Rapid prototyping (RP) can be used to make complex shapes with very little or even no constraint on the form of the parts. New design methods are needed for parts that can take advantage of the unique capabilities of RP. Although current synthesis methods can successfully solve simple design problems, practical applications with thousands to millions elements are prohibitive to generate solution for. Two factors are considered. One is the number of design variables; the other is the optimization method. To reduce the number of design variables, parametric approach is introduced. Control diameters are used to control all strut size across the entire structure by utilizing a concept similar to control vertices and Bezier surface. This operation allows the number of design variables to change from the number of elements to a small set of coefficients. In lattice structure design, global optimization methods are popular and widely used. These methods use heuristic strategies to search the design space and thus perform, as oppose to traditional mathematical programming (MP) methods, a better global search. This work propose that although traditional MP methods find local optimum near starting point, given a quick convergence rate, it will be more efficient to perform such method multiple times to integrate global search than using a global optimization method. Particle Swarm Optimization and Levenburg-Marquardt are chosen to perform the experiments.