Based on the definition for form error and theory of mathematics programming, a non-linear programming model was built for form error evaluation. This non-linear model was further proved, in essence, to be a multi-target optimization problem, and could be transformed into a single-target optimization problem. A unified and efficient successive quadratic programming method (SQP algorithm) was proposed to solve these models. As these programming models were convex, and SQP algorithm could retain the non-linear information of the models during the process of calculation. So it has very loose requirement for initial parameters. The method is also stable, reliable and high-efficient. The results of several experiments of form error evaluation showed that the method fitted the requirements for convex programming's global optimization very well which proved the above-mentioned conclusion. 