Yu-Zhong Wang*
The Collaborative Innovation Center for Eco-Friendly and Fire-Safety Polymeric Materials (MoE), National Engineering Laboratory of Eco-Friendly Polymeric Materials (Sichuan), State Key Laboratory of Polymer Materials Engineering, College of Chemistry, Sichuan University, Chengdu, 610064, China.

EXTENDED ABSTRACT: The flame retardation of polymeric materials plays an important role in preventing the occurrence of fire, and the flame-retardant polymeric materials have urgent application needs in many fields. Although the quantitative relationship between material structures has been clarified for specific flame-retardant material systems, there is still no universal principle that reveals the quantitative contribution of specific group structures to the flame-retardant properties of polymer materials. In response to this recognized basic scientific problem, we propose a genetic engineering strategy for flame retardant materials that combines experimental big data with machine learning algorithms. Unlike the inherent properties of polymers, which are determined by their own structures, the flame-retardant properties of polymers are more closely related to the flame-retardant groups involved in the complex burning processes. We first create the polymer burning operando analyzer to research the burning process of the polymers, further obtain the real combustion data of massive polymers to form a flame-retardant polymer material database; combine with the multi-level structure information of the polymer and the flame-retardant group characteristic information, the analysis and descriptors that are strongly related to flame retardant properties; further use machine learning algorithms to establish models that can accurately predict flame retardant properties from polymer structures. This work enriches the scientific theory of flame retardance and lays a scientific foundation towards the high-throughput design.

Keywords：Flame retardant materials; Real burning; Motif groups; Polymer database