Prediction of Flash Points of Organosilicon Compounds by Structure Group Contribution Approach

Abstract

Flash point (FP) is the primary property to evaluate fire hazards of a flammable liquid. In most countries regulations for safe handing, transporting, and storage of liquid chemicals mainly depend on the FPs of liquid chemicals. Due to the advancement of technology in discovery or synthesis of new compounds, FP data are desirable for related industries, but there is often a significant gap between the demand for such data and their availability. Thus, a reliable method to predict the FPs of flammable compounds seems very important in this regard. In the present work a predictive model of FP for organosilicon compounds is proposed via the structure group contribution (SGC) approach. This model is built up by using a training set of 184 organosilicon compounds with the fitting ability ( R 2 ) of 0.9330, the average error of 8.91 K, and the average error in percentage of 2.84%. The predictive capability of the proposed model has been demonstrated on a testing set of 46 organosilicon compounds with the predictive capability ( Q 2 ) of 0.8868, the average error of 11.15 K, and the average error in percentage of 3.66%. Because the known error for measuring FP by experiment is reported to be about 6–10 K, the proposed method offers a reasonable estimate of the FP for organosilicon compounds. Moreover, the proposed SGC model requires only the molecular structure of a compound to estimate its FP, so it also offers an effective way to approximate the FP of a novel chemical for which its quantity is still not readily available for measuring its FP by experiments.