Analysis on the accuracy of laboratory measurement of apparent diffusioncoefficient of adsorbed gas in matrix based on the competition relationship between gas flow and gas diffusion

Due to high speed free gas from initial desorption,the desorption curve shape may experience a sharp variation with different loss time,which brings some barriers to the laboratory measurement of gas apparent adsorption diffusion coefficient in matrix and the accurate measurement of K1 value and loss gas content in engineering practice. Based on the competition relationship between gas flow and diffusion,the whole concentration apparent diffusion model considering the free gas flow influence was established,and the contribution of gas flow and gas diffusion in initial desorption was obtained. Meanwhile,the attenuation model of K1 value as the loss time goes was established. A self developed gas desorption automatic test apparatus was used to the initial gas desorption curve,and the apparent gas diffusion coefficients with loss times of 1 min,5 min and 10 min were obtained. The results show that the whole desorption can be divided into three stages as the desorption goes,which are free gas flow dominant stage,transition stage and adsorbed gas diffusion dominant stage. The free gas volume in fractures and the desorption time obey a linear relationship,while the adsorbed gas in matrix and the desorption time obey the unipore diffusion model,and at a short time,they obey the model. With the increase of loss time,the contribution of free gas in the whole desorption becomes smaller,while that of adsorbed gas becomes larger. The linear stage of the desorption curve shortens and the curvature of desorption curve increases at the macro level. The deduced apparent diffusion coefficient and K1 value become smaller and the decreasing speed is large in the initial stage. With the attenuation model of gas apparent diffusion coefficient,the effective measuring time zones of K1 value and loss content were obtained. Besides,a new model to deduce loss content with higher accuracy was given. Compared to the original model,the fitting degree of the experimental data is improved by 20.8%.