Estimation of pressure drop in two-phase flow using fully connected neural networks: A short communication

It is well known that the pressure drop estimation is crucial for optimizing flow systems in various industrial applications. Although several available conventional models, such as the semi-empirical model, have been used for estimating pressure drop for a two-phase flow, it is challenging to achieve accurate results owing to the complexity of two-phase flow. In the present work, we are considering the problem of predicting pressure drop in two-phase flow in pipes using fully connected neural networks (FCNNs). Motivated by experimental data published in the literature, the FCNN model has been trained and then the results have been predicted. It is worth noting that the used experimental dataset was collected for a horizontal flow loop with a length of 9.15 m and diameter of 0.0254m, conveying a two-phase flow system. A critical comparison between the performance of the present FCNN model and mechanistic model is discussed, where results demonstrate how the present model is pre-eminent to handle the estimation of pressure drop in two-phase flow over the mechanistic models. However, results show that the FCNN model accuracy is estimated at 76 %, where the model is superior for the high pressure drop case over the low pressure drop case. The present FCNN model might be applicable to similar systems, without the need for further experimental measurements. By online uploading out the Python FCNN codes, we hope to fast-track the readers in applying FCNN algorithm to their own problems.