The development of pharmaceutical manufacturing processes has been facilitated by recent advances in the process simulation area. However, since the simulations are usually complex and the analytical expressions of the model can be unknown, it is difficult to directly apply traditional optimization techniques in such cases. Therefore, we propose a novel surrogate-based optimization approach to address such difficulties. Our algorithm is suitable to solve black-box constrained optimization problems. The effectiveness of this method is illustrated with several two-dimensional benchmark problems and a case study from the pharmaceutical manufacturing process.