Core configuration significantly affects the performance of sandwich panels. However, existing designs are mostly uniform, despite the potential benefits of non-uniform configurations. Non-uniform design involves variables from different levels, including number, position, and thickness of stiffeners, yet it has not been systematically studied. When attempting to obtain optimized designs, we found that existing methods formulate the problem as a mixed-integer bilevel optimization, which suffers from low efficiency. To address this, we propose a unified optimization method that transforms the bilevel problem into a single-level problem. Specifically, the panel is discretized using equidistant nodes, each assigned a unified design variable, and stiffeners are constructed by connecting activated nodes. This approach enables efficient optimization of non-uniform configurations. To validate the method, we conducted multi-objective optimization on triangular and trapezoidal corrugated cores under static out-of-plane pressure and dynamic planar compression. The results demonstrate that non-uniform designs achieve superior performance compared to uniform configurations. Furthermore, compared to the baseline bilevel optimization method, our approach achieves a higher hypervolume under the same evaluation budget, making it more efficient for designers to rapidly obtain optimized configurations.