predicted, method validation was performed training predicted values, comparing specified variance characteristic. not only advances our comprehension physics but also contributes informed optimization materials science applications.

Язык: Английский

---

Heat-Resistant Polymer Discovery by Utilizing Interpretable Graph Neural Network with Small Data

Macromolecules, Год журнала: 2024, Номер 57(8), С. 3515 - 3528

Опубликована: Апрель 12, 2024

Polymers with exceptional heat resistance are critically valuable in numerous domains, particularly as essential components of flexible organic light-emitting diodes. Among these, polyimides (PIs) demonstrate significant potential substrate candidates for these next-generation displays due to their robustness. However, traditional Edisonian approaches struggle navigate the vast chemical space PIs and also pose challenges small data, which constrains learnable machine learning (ML). In this study, we propose a chemical-knowledge-based strategy facilitate design high glass transition temperature (Tg) utilizing an atom-wise graph neural network data. Inspired by intuition, our leverages available data on same property (i.e., Tg) from other polymers, is beneficial expanding used ML. The trained ML model achieves impressive performance predicting Tg polymers. We have investigated impact encompassed sets models. Through interpretability analysis, it has been demonstrated that learned more accurate knowledge. Utilizing model, 89 were rapidly discovered over 106 candidates, experimental validation confirmed most promising PIs, found possess exceeding 405 °C even 450 °C. These results, along accelerate discovery polymer materials display devices.

Язык: Английский

---

Machine learning and molecular design algorithm assisted discovery of gas separation membranes exceeding the CO2/CH4 and CO2/N2 upper bounds

Chemical Engineering Science, Год журнала: 2024, Номер 291, С. 119952 - 119952

Опубликована: Март 1, 2024

Язык: Английский

---

Predicting polymer solubility from phase diagrams to compatibility: a perspective on challenges and opportunities

Soft Matter, Год журнала: 2024, Номер 20(29), С. 5652 - 5669

Опубликована: Янв. 1, 2024

Advances in physical models and data science are improving predictions of polymer–solvent phase behavior we discuss the different approaches taken today remaining barriers to making broadly useful predictions.

Язык: Английский

---

PolyNC: a natural and chemical language model for the prediction of unified polymer properties

Chemical Science, Год журнала: 2023, Номер 15(2), С. 534 - 544

Опубликована: Дек. 6, 2023

PolyNC directly infers properties based on human prompts and polymer structures, enabling an end-to-end learning that encourages the model to autonomously acquire fundamental knowledge, in a multi-task, multi-type unified manner.

Язык: Английский

---

Machine learning-guided discovery of polymer membranes for CO2 separation with genetic algorithm

Journal of Membrane Science, Год журнала: 2024, Номер 712, С. 123169 - 123169

Опубликована: Авг. 8, 2024

Язык: Английский

---

Machine learning-assisted development of gas separation membranes: A review

Carbon Capture Science & Technology, Год журнала: 2025, Номер 14, С. 100374 - 100374

Опубликована: Янв. 30, 2025

Язык: Английский

---