In view of the characteristics of high nonlinearity and high computational complexity in the optimization design of construction machinery, as well as the demand for high-fidelity and low-cost simulation models, an optimization design method for construction machinery structure based on an adaptive Kriging surrogate model was proposed. Firstly, the design variable sample space of the construction machinery structure was partitioned using the Voronoi diagram method. The global sparsity index (GSI) of the sample points was defined to identify subspaces requiring node-adding, and global points were generated within these subspaces using the max-min criterion. Next, local node-adding subspaces were identified based on the potential optimal parameters, and local points were generated accordingly. Finally, by integrating a local refinement strategy, the generated global points, local points and potential optimal solutions were filtered and incorporated into the sample set to update the surrogate model, thereby improving the fitting accuracy of the surrogate model. It was applied to the surrogate model construction phase in the construction machinery structure optimization design to significantly shorten the computational time for traditional high-fidelity model simulation and improve the efficiency of the optimization design. The numerical example verifications and engineering application results demonstrated that, compared with the existing methods, the proposed method achieved faster convergence within an acceptable error range under the same node-adding number. In the lightweight design of the upper structure of the arm-type bucket wheel reclaimer and the front boom structure of the tower crane, after only one iteration, the maximum relative errors of the optimal solution were reduced to 0.09% and 0.59% respectively. The proposed method is feasible and effective in the structural optimization design of construction machinery.