Abstract: This paper analyzes the factors that influence the measure of the metal linear coefficient of expansion in light lever experiment and finds that calibration is nonlinear variation with temperature changing. Artificial neural network （ANN） has more abilities to learning and highly nonlinear reflecting than other traditional method. It is used to error compensation and analysis for experiment, and the simulation result obtained shows that the method is feasible and the calculation is precise.

Key words: metal linear expansion coefficient, artificial neural network （ANN）, error compensation