In design optimization of crane metal structures, present approaches are based on simple models and mixed variables, which are difficult to use in practice and usually lead to failure of optimized results for rounding variables. Crane metal structure optimal design(CMSOD) belongs to a constrained nonlinear optimization problem with discrete variables. A novel algorithm combining ant colony algorithm with a mutation-based local search(ACAM) is developed and used for a real CMSOD for the first time. In the algorithm model, the encoded mode of continuous array elements is introduced. This not only avoids the need to round optimization design variables during mixed variable optimization, but also facilitates the construction of heuristic information, and the storage and update of the ant colony pheromone. Together with the proposed ACAM, a genetic algorithm(GA) and particle swarm optimization(PSO) are used to optimize the metal structure of a crane. The optimization results show that the convergence speed of ACAM is approximately 20% of that of the GA and around 11% of that of the PSO. The objective function value given by ACAM is 22.23% less than the practical design value, a reduction of 16.42%over the GA and 3.27%over the PSO. The developed ACAM is an effective intelligent method for CMSOD and superior to other methods.