A LTHOUGH the computerized prediction of phase diagrams by thermodynamic method has made great achievements, the lack of thermodynamic, data is one of the obstacles for unknown phase diagram prediction. Miedema has proposed to use atomic parameters for thermodynamic property estimation of alloy systems. Chen and co-workersL have used artificial neural network and atomic parameters to predict some thermodynamic properties, such as melting point, melting types. In this work, we presented a new method for phase diagram calculation by combining thermodynamics with atomic parameters and artificial neural network. In this method, the interaction parameters of the solutions were first computed from phase diagram of the binary systems. Then artificial neural network and atomic parameters were used to predict A LTHOUGH the computerized prediction of phase diagrams by thermodynamic method has made great achievements, the lack of thermodynamic, data is one of the obstacles for unknown phase diagram prediction. Miedema has proposed to atomic parameters for thermodynamic property estimation of alloy systems. Chen and co-workersL have used artificial neural network and atomic parameters to predict some thermodynamic properties, such as melting point, melting types. In this work, we presented a new method for phase diagram calculation by combining thermodynamics with atomic parameters and artificial neural network. In this method, the interaction parameters of the solutions were first computed from phase diagram of the binary systems. Then artificial neural network and atomic parameters were used to predict