易燃液体闪点预测模型综述

A review of flash point prediction models for flammable liquids

闪点(FP)是易燃液体及其分类标准的重要划分依据,同时也是衡量可燃液体火灾危险性的重要参数。闪点的确定将影响危险化学品的分类、储存、运输、使用、防火及危险品公示等各方面。为弥补实验测定的不足,借助模型预测来计算闪点具有重要的理论意义和实用价值。本文综述了易燃液体闪点的估算方法,主要分为三类:经验关联计算,基团贡献法计算和基于分子结构的模型预测,并讨论了三类方法各自的优势和不足。经验关联计算形式上简单,并且易于从实验数据中构建,一般与沸点相关联,使用数学回归或人工神经网络(ANN)方法获得。基团贡献法(GCM)是假设分子的性质是构成分子的所有基团贡献的函数,通过分子官能团贡献对闪点建立线性或非线性模型。基于分子结构的定量结构-性质关系,(QSPR)模型的建立与精度关键在于分子描述符的计算与筛选、模型建立的不同方法。近年来,鉴于各模型的优势与不足,将QSPR与其他预测模型和先进技术结合起来研究闪点与分子结构的相关性,是闪点预测的研究方向和热点,也为易燃液体混合物闪点的预测模型打下基础。

Flash point is an important basis for classification of flammable liquids and their classification criteria, and is also an important parameter for measuring the fire risk of flammable liquids. The determination of flash point will affect the classification, storage, transportation, use, fire prevention and publicity of dangerous goods of dangerous chemicals. To make up for the lack of experimental determination, it is of great theoretical and practical value to calculate the flash point by means of model prediction. This paper reviews the methods for estimating the flash points of pure flammable liquids, which are divided into three categories: empirical correlation model, group contribution model and molecular structure-based model. The advantages and disadvantages of the three models are discussed. Empirical correlation models are formally simple and easy to construct from experimental data, generally associated with boiling points, using mathematical regression or artificial neural network (ANN) approach. The group contribution model (GCM) is the assumption that the nature of the molecule is a function of the contribution of all groups that make up the molecule, and a linear or non-linear model of the flash point is established by contributions of molecular functional groups. The key to the establishment and accuracy of quantitative structure-property relationship (QSPR) model based on molecular structure lies in the calculation and screening of molecular descriptors and the different methods of model establishment. In recent years, in view of the advantages and disadvantages of each model, combining QSPR with other prediction models and advanced technologies to study the correlation between the flash point and the molecular structure is the research direction and focus of the pure component flash point prediction. It also lays the foundation for the prediction model of the flash point of the flammable liquid mixtures.