期刊
CURRENT OPINION IN GREEN AND SUSTAINABLE CHEMISTRY
卷 41, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.cogsc.2023.100801
关键词
Polymer formulation; Machine learning; Molecular dynamics; Experimental design
Chemical recycling of polymers is gaining momentum as a circular technology. However, the high energy demand and sensitivity of complex formulations hinder efficient recycling. To address this, a hybrid model-based framework is proposed, combining machine learning, structure-property models, and experimental data. This approach aims to substitute virgin feedstocks with complex recyclates and improve the efficiency of polymer recycling.
Chemical recycling of polymers is taking off as a circular technology, typically targeting pure recyclates. However, this is often not achieved efficiently due to high energy demand of separation and purification steps. In addition, many polymer applications have complex formulations that may be sensitive to impure feedstocks. Substitution of virgin feedstocks by complex recyclates (often containing impurities) requires a good knowledge of the structure/composition-property re-lations of polymer formulations. As this is often not the case, current practice relies on costly and rather inefficient enumeration experiments, or, at best, classical design-of-experiments approaches. We review the state-of-the art in structure-property modeling, present an example for poly-urethane formulations, and propose a hybrid model-based framework. This involves a machine learning workflow for substitution problems in complex polymer formulations, combining existing data, novel reaction kinetics, structure-property models, molecular dynamics, and a mini-mum of experimental-analytical data where necessary, to build and validate the model.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据