Flow characteristic optimization of a multi-stage orifice plate using surrogate-based modeling and Bayesian optimization

The efficient design of a throttle element is still a challenging open problem in fluid flow engineering, considering the flow characteristics, turbulence, and cavitation phenomenon. An optimization framework based on the Optimized Latin Hypercube Sampling method and Bayesian optimization is proposed, where the Optimized Latin Hypercube Sampling method is used to build a preliminary surrogate model within limited sampling points, then the best candidate location is guessed by the Bayesian optimization. First, two numerical cases are discussed to verify the correctness of the proposed optimization framework in mathematics, both cases find the best candidate in less than 200 iterations. Second, a two-stage throttle plate is used to verify whether the optimization framework is suitable for an engineering problem. The computational fluid dynamics is less than 5% error compared to the experimental data. In the two-variable case, the optimal candidate is better than the candidate of the exhaustive method with fewer simulations. Some common features of the geometric structure and flow characteristics are discussed in the five-variable case. Finally, the complex engineering design of the multi-stage orifice plate is optimized, and the best candidate is generated after 130 iterations, showcasing close flow velocity performance comparable to the two-stage case, a unique "X"-shaped flow path is observed in this candidate. The proposed optimization framework efficiently realizes high-performance orifice plates with less computational resources.