Modeling of hydrogen flow decompression from a storage by a two-stage Tesla valve: A hybrid approach of artificial neural network, response surface methodology, and genetic algorithm optimization

Abstract

Hydrogen, as an energy carrier, requires efficient handling and regulation, especially during storage and release phases. The efficient hydrogen release from the storage tanks is often coupled with a rapid decompression process. Pressure regulators, hydrogen sensors, and monitoring systems are commonly used to control the pressure of hydrogen gas in storage tanks and distribution systems. Besides the mentioned methods, Tesla valves offer a compelling solution due to their unique design and functionality. In the current research, a novel two-stage Tesla valve was designed to regulate the hydrogen flow in both forward and reverse directions. Artificial neural networks (ANN) and response surface methodology (RSM) were applied to provide the prediction models. Presenting these ready-made and pre-determined models can help researchers and developers of Tesla valves to achieve their goals in the shortest time and with the lowest cost. In this research, two different prediction models were employed to predict the diodicity (Di) and absolute pressure drop ratio (APDR) in the Tesla valve. The distance between stages (X), length of divider (L), valve angle (θ), and depth of divider penetration (H) are the input variables of the prediction models. These variables, as the most influential parameters, are carefully selected and manipulated within specified ranges. Moreover, single- and multi-objective optimizations were performed by genetic algorithm, and the Pareto fronts for Di and APDR were presented. For the first time, the combination of these three methods (ANN, RSM, and genetic algorithm) was employed herein to investigate the decompression performance and hydraulic behavior of Tesla valves. Based on the findings, the ANN models exhibited higher accuracy with R2 values of 0.991 relative to RSM models (for both responses). The optimized design significantly enhanced the pressure drop of hydrogen flow in the reverse direction, resulting in an improvement of about 379.5 % in diodicity. Besides, the novel two-staged Tesla valve enhanced the pressure drop by about 4830 % and 1354 % in reverse and forward directions compared to the simple straight channel.