Direct contact membrane distillation (DCMD) is a nonisothermal technology applied for the separation of nonvolatile components from aqueous solutions. Nowadays, a huge number of publications are dedicated to modeling of DCMD, however all of the presented models have at least one of the following disadvantages: oversimplification, the use of empirical heat and mass transfer coefficients and poor prediction for cases which are out of the experimental data range. To overcome these drawbacks, a multipurpose general predictive model of DCMD has been developed. The proposed model is suitable for hollow fiber and flat sheet configurations with or without spacers. For each compartment of the DCMD process, our model describes the momentum, mass and heat balances by systems of ordinary differential, partial differential and algebraic equations. The performance of the model has been analyzed in terms of the operating parameters (concentration of a feed solution, feed flow rate and feed temperature) and membrane thickness and length. The simulated results were in very good agreement with experimental and literature data. The broad parametric study demonstrates the great potential of application of the proposed model not only in the process optimization but in design of DCMD modules.