Structure optimization of steam turbine cascade considering the non-equilibrium condensation using machine learning method

This paper proposes a novel optimization strategy for the structure of turbine cascades based on the theory of non-equilibrium condensation and genetic algorithms, aiming to effectively reduce the intensity of non-equilibrium condensation of steam within the cascades. A comparative analysis of numerical calculations and predictive results from genetic algorithms reveals that the error between the two is less than 1 %, validating the scientificity and practicality of the proposed optimization method. Further investigation into the parameters of the optimized cascade model and its non-equilibrium condensation characteristics reveals that the optimization strategy advances the steam expansion process to within the cascade channel. This change significantly reduces the maximum logarithmic condensation nucleation rate of steam (from 20.87 to 20.36). Meanwhile, the pre-expansion effect of steam effectively decreases the supercooling degree of steam, thereby inhibiting the growth of droplets (droplet radius reduced from 0.064 μm to 0.053 μm). Ultimately, the average mass flow rate of droplets at the outlet of the optimized cascades achieves a 12.47 % reduction. Notably, the optimized cascades exhibit good performance even under conditions of low expansion rates.