Study on the flow fluctuation characteristics of dual-pump deep-sea mud lifting systems

This study investigates transient flow fluctuations in a dual-pump deep-sea mud lifting system during transitions from single-pump to dual-pump series and parallel operation. An indoor test platform was combined with Computational Fluid Dynamics–Amesim modeling to characterize the transient responses and evaluate model accuracy. The predicted steady-state flow rates deviated from the reference value by 0.83% and 5.83% for series and parallel operation, respectively. An L81 orthogonal design was applied independently to the two switching modes, generating 162 simulation cases for the prediction of the normalized flow fluctuation amplitude. On independent test sets, the RIME-extreme learning machines model achieved root mean square error (RMSE) values of 1.3087 and 0.6695, mean absolute percentage error values of 5.7560% and 3.9586%, and R 2 values of 0.9855 and 0.9895 for series and parallel switching, respectively. Compared with conventional extreme learning machines, the corresponding RMSE values were reduced by 55.9% and 72.3%. Extended Fourier amplitude sensitivity test analysis identified pump speed as the most influential factor in both switching modes (S i > 0.28; S Ti > 0.45), while the interaction between pump speed and the switching-valve opening exhibited the strongest second-order effect (S ij > 0.06). These results indicate that coordinated control of pump speed and valve opening is particularly important for mitigating transient flow fluctuations during series switching.