The Sheet Molding Compound (SMC) process is widely used for producing composite structures in high-volume manufacturing due to its efficiency and scalability. A critical challenge, however, is determining the initial charge shape to achieve full mold filling without overflow, often requiring time-consuming trial and error. Although simulations can predict final mold filling, they often fail to optimize the initial preform shape, leading to inefficiencies in time and material usage. This research introduces a novel simulation-based methodology that accurately predicts the initial charge preform for two-dimensional (2D) mold geometries. Using Darcy’s Law and a fixed mesh grid system, the approach inverts the flow direction to simulate reverse material flow, thus identifying the optimal preform shape. This method reduces trial and error while ensuring precise control over the preform’s geometry. In tests, the simulation accurately predicted preform shapes, significantly reducing iterative adjustments. An accompanying machine learning (ML) model was then trained to predict preform shapes directly from mold geometry, final thickness, and initial charge thickness. Acting as a digital twin of the SMC process, the ML model provides results as accurate as the simulations but with far greater computational efficiency, avoiding the convergence issues found in traditional simulations. This integrated simulation and ML framework offers manufacturers a highly efficient and accurate tool for optimizing SMC processes, reducing waste and production time.