Environmental risks are inevitable in the operation of any complicated chemical process industry. In the chemical industry, a toxic and harmful gas leak could easily lead to accidents. Many leaking gases retain their flammable and explosive properties. The indoor release of a dense gas is a primary concern. Therefore, this paper presents a method for establishing a prediction model of leakage gas concentration. The general process of indoor dense gas leakages and dispersion was analyzed using the pi theorem. Then, we obtained the gas leakage diffusion process model by the similarity criterion equation. An experiment was conducted on the continuous indoor leakage and diffusion of carbon dioxide gas. The experimental apparatus is primarily composed of an indoor space, a gas discharge system and a gas concentration data acquisition system. The experimental results are used to solve the similarity equation. Quantitative analysis of the indoor carbon dioxide diffusion process is conducted, and the corresponding mathematical expressions are derived. Through this analysis, information can be obtained on the leakage concentrations in various locations, which facilitates a better understanding of the distribution of large gas leak concentrations in an indoor space. This research provides a reliable forecasting method for the trend of dense gas diffusion during civil and industrial leakage accidents and could be conducive to building a pertinent emergency rescue plan for effectively mitigating the damage due to accidents. (C) 2017 Elsevier B.V. All rights reserved.