Multiple Performance Optimization of a Single Stage Centrifugal Pump using an Intelligent System Approach

This study aims the use of an intelligent system to analyze and enhance the performance of a single stage centrifugal pump (SSCP) with respect to the blade number of impeller, which is crucial for the centrifugal pump design. In general, maximizing the efficiency (h) is the most common performance cost. Moreover, maximizing the pump head (H) and minimizing the power (P) are the other important criteria that should be considered for the optimal blade number. These goals are not simultaneously considered usually in the studies on this topic. The motivation of this study is to expand the system management perspective by evaluating the performance of the system with these multiple criteria. The centrifugal pump design is a typical Multi-Objective Optimization (MOO) problem. The MOO approach consists of five decision variables and three objective functions with different blade numbers for impellers. To determine the optimal solution of this MOO problem, the experimental data is initially expanded using a learning by examples methods based on an intelligent network model training. Then, extended data is used for calculating performance measures at each input-output pairs. In order to evaluate the results of the proposed approach on a real-time system, B50-200/100 type model pump with the speed of 2950 rpm and an outlet angle of 22.150 was tested at State Sugar Machine Plant, Eskisehir (Turkey). Since the main purpose is to determine the optimal blade number, different blade numbers has been used while other geometric parameters were kept constant. To determine the optimal solution, experimental data has used to train a selected soft computing model known as FWNN model. This model has advanced to express the relationship between the inlet and outlet values of the centrifugal pump. FWNN model achieves the general characteristics of the performance measure. The analysis with the use of FWNN model shows that, the optimum number of blades by considering the specified performance parameters for the centrifugal pump design is seven. Comparing with pump impeller with number of blades 5, 6, 8 and 9 increases in efficiency rates are 0.4%, 2.57%, 6.4% and 7.2%, respectively. The FWNN model over the performance analysis algorithm completes the missing data if exists and indicates the best performance solution as given.