CFD Investigation on Resistance Balance of Complex Pipe Networks and Optimization Based on Artificial Neural Network

Abstract

Large-scale pipe network systems for dust removal are characterized by multiple dust sources, complex branching, and variable layouts. Resistance balance of the pipe network is critical for its efficient operation which is not easy to reach. Engineers often have to make their design without theoretical basis resulting in inevitable waste of energy. In this study, a simplified model is established and based on which the resistance characteristics of the pipe networks are numerically simulated using computational fluid dynamics (CFD). First, the local resistance coefficients of various bends are numerically investigated, focusing on the effects of geometry and operating conditions. Furthermore, a simplified model describing the resistance characteristics of the bend is established using a porous jump model. The simplified model is then used to simulate a laboratory pipe network with an improvement of 30% in computational efficiency. Finally, an actual engineering problem with a complex pipe network is simulated to analyze with its resistance imbalance problem optimized, and an artificial neural network model is also proposed to tune the resistance balance of the pipe network automatically. This study provides a significant reference for the simplification of pipe fittings in pipe networks, and also facilitates the accurate design and stable operation of pipe network systems.