The process parameters for gravity casting bearing seats in wind turbines are optimized using a multi-objective optimization framework that combines casting process numerical simulation, response surface methodology (RSM), the dragonfly optimization algorithm (DOA), and the X-ray detection of casting product quality. The experiments follow a Box–Behnken design (BBD) to evaluate key casting parameters and response indicators. The crucial variables included are pouring speed, pouring temperature, and mold temperature. Defect volume fraction (DVF) and secondary dendritic spacing serve as response markers. Regression analysis develops nonlinear input–output models, whereas the analysis of variance assesses model significance. DOA expands the population to identify optimal casting parameters, with numerical simulations confirming the best option. Results demonstrated the optimized parameter set of mathematical models. That is, the mold temperature is 100 °C, the pouring speed is 0.07 m/s, and the optimal pouring temperature is 1380 °C. These casting parameters were used to fabricate the bearing seat experimentally, and defects in the products were detected using sensors. Metallographic observation by X-ray detection confirmed that the bearing seat exhibited strong mechanical properties without any flaws detected by X-ray. The proposed framework faces limitations and pending problems including nonuniform casting quality owing to optimized parameters from simplified models and time-consuming, costly experimental validation. Overcoming these challenges is central to our subsequent work.