To address the challenges of insufficient convergence drive and the loss of elite solutions in user-preferred regions during the solving of Many-Objective Optimization Problems (MaOPs), this paper proposes an improved NSGA-III algorithm. First, a Three-Stage Adaptive Differential Evolution (TS-ADE) mechanism is designed to balance global exploration and local exploitation by dynamically adjusting evolutionary operators and control parameters throughout the process. Second, an elite-priority dual-layer niche selection strategy is introduced. By incorporating a “relative optimization degree” metric, this strategy prioritizes the preservation of superior individuals within user-preferred regions, overcoming the flaw of standard algorithms that accidentally delete engineering elites due to an overemphasis on geometric uniformity. To validate the algorithm's performance, it is applied to a complex engineering instance: the five-objective parameter optimization of a high-speed train body section. Experimental results demonstrate that the proposed algorithm outperforms the standard NSGA-III in 83.3% of benchmark tests. In the engineering application, the Hypervolume (HV) indicator is improved by approximately 5.7%, and the convergence towards the Pareto front is enhanced by 35.2%. The final engineering scheme achieves a 4.33% mass reduction and significant stiffness improvement under all constraints, verifying the robustness and effectiveness of the proposed algorithm in handling high-dimensional, nonlinear complex engineering optimization problems.