Diagrams showing relationship of the temperature (T), An and △131一131 (OR σ)of sodic plagioclase were advanced in succession by Christie (1962), Smith(1972) and Zhang Yijun(1985). Later, the formulae for deformation temperature of sodic plagioclase were put forward by Zhang Yijun. Nevertheless, in view of the inconvenience of the diagram, the uncertainty of the dividing point in the formulae and the nonlinear relationship amongσ，An and T, we present a new method for calculating the temperature of the deformation-artificial neural network，which has accurate nonlinear projection.In this paper, an account is first given about the study of the deformation thermometer,followed by a description of the ANN (Artificial Neural Network) model. With this model, the temperature of the deformation is estimated. A comparison shows that the artificial neural network is very effective in the estimation of the deformation temperature.