Applying machine learning for optimization of dehumidification strategy on the modified model for the non-equilibrium condensation in steam turbines

Abstract

In this study, the dehumidification strategy has been used to reduce the damages caused by the non-equilibrium condensation flow. In the first stage, the non-equilibrium condensation phenomenon modeling has been performed after a sensitivity grid size test, and validated by an Eulerian-Eulerian approach for the machine learning (ML) initial data. In the next step, the location of nozzle have been evaluated for the ML to predict the non- equilibrium phenomenon. Then, the output data of the numerical method have been used as the initial data for the ML method. Finally, after accurately ensuring the prediction of the ML from the behavior of the non-equilibrium condensing flow compared to the experimental data, the dehumidification technique has been linked to the ML. By help of the ML approach, a performance optimization has been carried out to reduce the presence of water droplets on different locations of the blade surface. In this way, the numerical simulations result by applying the ML have been linked and the final optimized model according to the defined parameters has been developed. It is noteworthy to mention that the objective functions are defined according to the energy of flow and the corrosion rate. The outcomes from such optimization could lead the designers to find the best location for suction dehumidification technique for the turbine blades. The machine learning has been linked to three optimization algorithms, including drone squadron optimization, particle swarm optimization, and genetic algorithm. According to the obtained results of the employed machine learning and optimization methodology, for the optimum case compared to the original case, 99%, 1%, 18%, and 90% reductions have been achieved for the erosion rate, entropy, loss of moisture, and droplet sizes, respectively. It means that, by performing such research one could study some optimization parameters and link them by artificial intelligence to investigate the best place for the suction dehumidification technique.