Prediction of energy mass loss rate for biodiesel fire via machine learning and its physical modeling of flame radiation evolution

Biodiesel is an emblematic energy of green power and developing vigorously, which is conducive to an important strategic significance covering the sustainable development of the global economy, the promotion of energy substitution, the reduction of environmental pressure, and the control of atmospheric contamination. In order to study the biodiesel energy burning characteristics, a series of experiments were carried out in an ISO9705 full-scale room, and the effects of pool size on the burning rate, flame height, flame oscillation frequency, and flame radiation fraction of biodiesel energy fire were systematically analyzed. This study uses genetic algorithm-back propagation neural network (GA-BPNN) algorithms for real-time prediction of transient fire mass loss rates. Three parameters (pool size, liquid depth, and burning time) are paired with the fuel mass loss rate and trained using a GA-BPNN algorithms model. The results show that the GA-BPNN algorithms predictions have a good correlation with the validated experimental values and the relative is less than 15%. Furthermore, the ratio of intermittent flame height and continuous flame height to their mean value is calculated respectively at about 1.72 and 0.58. The flame oscillation frequency decreases following the increase in oil pool size, the correlation can be calculated by f=0.3(D/g′)1/2. Finally, a new correlation χrad=0.2Q˙*−2/3 is proposed to predict the flame radiation fraction by increasing the flame viewing factor coefficient. The proposed correlation can be used to describe its evolution under different oil pool sizes, which can be essential to estimate the flame radiation impact on surroundings for such biodiesel energy fires.