Multi-objective optimization of inlet and outlet in a circular culture fish tank using the GBDT-NSGAII algorithm and computational fluid dynamic

In a circular culture tank, hydrodynamics plays an essential role for the growth and development of aquatic products in the recirculating aquaculture system (RAS). The challenge of maintaining the maximum effective energy utilization rate and achieving uniform velocity distributions remains a significant hurdle. To solve this problem, a novel multi-objective optimization approach was employed, utilizing the Gradient Boosting Decision Tree-Non-Dominated Sorting Genetic Algorithm II (GBDT-NSGAII) algorithm, to optimize the parameters of inlet and outlet of culture tanks as achieving suitable hydrodynamics. Specifically, the initial circular tank is regarded as the benchmark, four vital parameters of circular tank including nozzle quantities, nozzle diameters, outlet diameters, and inlet angles has been investigated by CFD model. The development of a GBDT-NSGA-II-based model and parameter optimization is conducted to obtain the Pareto front, with the objective of maximizing effective energy utilization rate while minimizing velocity STD. The main results show that: 1) At an inlet velocity of 0.5 m/s, the benchmark successfully establishes dynamic homogeneous velocity fields, while exhibiting a highly pressurized domain in the inlet pipe primarily influenced by the nozzle diameter. 2) The GBDT model demonstrates excellent predictive capability for the CFD database, with an RMSE of 0.002 m/s for average velocity and an RMSE of 0.304 % for velocity STD.3) The diameter of nozzles (12.76 mm) and the number of nozzles (7) have a highly significant influence on both average velocity and velocity uniformity compared to the benchmark model among the 63 groups of optimal Pareto fronts. 4) The optimal inlet-outlet parameter combination mainly includes an inlet velocity of 0.49 m/s and an outlet diameter of 22.26 mm, which improves the energy utilization rate up to 88 % compared to the benchmark model. The GBDT-NSGA-II model, driven by CFD simulation dataset, can effectively replace complex CFD calculations and save computer resources, exhibit commendable performance in accurately predicting the hydrodynamics of circular culture tanks in RAS.