Limited to the high cost, low quantity and few species for the measurement of infrared spectrum of pollutant gas cloud, it is extremely important to simulate infrared spectrum with which to improve spectral identification algorithms. Traditionally, empirical probabilistic model and empirical or semi-empirical parameters were used to simulate the real-time explosion of pollutant gas cloud and furthermore simulate the infrared spectrum. A physical-based model was established, which compensated for the low accuracy of the empirical or semi-empirical models, to finely simulated the explosion of pollutant gas cloud. The mechanism of the physical-based explosion model and the way of infrared spectrum simulation based on the physical-based explosion model were studied. Finally, a comparison between the simulated infrared spectrum based on physical model, simulated infrared spectrum based on probabilistic model and the practical one was made and it got a more accurate result that improved by 14 percent at most as far as spectral residual is concerned and figured out the application scope of two models. The physical-based explosion model and the way of infrared spectrum simulation established are significant to finely simulate infrared spectrum of pollutant gas cloud and study high-quality spectral identification algorithms.