An optimization model of matching parts assembly based on process capability index was constructed to improve the precision of selective assembly accordingly by upgrading the process capability so as to make the assembly process be flexible, reduce the ratio of waste product, boost the productivity and quality of assembly and create more benefit for increment of product. Based on reducing the cost of matching assembly process, by way of process of Markov chain in the random condition, the tolerance index of matching assembly and others was described, the multi-object restriction primary model of maximal process capability index and minimal cost of matching assembly process among the precision of the assembly, tolerance and the process capability index were constructed. By using the sequential unconstrained minimization technique, this restriction model of matching assembly based on dynamic multi-object programming has been solved, and the model has been validated through a practical example.