X-ray variable energy imaging can obtain the sectional information of complicated structural component successively, and get the whole information by multi-spectrum fusion. Now the energy parameters of X ray imaging mainly depend on man-made setting with the certain step voltage. However this modulation doesn’t match to the attenuation thickness variation of the object. Therefore, this paper proposes an optimum tube voltage prediction algorithm based on variable energy imaging. It extracts the effective thickness (ET) and near the effective thickness (NET) in the image sequences which are acquired by pre-scanning the detected object. Then it establishes a physical model between image gray, tube voltage and X ray spectrum. And the model of voltage and gray difference between the ET (high quality area) and NET (prediction area) is also established. On the basis of these two models, the optimal imaging energy forecasting model of NET is modeled. Then, solve the model and get the optimal voltage for NET. At last, by the experiment of the steel blocks with different thickness, testify this prediction algorithm. The results compared with the actual values showed that the prediction algorithm can accurately predict 3 or 4 mm at low voltage and 7 or 10 mm at high voltage. Prediction accuracy is over 95%.