| Centrifugal pumps have important applications in the national economy,the national defense,and the military industry,but the pump technology for some key devices is still controlled by others.The complex internal flow of a centrifugal pump is the key factor that restricts its hydraulic performance optimization and stability improvement.Deeply exploring the flow field characteristics in the pump is an important foundation for improving the hydraulic performance and flow stability of the centrifugal pump.Based on this,the complex flow field characteristics and hydraulic performance optimization in the centrifugal pump from the perspective of flow field decomposition were studied in this paper.The characteristics of complex flow field in the centrifugal pump were analyzed by proper orthogonal decomposition(POD)and dynamic mode decomposition(DMD)methods.An adaptive POD surrogate model based on the clustering algorithm was proposed,and the blade inverse design and performance optimization of the centrifugal pump were carried out based on this model.The knowledge mining research on the blade shape and flow field after multi-objective optimization of the centrifugal pump was conducted.The main work and research results are as follows:1.Based on the large eddy simulation(LES)numerical calculation results of the internal flow in the centrifugal pump,the dynamic modal analysis of the flow field characteristics for the unsteady flow in the impeller was performed.The results show that the DMD method can extract the main flow field structures in the impeller,and decompose its complex flow field into flow field characteristics with different energies and frequencies,including basic mode characteristics,dynamic mode flow field characteristics of rotor-stator interaction and its high-order harmonic behavior.The basic mode can reflect the high-speed jet region and low-speed wake region at the impeller outlet and the flow separation region on the blade suction side caused by the geometry of flow passage.The dynamic modes reflect the flow field characteristics of flow separation and unstable vortex structure shedding and dissipation caused by its rotation and volute interference in the impeller.The DMD method can effectively approximate the dominant flow field structure in the impeller in low dimensions,and clearly analyze the complex transient flow field characteristics in the impeller of the centrifugal pump.2.The complex transient flow field in the volute was decomposed by POD and DMD respectively,and the main coherent structures and flow field characteristics were analyzed.The results show that the transient flow field in the volute is decomposed into flow structures with different frequencies by the DMD method.Under the design flow condition,the DMD method obtains the flow field characteristics of rotor-stator interaction and its high-order harmonic behavior reflecting the wake flow structure and its shedding caused by blade rotation at the volute inlet.Under the low flow rate condition,the flow field characteristics of low-frequency in volute are further obtained by the DMD method.The flow field in the volute is decomposed into time-varying flow structures with different energy levels by the POD method.Under the design flow condition,the lower order modes reflect the large scale wake flow structure in the volute,and the higher order modes reflect the small scale wake flow field structure.Under the low flow rate condition,the higher order modes contain multiple frequencies with higher amplitudes,and the flow field structure represents the mixed superposition of flow field characteristics with different scales.Both POD and DMD can capture the flow field characteristics of rotor-stator interaction and its high-order harmonic behavior in the volute.Compared with the POD method,DMD can completely decouple the flow structure of each frequency characteristic in the flow field and judge the stability of different frequency characteristics according to the mode eigenvalues,providing certain theoretical support for the improvement of the stability of the inner flow field in the centrifugal pump.3.The proper orthogonal decomposition-radial basis function(POD-RBF)hybrid model method was introduced to reconstruct and analyze the flow field in the centrifugal pump impeller.Compared with the Gappy POD method based on linear regression,the POD-RBF model based on nonlinear regression has higher fitting accuracy and convergence stability,and its prediction accuracy gradually improves with the increase of the number of samples.The reconstruction results show that the POD-RBF method can accurately reconstruct the flow field in the impeller.The root mean square error of pressure prediction is 0.49%,the root mean square error of relative velocity prediction is 0.85%,and the time required for flow field prediction is only 1/240 of that calculated by CFD.The POD base modal analysis was carried out on the sample set,and the characteristics of the flow field and energy distribution were analyzed.4.An adaptive POD surrogate model based on the K-means clustering algorithm was proposed,and on this basis,the inverse design and performance optimization of the centrifugal pump blade were carried out.The results show that compared with the traditional fixed sample POD method,the adaptive POD method based on the clustering algorithm has higher flow field prediction accuracy,and the root mean square error of pressure prediction in the middle section and meridional plane of the impeller is only 1.7%and 0.96%,respectively.Compared with the fixed sample POD method,the prediction accuracy of the adaptive method in blade inverse design is significantly improved.The adaptive POD method was used to predict the flow field under the perturbation of control parameters in the process of impeller optimization,which greatly decreases the computation cost and significantly improves the accuracy of flow field prediction.The optimization results show that the efficiency of the pump is improved at all flow rate conditions after optimization,by 1.27%under the design flow condition(Q_d)and 2.53%under 0.6Q_d.5.Multi-objective optimization of the centrifugal pump head and efficiency under the rated flow was carried out,and the POD method was used to conduct knowledge mining for the optimized Pareto frontier solution set.The results show that a certain constraint relationship exists between the head and efficiency of the centrifugal pump.The POD decomposition of the blade shape and flow field data of the Pareto solution set reveals the complex relationship among the impeller geometric flow channel,internal flow field,and hydraulic performance in the pump.It is found that a larger blade angle and smaller blade wrap angle increase the average kinetic energy in the impeller,resulting in a higher head;a smaller blade angle and larger blade wrap angle reduce the velocity gradient from blade pressure surface to suction surface,thereby reducing the hydraulic loss and improving the efficiency of centrifugal pump. |
