Modelling development and optimization on hydrodynamics and energy utilization of fish culture tank based on computational fluid dynamics and machine learning

Hydrodynamics of culture tank plays an essential role in the recirculating aquaculture system (RAS), maintaining the maximum effective energy utilization rate and uniform vortex distributions for fish growth is still a great challenge. To solve this problem, a novel approach to optimize physical parameters of octagonal tank using computational fluid dynamics (CFD) and machine learning (ML) is proposed. The initial tank is regarded as the benchmark, six vital parameters of octagonal tank including inlet and outlet diameters, fillet radius, inlet height, inlet angle and inlet velocity have been numerically investigated by CFD. Modelling development and optimization based on the Artificial Neural Network (ANN) and Nondominated Sorting Genetic Algorithm CYRILLIC CAPITAL LETTER BYELORUSSIAN-UKRAINIAN ICYRILLIC CAPITAL LETTER BYELORUSSIAN-UKRAINIAN I (NSGA-II) are developed to obtain the Pareto front for maximizing effective energy utilization rate and minimizing the vortex STD. Some key findings are found that: 1) The model provides high predictive capability: RMSE of average velocity is 0.002.2) NSGA-II combining ANN is applied in the optimization process to obtain 78 groups of optimal Pareto front.3) A series of optimal parameters are achieved by LINMAP, the optimal parameter combination could improve the energy utilization rate up to 12.32 corresponding with the benchmark. 4) Inlet velocity of 0.1 m/s and inlet diameter of 103.9 mm in the culture tank can be more significant for improving flow uniformity but also raising effective energy utilization rate. The proposed CFD-ML model prediction and optimization have satisfactory performance in hydrodynamics of culture tank.