PREDICTION OF THE PARAMETERS RADIAL CENTRIFUGAL PUMPS WITH ARTIFICIAL NEURAL NETWORKS

In this study, an estimation of some structural parameters in radial centrifuge pumps was performed with Artificial Neural Networks having properties such as production of suitable outputs for defined system inputs of which the output values are unknown. For Artificial Neural Network, experimental parameter values were used. Therefore, an Artificial Neural Network model was established with the pump RPM, flow rate and manometer height parameters as 3 input data and impeller inlet diameter, impeller outlet diameter, impeller blade number, pump inlet pipe diameter, pump outlet pipe diameter as 5 output data. For Artificial Neural Network Model, Levenberg Marquart education algorithm which belongs to Multi-Layer Perceptron structure was used and the results obtained by Artificial Neural Network Model were compared with the results obtained by measurement, regression equations and theoretical calculations. It was observed that the results obtained by Artificial Neural Network were very compatible with the results of measuring.