In this master's thesis we discuss the use of a meta model on a parameterized numerical model of a double suction centrifugal pump. Compared to numerical modelling of machines which become enforced with each passing day, experimental testing of machines is expensive as well as time consuming. Combined with numerical calculations a state of the art approach with a meta model can be applied. A state of the art approach with a meta model covers various techniques of statistical and mathematical methods. The input parameters of numerical model may be subject to change. Using a correct sampling method for simulations and an approximation of the response within the observed region, an observed response is obtained. The response may also serve as a base for optimization methods in order to find the global optimum in the response. In this case, the input geometric parameters of the outlet casing pump have been changed in the prescribed manner and so the flow condition response in wide range of pump specific speeds are obtained. The results were compared to the literature and showed a high degree of compliance. Simultaneously, We have analyzed the remaining hydraulic conditions in this segment of the pump. In the results we have presented the optimal geometry of the outlet casing, which was obtained using an optimization method based on genetic algorithms.
|