Vacuum chambers can effectively suppress gas explosion without relying on any explosion suppression material. The effect of explosion suppression by a vacuum chamber is correlated with the negative pressure of the vacuum chamber and flame front position when the diaphragm breaks. Accurate control of the experimental conditions of explosion suppression is challenging, and the use of experimental methods alone for analyzing their interrelation is difficult. This study analyzed considerable experimental data on explosion suppression by a vacuum chamber, investigated factors that influenced the explosion suppression, and evaluated parameters of the explosion suppression by using the Delta P-I principle. This study established a BP neural network optimized by a genetic algorithm, and discovered the non-linear relations among all parameters. The trained neural network was used to analyze the influence of vacuum degree and flame front position on explosion suppression when the diaphragm breaks. The results demonstrated that a good explosion suppression was achieved when a vacuum degree of P-0 was >0.06 MPa and the optimal position of the flame front index S was 0 m. Explosion suppression of the vacuum chamber can be predicted by a fitting curve I = 5.61e ((-P0/0.028)) + 1.70, which was obtained using the trained neural network and S = 0.86 m. (C) 2016 Elsevier Ltd. All rights reserved.