Artificial neural network or pattern recognition with chemical bond parameters as imputs or features is applied to reciprocal ternary systems. After training with the known data, the mathematical models obtained can be used to make computerized prediction of the formability, the melting point of intermediate compound. Predicted results indicate that the Cs^+, Mn^2^+|Cl^-, SO4^2^- system can form a ternary compound and the melting point of CsCl. MnSO4 is 677℃. The new compound is confirmed by X ray diffraction and the predicted value of CsCl. MnSO4 melting point is in agreement with the experimental one.

Key words: X-RAY DIFFRACTION ANALYSIS, CHEMICAL BONDS, CESIUM CHLORIDE, PATTERN RECOGNITION, PHASE DIAGRAM, MANGANESE SULFATE, NEURAL NETWORK