<p>Hollow-divided-flow forging method was studied to achieve thin-walled multi-cavity parts with different cavity cross-sectional areas, using blanks of uniform thickness to form parts with similar wall heights around different cavities under smaller forming force. The feasibility of the hollow-divided-flow forging method and the effects of hollow cavity dimensions including cross-sectional area and thickness of blank on forming force and forming uniformity were analyzed. Then the BP neural network was employed to establish a surrogate model for hollow-divided-flow forging of thin-walled multi-cavity parts, replacing the complex trial and error of process design and avoiding studying complex material flow patterns. An optimization method for blank dimensions and reduction amount targeting minimized forming force and minimized deformation uniformity coefficient was designed based on genetic algorithm and the surrogate model, with forming experiments validating the accuracy of the optimization results. Based on the part design dimension and the maximum machining allowance, the optimization method can accurately and quickly predict the combination of blank dimensions and reduction rate to achieve minimized forming force and minimized deformation uniformity coefficient.</p>