Abstract

This paper presents a data-driven model of pressure drop in spiral wound reverse osmosis membrane channels with the aid of computational fluid dynamics (CFD) simulations and multilayer artificial neural networks (MLN) method. Typical industrial spacer structures are considered in the CFD models to ensure practicability. To generate adequate data for MLN training, three-dimensional (3D) CFD model problems with respect to 500 different design parameter combinations are solved with a high-performance computing (HPC) strategy. With the HPC strategy the efficiency of CFD simulations is improved by about 50 times. The proposed MLN method is then used to correlate the pressure drop considering 7 design parameters, which enables a quantitative description of geometric structures and operation conditions for improvement. The accuracy of the MLN prediction on the test set is 99.734%. The obtained results show that the performance of the predictive model is comparable to alternative approaches in previous literatures. The proposed method has potential applications in intricate PDE-constrained optimization problems involving CFD simulations due to its high computational efficiency.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call