藉由機器學習方法探討固化劑結構對環氧樹脂系統玻璃轉移溫度的關鍵因素

Abstract

環氧樹脂因其卓越的黏著力、機械強度和耐熱性，已廣泛應用於電子、航空航太和土木工程等領域。為了滿足這些應用的需求，開發具有高玻璃轉移溫度（Tg）的環氧樹脂成為一個重要且迫切的課題。為了更有效地開發高Tg環氧樹脂的新型固化劑，本研究利用監督式機器學習方法來實現固化劑化學結構與基於DGEBA的環氧樹脂Tg之間的關聯（結構-性質關係）。在本研究中，我們首先訓練機器學習模型，通過輸入固化劑的化學結構來快速且準確地預測DGEBA-based環氧樹脂的Tg，接著透過分析該模型來確定固化劑結構中影響環氧樹脂Tg的關鍵因素。過去開發優良材料通常是基於經驗法則進行，因此總是需要大量的時間和金錢成本。我們期望本研究不僅能為實驗專家提供一種更明確的設計策略以開發耐高溫的環氧樹脂新型固化劑，還能提供一種材料開發的創新方法，進而促進其他領域的發展。

Epoxy resins have been widely used in several fields such as electronics, aerospace, and civil engineering because of their outstanding adhesion ability, mechanical strength, and heat resistance. To meet demands in these application, developing epoxy resins with high glass transition temperature (Tg) is one of the significant and urgent issue. To develop novel curing agents of high-Tged epoxy resins more efficiently, we utilize supervised machine-learning method to realize the correlation between chemical structures of curing agents and Tg of DGEBA-based epoxy resins (structure-property relationship). In this study, we firstly train a model which can rapidly and accurately predict Tg of DGEBA-based epoxy resins by inputting chemical structures of curing agents, then figure out critical factors in Tg of epoxy resins via analysis of our model. In the past, finding a useful material always spent lots of time and money because it was executed by empirical methods. We expect our research can provide not only a more distinct design strategy to experimental experts to develop novel curing agents of high-temperature resistant epoxy resins, but an innovative way, which can obviously decrease costs in experiments, to develop new materials in other fields.