Prediction of the maximum temperature of sulfur-containing oil using Gaussian process regression for hazards prevention

An oxidation self-heating process of sulfurized rust usually results in a fire or an explosion in crude oil tanks due to the oil’s maximum temperature (T_max ) exceeding the critical temperature at which the fire and explosion happens. Some previous studies have shown that T_max is determined by the five main factors including water content, mass of sulfurized rust, operating temperature, air flow rate, and oxygen concentration in the safety valve. In this paper, based on a collected dataset about the five factors and T_max , the Gaussian process regression (GPR) method is adopted to build a nonlinear model describing the relationship between T_max and the five factors, and the new model is then used to predict ^T _max of other similar processes by inputting the data corresponding to the five factors. The results show that the GPR model can reach the prediction accuracy and the prediction result by the GPR model is more accurate than that by the model of Support Vector Machine (SVM). This indicates that the GPR method can be applied to predict T_max of the oxidation self-heating process of sulfurized rust. The prediction of ^T _max using the GPR model is of great significance to industrial risk control and accident prevention of sulfur-containing oil in production and transportation.