Today inorganic membranes attract a lot of interest as a growing field. Main focus of those activities is on the development of membrane materials, which can offer high permselectivities with acceptable high permeances. The need for high permselectivity beyond those limited by Knudsen flow requires the estimation of the factors, which determine the permselectivity. Plausible theoretical models based on physical or chemical reasoning is desirable to guide systematic development efforts for designing next generation inorganic membranes. Here we reviewed our attempts to generate theoretical models based on the molecular dynamics method for this purpose. As a first attempt, simulation was performed at specific conditions where the Knudsen theory can be applied and can be reproduced well by our simulation methodology. Molecular dynamics simulation at 373 K of the permeation of iso- and n-butanes through ZSM-5 type silicalite membrane are presented. After 200 ps of simulation time the permeation of n-butane was observed whereas the permeation of iso-butane was not observed. The calculated permeability of n-butane, which is close to experimental data, is also presented. A study on the affinity membrane for the separation of CO2 at high temperature is presented and the prospect of permselectivity of CO2 is demonstrated.