Zhiyi Yu

Physics-informed neural networks (PINNs) enable mesh-free solutions of partial differential equations (PDE) but can suffer from training stagnation or convergence to non-physical solutions in moderate-to-high Reynolds number non-Newtonian bluff-body flows. This study investigates steady confined channel flow around a square cylinder for power-law fluids and evaluates, within a data-free PINN framework (boundary and PDE residual constraints only), the effects of Fourier feature embedding and a re-initialization training strategy. Predictions are performed for Reynolds numbers Re=10,40, and 100 and power-law indices n=0.5,1.0, and 1.4 and are compared against benchmark computational fluid dynamics results. Fourier features improve accuracy at low Reynolds numbers but can yield low-residual yet non-physical solutions at higher Reynolds numbers, manifested by windward stagnation point displacement and attenuation of predicted magnitudes. Re-initialization mitigates training stagnation and reduces errors but may induce non-physical wake deflection at the highest Reynolds number. These findings clarify how feature embedding and re-initialization affect PINN robustness for confined non-Newtonian flow simulations.

In the chemical and petroleum industry, the axial flow pump is widely used for the circulation pipeline system, and most of the transportation mediums are the shear-thinning non-Newtonian fluids. However, previous investigations on axial flow pumps are focused on water, which leads to a considerable deviation between the actual application and the research finding. In this work, shear-thinning non-Newtonian fluid (CMC solution) and viscous Newtonian fluid (the viscosity equals the apparent viscosity of CMC solution as the flow index is 1) are selected as the working medium. Based on the research output, lower apparent viscosity occurs in the near-wall and rotor–stator interaction region due to the larger velocity gradient. The shear-thinning property results in an increased tip leakage flow rate, and a sharp decline in friction loss. Compared to the viscous Newtonian fluid, the head and efficiency of the pump improves substantially for the shear-thinning fluid. The discrepancy is observed to increase with a higher flow rate. The comprehensive analysis of flow field and energy performance reveals that friction loss is still the main part of the total loss in the shear-thinning fluid.

The multiphase rotodynamic pump is widely used in petroleum and gas exploitation, and blade tip clearance may cause flow instability and performance deterioration. In the present work, the influence of tip clearance on the transportation characteristic in a multiphase rotodynamic pump is investigated based on the non-uniform bubble model, in which the bubbles’ coalescence and break-up are considered. The influence mechanism of tip clearance on the energy performance is revealed. The results show that the leakage flow rate increases linearly with the increase in tip clearance, but variation in pump energy performance shows the opposite trend. In addition, a larger tip clearance results in a sharply decreased pressure increment in the impeller, while in the guide vane, the increment is raised slightly. For the 0 mm tip clearance condition, a positive vortex (relative to the impeller rotation direction) is observed in the impeller passage. However, the opposite leakage vortex is also found in the region near the tip clearance when the tip clearance is considered, and the vortex strength increases for a larger tip clearance.

Visualization experiments and numerical simulations of two-phase flow are conducted to study the pressure fluctuation characteristics of oil-gas flow in horizontal pipes. The two-fluid model based on the Eulerian–Eulerian method is adopted, and the bubble number density equation (BNDE) is introduced to the simulation to predict the bubble size and distribution within the pipe. The bubble size and pressure variations in the pipe obtained from the simulations agree well with the recorded values from the experiments. The fast Fourier transform (FFT) algorithm is used to analyze the characteristics of pressure fluctuations, and the results show that the sudden pressure increase in the pipe sections is related to gas injection. The bubble number density increases with liquid flow rate (Q_l), which causes the oil-gas flow to be more turbulent while increasing the amplitude of high-frequency fluctuations. The maximum amplitude for the dominant frequency is observed near the pump inlet for low liquid flow rates. At high liquid flow rates, more liquid vortices are found near the gas orifice, and there is a maximum amplitude for the dominant frequency in this section. Due to the high swirling strength at larger inlet gas volume fraction (IGVF), there is an obvious increase in the amplitude of low-frequency fluctuations, while the amplitude of high-frequency fluctuations is nearly the same under all IGVF.

Study of bubble distribution characteristics is the basis to understand the mechanism of gas-liquid two-phase flow in multiphase pumps. In the present work, a method based on Eulerian-Eulerian approach was used for the simulation of the flow in a multiphase rotodynamic pump. The bubble size was calculated through a bubble number density equation. The distribution characteristics of bubbles were analysed in conditions of different inlet gas volume fraction (IGVF). The results show that bubbles move to the hub and a homogeneous bubble size distribution can be observed due to the rotational effect and stirring action of the impeller blades. The bubble size increases in the vicinity of interaction region because of the rotor-stator interaction. Collision rate of bubbles increase and a large bubble size is shown in the diffuser, which is resulted from the decrease of bubble velocity. It is found that high IGVF cause small distance between bubbles so that collision rate of bubbles increase. Largest bubble size and gas volume fraction are obtained when IGVF is 15%.

Pressure fluctuation in single-phase pumps has been studied widely, while less attention has been paid to research on multiphase pumps that are commonly used in the petroleum chemical industry. Therefore, this study investigates the pressure fluctuation for a multiphase rotodynamic pump handling air–water two-phase flow. Simulations based on the Euler two-fluid model were carried out using ANSYS_CFX16.0 at different Inlet Gas Void Fractions (IGVFs) and various flow rate values. Under conditions of IGVF = 0% (pure water) and IGVF = 15%, the accuracy of the numerical method was tested by comparing the experimental data. The results showed that the rotor–stator interaction was still the main generation driver of pressure fluctuation in gas–liquid two-phase pumps. However, the fluctuation near the impeller outlet ascribe to the rotor–stator interaction was weakened by the complex gas–liquid flow. For the different IGVF, the variation trend of fluctuation was similar along the streamwise direction. That is, the fluctuation in the impeller increased before decreasing, while in the guide vane it decreased gradually. Also, the fluctuation in the guide vane was generally greater than for the impeller and the maximum amplitude appeared in the vicinity of guide vane inlet.

Recent researches show that the pressure increment of a multiphase pump is affected by bubble size and distribution. In order to study the bubble distribution characteristics in such pumps, a novel approach describing the variable bubble size in the pump is proposed. The bubble number density equation, which has taken into account the phenomena of break-up and coalescence, is introduced into the flow simulation, and the drag coefficient is revised because of the interaction of multiple bubbles. The reliability of the approach is verified by comparison with the experiment. It was established that the bubbles move to the impeller hub due to the difference in centrifugal force between gas and liquid. Despite the high collision rate near the hub, bubble size changes little with the stirring action of the impeller. The mixture flows in a disorderly way and the bubble diameter increases due to the rotor–stator interaction. Owing to the increasing flow area in the diffuser, bubbles move to the mainstream region, and bubble size reaches its maximum owing to the flow separation near the hub. The distribution of bubbles is also analyzed under a different inlet gas volume fraction (IGVF) and inlet bubble diameter (d₀). Bigger IGVF brings about a higher collision rate of bubbles, while smaller d₀ makes the diffusion of bubbles easier.