Performance improvement of a vortex pump based on volute shape modification Authors: Kun Lan, Yaguang Heng, Qifeng Jiang, Yunchuan Tu, Chuan Wang, Jie Ge Published: 2025 ABSTRACT AbstractVortex pumps are widely used in many engineering fields because of their small size and high lift head. However, the efficiency of this pump is quite low compared with that of other pumps. Therefore, the objective of the present study is to increase the efficiency of a vortex pump while increasing the head and reducing the power consumption. Numerical simulations are performed to obtain the internal flow pattern of the pump, and the nonideal double-vortex structure appears under all flow conditions. Modification of the volute shape was then proposed to optimize the double-vortex flow pattern, and the results revealed that the trapezoidal-shaped volute greatly improved the internal flow pattern. Additional modifications are proposed to reduce the power consumption of the pump. The experimental results of the optimized model show increase in the head and efficiency of 3.6 m and 3.2% (under nominal conditions), respectively, and a decrease in the power consumption from 10 to 20 W. This work highlights the importance of the volute shape on the performance of vortex pumps; a proper volute diffusion angle effectively suppresses the nonideal double-vortex structure and then avoids unnecessary energy losses.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-30084-4.Keywords: Vortex pumps, Performance optimization, Internal flow analysis, Volute optimization, Entropy production analysisSubject terms: Mechanical engineering, Fluid dynamics Introduction A vortex pump is a special type of pump characterized by a high lift head, low flow rate, and compact structure1. The impeller structure and internal flow pattern of vortex pumps are quite different than those of conventional vane pumps, the impeller structure is similar to that of a gear, and many small blades are distributed at the impeller periphery, as shown in Fig. 1. During the operation of vortex pumps, fluid enters from the inlet and moves to the outlet, the streamlines are close to the spiral, and the fluid receives energy periodically by entering the blade passages, which results in vortex pumps with very high lift heads in a single stage. However, strong impact losses are generated because the fluid enters the blade passages periodically. In addition, the internal flow pattern is swirling flow, which consumes additional energy. Consequently, the pump efficiency is always very low2. As reported by Karlsen-Davis and Aggidis3 and Sreekanth et al.4, the maximum efficiency of vortex pumps might be physically limited to 50%. Nevertheless, the small size and high lift head are advantages of vortex pumps; thus, many studies of vortex pumps have been conducted. Structure of the vortex pumps. Because of the low efficiency of vortex pumps, numerous studies have focused on optimizing the performance of vortex pumps by investigating the effects of various geometrical/operational parameters on pump performance. Kang and Ryu5 investigated the effect of the Reynolds number (between 5.52 × 103 and 1.33 × 106) on the performance of a small vortex pump and reported the strong dependence of the internal flow pattern on the Reynolds number. When the Reynolds number is small, the momentum exchange decreases because of the increasing viscous effects. When the Reynolds number decreases, the maximum efficiency point shifts toward a lower flow rate, and both the maximum flow rate and head decrease. Quail et al.6,7 analyzed the effects of impeller geometrical parameters (ten blade profiles are designed) on pump performance via computational fluid dynamics (CFD) and experimental verification. The results revealed that the internal helical flow pattern can be obtained from the CFD results, that the circumferential eddy distribution and pressure fluctuations are consistent with the experimental results, and that the CFD performance error is within 3%. These conclusions are important references for optimizing the performance of vortex pumps. Choi et al.8 investigated the effect of the blade angle on the performance of a vortex pump and reported that the pump performance strongly depends on the blade angle and blade geometry; a chevron blade with an angle of 30° exhibited the best performance. Nejadrajabali et al.9 also investigated the effect of the blade angle on the performance of a vortex pump, and 12 impellers with different blade configurations were tested experimentally. The results revealed that the lift heads of forward blade angle designs are greater than those of radial and backward blade angle designs, and the best performance is obtained with the model with a symmetric forward blade angle. Verma et al.10 designed a high-speed vortex pump via the CFD approach and reported that the lift head increases with increasing side channel width but decreases significantly when the vane thickness increases. Li et al.11 analyzed the effect of the tongue clearance of a vortex pump and reported that a change in the tongue channel section modified the circumferential vortex distribution and that the vortex structure shifted to a larger radius when the tongue width increased. In addition, the circumferential velocity and pressure gradients decrease with increasing tongue section. Chen et al.12 investigated the effects of geometric parameters (number of blades, impeller radius, and impeller radius ratio) on pump performance and reported that the best configuration was 35 blades and a radius ratio of 0.65. An internal flow analysis revealed that a decrease in the flow rate causes an increase in the axial velocity and a decrease in the circumferential velocity. The increased axial velocity improves the mass transfer and consequently results in an increase in pressure. On the other hand, the decrease in the circumferential velocity component increased the number of interactions between the fluid and the blade, which caused an increase in shock losses. Recently, optimization approaches have also been applied to achieve high-performance vortex pumps. Meakhail and Park13 proposed an improved model to predict the performance of vortex pumps, which can be used to design twisted blades for performance optimization. This model shows that the performance of vortex pumps strongly depends on the side-blade exit angle. The pressure fluctuation in this vortex pump is more significant than that in other pumps because of the periodical interaction between the fluid and the impeller. On the basis of improved momentum exchange theory, Yoo et al.14 proposed a design procedure to determine the best configurations of a vortex pump for achieving the desired lift head and best efficiency under nominal flow conditions. Nondimensional analysis of the effects of geometrical parameters on pump performance was conducted, and the performance of the designed pump prototype was validated via experimental measurements. Salimi et al.15 used response surface methodology (RSM) to optimize a vortex pump by selecting the S-shaped impeller, blade number, thickness, and curvature as the optimization variables, and the efficiency increased by 2.85% via single-objective optimization. Based on orthogonal design method (ODM), Yang et al.16 optimized a vortex pump model by selecting the impeller diameter, blade groove radius, blade number, and impeller width as impact parameters, and their results revealed that the number of blades is the most significant impact factor. To obtain a deeper understanding, some studies have focused on the internal flow pattern or performance characteristics of vortex pumps. Zhang et al.17 numerically investigated the pressure fluctuations in a vortex pump installed with two different impellers, and the results revealed that the dominant frequency is the blade passing frequency, the pressure amplitude fluctuates from the inlet to the outlet, and the maximum amplitude appears at the stripper. The pressure fluctuation in the impeller with unequal circumferential spacing between two adjacent blades is weaker than that in the model with equal spacing, mainly because the harmonics reduce the dominant frequency without decreasing the pump performance. Mattern et al.18 visualized the complex three-dimensional swirling flow pattern of a vortex pump via time-resolved particle image velocimetry (PIV) measurements. Their study enables a better understanding of the momentum exchange and working principle of the vortex pump internal flow field. Li et al.19 investigated the energy characteristics of a vortex pump and validated the presence of circulation flow in the pump. This flow can be evaluated by analyzing the circulation number, entropy production and axial distance. These authors also established the relationship between these parameters and pump performance, which provides a reference for performance optimization. Nejadali20 employed a vortex pump as a turbine and proposed a theoretical model to predict the performance of a vortex pump in reverse mode. The results revealed that the most significant energy losses appear in the blade area and that the pump efficiency decreases by 20% in reverse mode. Yang et al.21,22 analyzed the principles of energy loss and conversion in a vortex pump by applying the energy transport equation and reported that energy losses caused by vortex alterations are significant and that the vortex structure is the key factor affecting pump performance. Mojaddam et al23 summarizes the design methods of the volute of the radial compressor, and conducts parametric studies on the spiral cross-sectional area, circumferential area distribution, outlet cone and tongue area of the volute of the compressor to determine the optimal value. The results show that the volute design method is different at the design point and off-design conditions. Increasing the pressure ratio at the design point and reducing the total pressure loss at the design point will lead to the deterioration of the off-design conditions and the increase of the radial force on the impeller. The above studies systematically investigated the effects of various parameters on pump performance and revealed the internal flow patterns and characteristics of vortex pumps. However, conclusions of the effects of parameters cannot be drawn for all cases because only several parameters were considered in each study. Because the internal flow pattern of a vortex pump is characterized by strong and periodic interactions or convection between the impeller and the volute, all the parameters affect the pump performance. Therefore, understanding pump performance behavior by considering several isolated parameters is insufficient. Nevertheless, these studies reach the consensus that the vortex structure or vortex distribution has a decisive effect on the vortex pump performance. Therefore, in the present study, we attempt to optimize the performance of a vortex pump by considering the optimization of the vortex structure. To achieve this goal, numerical simulations have been employed to obtain the internal vortex flow pattern of the original vortex pump model. By analyzing the deficiency of the vortex distribution, modifications to the volute shape are then proposed to optimize the vortex structure. The numerical results revealed that the optimization of the vortex structure effectively decreased the internal losses and improved the energy conversion efficiency. The optimized model is then manufactured and tested, and the experimental results validated that the head and efficiency are increased by 3.6 m and 3.2%, respectively, under nominal conditions and that the power consumption decreased by 10 W to 20 W. Setup of the numerical simulations 3D model UG NX 12.0 is employed to construct a 3D model of the vortex pump with the following important geometrical parameters: blade number N = 41, impeller outlet diameter D2 = 60 mm, impeller width B = 6 mm, blade tip-to-root ratio r2/r1 = 1.4, and blade height h = 9 mm. The cross-sectional shape of the volute is rectangular (the volute shape is presented in section "First attempt: Volute shape modification and comparisons"), and the whole fluid domain includes the inlet, impeller, volute and outlet, as shown in Fig. 2. Considering the stability of inlet flow and outlet flow, the inlet and outlet pipes have been extended to 2 times of the pipe diameter. Fluid domains of the original vortex pump model. Mesh model ANSYS Mesh 19.2 is used to generate the mesh model of the fluid domains. Both structured and unstructured meshes are employed to discretize the fluid domains. The structured mesh is characterized by less mesh number and high computational speed but is limited in terms of fitting models with twisted or sharp shapes. Compared with the structured mesh, the unstructured mesh is adaptable for any model shape, but the number of meshes is much more than that of the structured mesh and results in a slower computational speed. Based on these features, structured mesh is used to discretize the impeller fluid domain because of its regular geometric structure, and unstructured mesh is applied for the rest fluid domains. Owing to the strong velocity gradient near the wall surface, all the boundary layers are refined to increase the accuracy of the calculations. The boundary layer is set to number layers of 5, growth of 2, and maximum thickness of 0.08 m. For regions with steep gradients and small curvatures, more refinement is implemented. The final mesh models are shown in Fig. 3. The values of Y + for the impeller wall are below 150, and for the volute wall are below 60 in Fig. 4, which are sufficient for the calculations, as follows. Mesh models of the original vortex pump. Impeller and volute wall Y+ Turbulence model and boundary conditions The generated mesh models are imported to ANSYS CFX for the numerical setup. The governing equations, including the continuity equation and momentum equation, are defined as follows: where r represents density (kg/m3), t represents time (s), u represents velocity (m/s), τ represents stress (Pa), F represents mass force (N), and the subscripts i and j represent different directions. Considering the strong swirling flow in vortex pumps, so standard k–ε turbulence model is selected for the calculations. This turbulence model relies on two extra equations to calculate the turbulent kinetic energy k and turbulent dissipation rate ε, as follows: where Gk represents the turbulent kinetic energy-generated average velocity gradient, C1ε and C2ε represent empirical coefficients, and σk and σε represent the turbulent Prandtl numbers of k and ε, respectively. For the boundary conditions, the total pressure condition (1 atm) is specified as the inlet condition, and the mass flow rate is specified as the outlet condition. The impeller is set as a rotating component with a rotational speed of 2900 rpm, and the rest components are static domains. Interfaces are used between the rotating impeller and the static volute. The interface between the rotating impeller and the stationary volute is set to Frozen Rotor. Frozen Rotor is used for steady-state simulation. It does not simulate real transient interference, but retains the spatial inhomogeneity caused by different relative positions. The nonslip wall condition is specified for all surfaces. The number of convergence steps is specified as 1000 steps, and the convergence precision is 10–4, as shown in Fig. 5. The calculated performance parameters are also at a stable level in Fig. 6. Convergence curves. Pressure at monitoring points as a function of the number of convergence steps. Mesh independence analysis The mesh number is an important factor that directly affects the calculation results; an insufficient mesh number results in insufficient calculation accuracy and numerical divergence, and an excessive mesh number results in the increase of calculation costs in terms of time and computer resources. Therefore, the balance between calculation accuracy and computer resources must be considered in numerical simulations. Seven sets of mesh numbers are calculated (steady calculation) for the present model under nominal flow conditions (Qn = 1.0 m3/h, N = 2900 rpm):0.79 million, 1.5 million, 3.4 million, 4.8 million, 5.8 million, 6.4 million, and 7.6 million. The pump lift head and efficiency are chosen as indicators, and the results are shown in Fig. 7. The performance curves are stable at a mesh number of 5.8 million, and a further increase in the mesh number has almost no effect on the performance (head and efficiency variation are smaller than 0.5%). Therefore, all the calculations performed in the present study use a model with 5.8 million meshes (for each component: 1.77 million meshes for the inlet, 1.74 million meshes for the volute, 1.56 million meshes for the outlet, and 0.73 million meshes for the impeller). Mesh independence analysis. Experimental measurements and verification of the original model Experimental measurements are performed to validate the numerical results, and the tested pump model is shown in Fig. 8. Vortex pump test rig (original model). During the test, before the pump is started, the outlet gate valve is closed first to prevent the start-up duration from being too long and the motor current is too large. Then the motor is started to make the vortex pump start to run, gradually open the outlet gate valve, and slowly adjust the speed to the rated working condition, so that the operation slowly reaches a stable state. After the operation is stable, the control cabinet controls the opening of the outlet gate valve to adjust the flow rate, so as to carry out the external characteristic test under different flow conditions. During the measurements, the vortex pump operates in an open loop, two pressure transducers are mounted at the inlet and outlet pipelines, and the pump lift head H is calculated as follows: where z represents the vertical position (m), v represents the absolute velocity (m/s), p represents pressure (Pa), and subscripts 1 and 2 represent the pump inlet and pump outlet, respectively. A torquemeter is used to measure the torque and rotational speed, and the pump efficiency η is calculated as follows: where Q represents the flow rate (m3/s), M represents the torque (N·m), and ω represents the impeller angular velocity (rad/s). The measurement precision (device error) for each parameter is as follows: flow rate of ± 0.5%; rotational speed of ± 0.2%; pressure of ± 0.2%; and power of ± 0.5%. The total uncertainty for the head and efficiency are then estimated as follows: where R represents a variable, i represents the number of measurements, X represents the coefficient of the related variable, and UR represents the experimental uncertainty. This uncertainty is generated from experiment error because the head and efficiency are estimated by Eqs. (5) and (6) which contain many variables measured from experiments. Based on these formulas, the uncertainties of the head and efficiency are estimated as ± 0.28% and ± 0.77%, respectively. A comparison between the numerical and experimental results is shown in Fig. 9. The nominal flow rate is Qn = 1.5 m3/h because of the highest efficiency, and the numerical results agree with the experimental measurements. Both curves have the same tendency, but the numerical results are always slightly higher than the experimental results because of two reasons: firstly, the no-slip wall condition is selected and wall roughness is neglected in the calculations. The fluid velocity close to the volute wall is not zero, which is higher than those in experiments; secondly, CFD calculates only hydraulic efficiency, neglecting factors such as bearing friction losses and seal leakage losses which actually existed in experiment measurements. Consequently, the numerical results are higher than experimental results. For the head calculation, most of the results are quite close, with the exception of the lowest flow rate conditions in which the maximum deviation appears, probably because of the presence of strong instability in the experiments when the flow rate is low, which results in the experimental pump head being lower than that in the numerical calculations. The maximum efficiency deviation appears at the maximum flow rate because friction loss increases with increasing flow rate in the experiments, but the friction loss in the calculations is much lower because of the smooth wall boundary conditions. Relative error of head at rated flow rate is 7.4%, and the error in efficiency at rated flow rate is 2.4%. Comparison of the experimental and numerical results (original model). Optimization of the pump performance Analysis of the internal flow of the original pump To optimize pump performance, it is necessary to analyze the internal flow field to obtain the undesired flow pattern within the pump. Figure 10 shows the positions of the sections used in the following internal flow analysis, and Fig. 11 presents the streamline distributions of sections A, B and C, which represent the pump inlet, middle area and outlet area, respectively. Important locations and sections in the pump. Internal streamlines of sections A, B, and C under different flow conditions. Section A is near the inlet but at a certain distance from it, which allows the flow to develop fully. Therefore, the flow pattern in this section is used to represent the fully developed flow pattern at the inlet. Theoretically, based on the mechanism of vortex pumps, the ideal flow in each section should be a single-vortex flow pattern, as shown in Fig. 12, in which the pressurized fluid is discharged from the blade tip side, moves along with the internal surface shape of the volute, and finally reenters the adjacent blade area from the blade root side to obtain energy. Besides, the vortex core should be located near the interface between impeller and volute to ensure all fluids enter the impeller to get energy. Otherwise, for example, if the vortex core is located in the volute, then partial fluid will rotate around the vortex core and cannot get energy from the impeller. Ideal flow pattern in the flow channel. In figure, the general conclusions and key points can be drawn as follows: In section A: when Q = 0.6 m3/h, the flow pattern on the left-hand side is close to a single-vortex flow pattern, but on the right-hand side, two vortices are generated that interact with each other, which prevents partial fluid at the blade root side from entering the blade (indicated by a red circle). Consequently, these fluids cannot be pressurized by the blade, which reduces the energy exchange efficiency. When Q = 1.0 m3/h, although the flow fields on the left-hand and right-hand sides are close to a single-vortex flow pattern, but one can notice that near the root area of the right-hand side, some fluid still cannot enter the blade area (marked with a red circle). When Q = 1.5 m3/h, the right-hand side flow is close to the ideal flow pattern, whereas the left-hand side has a double-vortex structure. When Q = 2.0 m3/h, the double-vortex structure appears on both sides, but the left-hand side flow pattern differs from the previous double-vortex structure in that the locations of the two vortices change from vertical to horizontal. Near the blade root side of the right-hand side, the amount of fluid that cannot enter the blade area is greater than that at lower flow rates, which results in more hydraulic losses. Small vortices also exist in the top-middle area; however, the effect of these vortices on the pump performance is negligible. The flow patterns in Sections B and C are similar to those identified in Section A, and double-vortex structures are observed under different conditions. For example, the typical double-vortex structure can be identified in section B at all flow conditions but can be observed at only Q = 0.6 m3/h and Q = 1.5 m3/h in section C. When Q = 1.0 m3/h and Q = 1.5 m3/h, the flow patterns in section C are close to a single-vortex structure. In conclusion, the double-vortex flow pattern in the original model causes convection losses and blocks the fluid enter the impeller to get energy, this flow pattern must be modified to close the ideal flow pattern which has been proposed before. By doing that, convection losses and blockage generated by vortexes in the original model can be optimized and pump performance improvement can be expected, First attempt: volute shape modification and comparisons To optimize the internal double-vortex flow pattern, a convenient approach is to change the volute shape. According to the principle of fluid mechanics, the formation of the downside vortex is related to diffusion losses because of the excessive width of the volute downside casing, which gradually prevents the local fluid from entering the blades. During its movement toward the volute downside, the fluid gradually interacts with the outflow from the blade side, resulting in vortex generation. This phenomenon indicates that the diffusion angle of the original model is improper and must be modified. If the volute shape contracts from the blade tip to the blade loot, the diffusion angle decreases, and the fluid can be expected to enter the blade area more quickly and not interact with the blade outflow. Based on these guidelines and refers to a book2, In the Fig. 13, c = 0.5 b, h = b, a = (0.25–0.35) b, R = 0.5 b. Since h and a are determined in the prototype, they are used in the trapezoidal cross section. The 35-degree angle is determined to be tangent to the arc. A trapezoidal volute is designed to validate the assumption, as shown on the right-hand side of Fig. 14. The total cross-sectional area of the trapezoidal volute is the same as that of the original model. By using the same numerical setup, the performances of the two models are preliminarily compared, as presented in Fig. 15. Design parameters of trapezoidal volute. Comparison of the trapezoidal volute and original volute. Comparison of the numerical results between the trapezoidal volute model and the original model. The performance of the trapezoidal volute model is greater at high flow rates (Q = 1.5 m3/h and Q = 2.0 m3/h) and is quite close to that of the original model when Q = 1.0 m3/h; the head is lower only when Q = 0.6 m3/h. Nevertheless, the lowest flow condition is not very important because vortex pumps generally do not operate at the minimum flow rate. Specifically, the lift head increases by 2.3 m/1.5 m when Q = 1.5 m3/h/2.0 m3/h, and the efficiency improves by 2.5% at the best efficiency point (Q = 1.5 m3/h) and by 5% at the maximum flow rate. Considering the low efficiency of vortex pumps, these values can be considered significant improvements. To understand how the volute shape modification affects the pump internal flow field, the internal streamlines between the two models must be compared. Because the flow patterns in each section A/B/C are somewhat similar, the results in the middle section B are first presented in Fig. 16 for internal flow comparison. Overall, the streamline distributions of the trapezoidal volute model are clearly closer to the ideal single-vortex flow than those of the original model. Comparison of the streamlines in section B between the original model (top) and the trapezoidal volute model (bottom). When Q = 0.6 m3/h, the flow pattern on the left-hand side is greatly improved, and the double-vortex structure of the original model is modified to an ideal single-vortex structure. This helps to improve the energy exchange efficiency because the interaction between the vortices disappears. For the right-hand side, the flow patterns are similar for both models, but the secondary vortex is much smaller for the trapezoidal volute model. This finding is also observed for flow conditions where Q = 1.0 m3/h, and the size of the secondary vortex within the left-hand blade is slightly smaller. At the best efficiency flow condition, Q = 1.5 m3/h, and the maximum flow condition, Q = 2.0 m3/h, the flow pattern of the trapezoidal volute model is nearly ideal on both the left and right sides. Fluid is discharged from the blade tip and re-enters the blade loot area, and no flow convection or disturbance occurs in the main flow area. Small vortices are generated in the top-middle area, but this area has almost no effect on the main streamline and pump performance. In contrast, the main streamlines are always disturbed by the double-vortex structure in the original model (as indicated in the previous section), which results in significant hydraulic losses and reduces pump performance. Comparisons of the streamlines between the two models in sections A and C are additionally provided in Fig. 17 to show all the results. Undoubtably, the undesired flow patterns in the original model have been greatly improved by optimizing the volute shape, and analogous conclusions can be drawn, similar to the analysis in section B. Comparison of the streamlines in sections A and C between the original model (top) and the trapezoidal volute model (bottom). Second attempt: optimization of power consumption Based on the preliminary CFD results, it can be concluded that the modification of the volute shape effectively improved the performance of the original model under most flow conditions. However, one can also notice that the power consumption of the trapezoidal volute model is greater than that of the original model. Then, the second objective of the present study focuses on reducing pump power consumption. The power consumption is generally directly proportional to the volute cross-sectional area; therefore, the angle is reduced to 30 degrees, the volute cross-sectional area is reduced by 10%, and the shape remains the same. The reduction in the volute cross-sectional area causes a decrease in the pump head for two reasons: first, the circumferential velocity increases when the flow passage is narrower, which results in a decrease in the number of times the fluid enters the blade to obtain energy, as reported in Chen et al. (2024); second, the increased circumferential velocity increases the overall friction loss. To compensate for this decrease in head, the matching relationship between the impeller and the volute is modified by increasing the number of blades from 41 to 51. After these modifications, the new model is referred to as the optimized model for convenience. The optimized model is then manufactured to test its performance, the volute is manufactured via 3D printing, the comparison between the original model and the optimized model is shown in Fig. 18, and the experimental performance results are shown in Fig. 19. The head and efficiency curves of the optimized model are higher and the power consumption is lower than that of the original model under all flow conditions. The maximum performance improvement occurs at a flow rate of Q = 1.0 m3/h, in which the efficiency improves by 3% and the head improves by 3.6 m. For all flow rates, the power consumption of the optimized model is 5%–10% lower than that of the original model. The experimental results confirmed that multi objective optimization of the original pump was successfully achieved‌. Comparison of the original model (left) and the optimized model (right). Comparison of the experimental results between the optimized model and the original model. Analysis of the internal flow field of the optimized model Pump internal streamlines and total pressure distributions To investigate the internal flow pattern, numerical simulations of the optimized model are performed via the previous CFD setup. First, the streamline distributions of the optimized model are presented in Fig. 20. A comparison of Fig. 20 with Fig. 17 reveals that the streamline distributions of the optimized model are similar to those of the trapezoidal volute model in the same cross-sectional area, which indicates that a decrease in the cross-sectional area does not substantially change the internal flow pattern. Internal streamline distributions of the optimized model. Figure 21 shows the variation in the total pressure of the two models from the inlet to the outlet. From the inlet to the outlet, the total pressure increases gradually. With increasing flow rate, the total pressure generally decreases because the circumferential velocity increases, and then the number of times the fluid enters the blade to obtain energy decreases, which indicates that the fluid flows out of the pump more quickly, as shown in Fig. 22. In addition, the increase in velocity is accompanied by greater friction and local losses, which are proportional to the square of the velocity. Therefore, the total pressure decreases with increasing flow rate. Comparison of the total pressures (top: original model; bottom: optimized model). Streamlines under different flow conditions. Figure 23 shows more details of the total pressure distribution between the two models, which elucidates the internal pressurization of the blade. An analysis of Fig. 21 with the streamline distributions in Fig. 11 and Fig. 17 clearly reveals that for the original model, the lowest total pressure area in each section always corresponds to the double-vortex area, which includes not only the vortex core but also the interaction area. For the optimized model, the low total pressure area is greatly reduced because of the optimization of the internal flow pattern, and the vortex area is much smaller than that of the original model. In addition, the evolution of the total pressure from the blade root side to the blade tip side is more reasonable, which means that the total pressure gradually increases from the blade loot side to the blade tip side without any perturbation from the interaction between the vortices. Comparison of the total pressure in each section. For example, strong perturbation can be observed in section B of the original model when Q = 0.6 m3/h (marked in Fig. 23b(1)). These perturbations are generated by the interaction between vortices, as shown in Fig. 11b(1), and they occupy the internal flow area of the blade and interact with the internal flow of the blade, resulting in a discontinuous distribution of total pressure within the blade area. In contrast, the total pressure distribution within the blade is always continuous with the optimized model. The low-pressure areas/perturbations are located outside the blade area under most flow conditions and do not interact with the internal flow of the blade. This ensures high energy exchange efficiency. Entropy production analysis Entropy production is positively correlated with energy loss within the pump, which can be an indicator of pump efficiency. The change of entropy is due to the heat conduction caused by the temperature gradient between the computational unit and its surrounding environment, and the energy dissipation caused by viscous stress24. The internal flow pattern of vortex pumps is typically a swirling flow, which generates many vortices and results in substantial energy loss. Because the variation in temperature is not considered in the numerical simulations, therefore, the temperature gradient is constant to 0, so the wall entropy generation is ignored, only direct dissipative entropy production and turbulent dissipation entropy production are considered, and the effect of heat is disregarded. Direct dissipative entropy production is linked with the change in average velocity components, and turbulent dissipation entropy production is generated by turbulence fluctuations25–27. These variables are defined as follows: where Eq. (9) represents direct dissipative entropy production, Eq. (10) represents turbulent dissipative entropy production, T represents the temperature and is constant in the simulations, - represents the average value, and ' represents the fluctuating components. According to these equations, entropy production reflects the intensity of the change in the flow field, including the shear and linear deformation rates, which generally reflects the internal losses of the pump. The direct and turbulent dissipation entropy productions versus flow rate are plotted in Figs. 24 and 25 is entropy productions of impeller and volute versus flow rate. It can be seen from the diagram that the entropy generation of the impeller is higher, and the entropy generation in the impeller and volute of the optimized model is lower than that of the original model. Direct and turbulent entropy productions versus flow rate. Entropy productions of impeller and volute versus flow rate. The turbulent entropy production is clearly the dominant contributor to entropy production in the pump, and both values decrease with increasing flow rate, which indicates that the intensity of deformation of the internal flow field weakens at higher flow rates. The direct and turbulent entropy productions of the optimized model are smaller than those of the original model under all flow conditions. To investigate internal entropy production in detail, the total entropy production distributions of the two models are presented in Fig. 26. A comparison of Fig. 26 with Figs. 11 and 20 reveals that the high-entropy production area is concentrated in the vortex area, which is characterized by strong deformation in the flow direction and is consistent with the definition of entropy production. In addition, a high-entropy production area is observed near the wall surface, generally because the flow is greatly changed in the boundary layer. Undoubtedly, the size of the high-entropy production area in the optimized model is smaller than that of the original model under all flow conditions and in all sections because of the optimization of the internal vortex structure. Comparison of the entropy production rates of the two models. Conclusions In the present study, the performance of a vortex pump is optimized by modifying the volute shape, and the pump performance, internal flow field (streamline distribution and total pressure distribution) and entropy production of the vortex pump are analyzed via numerical simulations. The optimized model is experimentally tested, and the results show that the head and efficiency at nominal condition are increased by 3.6 m and 3.2%, respectively. Under all flow conditions, the power consumption is decreased by 10 W to 20 W. The main conclusions are summarized as follows: (1) Volute shape significantly affects pump performance by altering the internal flow pattern. The width of the original rectangular shaped volute from the blade tip side to the blade root side is constant, which results in strong flow diffusion and the formation of a double-vortex structure. These vortices not only interact with each other but also prevent fluid from entering the blade area to achieve pressurization, which results in extra energy losses. To reduce the diffusion angle, a trapezoid-shaped volute is proposed to guide the flow streamline, and the numerical results show that the internal flow pattern is improved from a double-vortex structure to an ideal single-vortex structure. (2) Additional modifications (volute cross-sectional area and number of blades) have been proposed to achieve multi objective optimization design. The optimized model was manufactured and tested in the laboratory, and the experimental results confirmed that the optimized model has a higher head and efficiency and lower power consumption than the original model. The streamline distributions show that the internal flow pattern is near ideal even when the volute cross-sectional area is reduced by 10%. (3) The total pressure and entropy production distributions in different sections are analyzed. The total pressure distribution of the original model is discontinuous because of the perturbation generated by the interaction between vortices, which results in additional energy losses. In contrast, the total pressure distribution of the optimized model is continuous under most conditions, and the low-pressure area generated by the vortex is always external to the blade area, which indicates that the vortex has minimal impact on the internal pressure of the blade. Entropy production analysis reveals that turbulent entropy production dominates entropy production and that the maximum entropy production area is located in the vortex area and the boundary layer because these areas are characterized by strong flow deformation, which proves how vortex structure optimization can improve vortex pump performance. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Information. Author contributions K.L.: Conceptualization, writing-original draft, data curation. Y.H.: Writing-review and editing, funding acquisition, supervision. Q.J.: supervision. Y.T.: Conceptualization, validation. C.W.: Resources. J.G.: Resources. All authors have reviewed the manuscript and agreed to the published version of the manuscript. Funding Science and Technology Department of Sichuan Province (2025ZYD0188, 2022YFH0017, 2025YFHZ0255); Key Laboratory of Fluid and Power Machinery, Ministry of Education (LTDL2021-008). Data availability Data is provided within the manuscript or supplementary information files. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Information. Data Availability Statement Data is provided within the manuscript or supplementary information files.