An Effective Model for Performance Prediction of a Centrifugal Pump with Nose Caps Using ANN

The centrifugal pump is to be cost-effective, efficient, reliable, and ensures safe operation at or near best efficiency point [BEP] and is heavily used in various day-to-day applications. Flow instabilities emerge in a centrifugal pump while operating outside of its design parameters. To reduce the flow instability, the conventional fastener for the impeller of a centrifugal pump is modified with the addition of nose caps. The authors created 12 different design profiles to minimize the flow randomness over the suction region and have witnessed a reduction in instability. The experimentation with different other design profiles will surely create a research analysis but this process consumes considerable time and resources. Computer programs based on machine learning (ML) have proven to employ algorithms to autonomously develop cost-effective solutions while saving a substantial amount of time thereby enhancing the systems' effectiveness, adaptability, and quality. The paper proposes an effective artificial neural network (ANN)-based network to predict the output parameters of centrifugal pumps for different nose caps and operating conditions. The model is trained using a dataset taken from the readings through experimentation and is validated using a test dataset. The performance of the model is validated using standard metrics like RMSE and R2 and is observed on analysis to be appreciable.