Application of artificial neural networks (ANN) to evaluate centrifugal pump characteristics

Abstract. This paper uses an Artificial Neural Network (ANN) technique to give an experimental and comparative examination of centrifugal pump characteristics. Comprehensive physical testing is a common component of traditional pump performance evaluation techniques, which may be expensive and time-consuming. In this study, we created an ANN model to forecast important performance metrics, such as speed, torque, pressure based on input data. These metrics include flow discharge, height, hydraulic power, motor power, and efficiency. The ANN model was trained and validated using experimental data that came from a series of carefully monitored experiments conducted on a typical centrifugal pump. Using a Levenberg Marquardt backpropagation approach, the ANN model was trained to attain high prediction accuracy by refining the topology of the network found to be high R-squared value of 0.98 and low RMSE by effectively predicting important parameters. Next, in order to assess the ANN model's predictive power, its performance was contrasted with the outcomes of the experiments. The results show that there is a strong degree of correlation between the predicted and experimental data, indicating that the ANN technique offers a dependable and effective way to forecast centrifugal pump characteristics.