Research on the parametric design of marine nuclear-powered turbine and the multi-objective intelligent optimization of aerothermodynamics performance under variable working conditions

Lei Zhang, G.b. Chen, Yindong Shi, Luotao XieResearch Square 2025 · DOI

TL;DR

Researchers have developed a multi-objective optimization framework for marine nuclear-powered turbines that significantly improves efficiency and reduces blade erosion. By combining 3rd-order Bezier curve parameterization with Kriging surrogate models and genetic algorithms, the team successfully optimized blade designs to perform better across both rated and overload operating conditions.

What Problem Does This Paper Address?

Marine nuclear-powered turbines face a unique challenge: they must operate reliably across a wide range of power levels (variable conditions) while dealing with "wet steam." In the final stages of a turbine, steam begins to condense into tiny water droplets. These droplets cause two major headaches:

  1. Efficiency Loss: The liquid phase doesn't contribute to work as effectively as the gas phase.
  2. Blade Erosion: High-velocity droplets strike the turbine blades, causing mechanical "pitting" and potential structural failure.

Most existing research optimizes turbine blades for a single "design point" (usually rated power). However, a ship rarely stays at one speed. This paper tackles the gap in multi-condition optimization, ensuring that the turbine remains efficient and safe even when pushed into "overload" conditions.

Methods

The authors proposed a sophisticated "intelligent optimization" pipeline:

  • Parameterized Reconstruction: Instead of treating blade shapes as arbitrary polygons, they used 3rd-order Bezier curves. This allowed them to describe complex, smooth blade profiles using only 12 variables (8 geometric parameters like inlet/outlet angles and 4 Bezier control points).
  • CFD and Wet Steam Modeling: They utilized Computational Fluid Dynamics (CFD) with a non-equilibrium condensation model (IAPWS-IF97) to simulate how wet steam actually behaves as it transitions from gas to liquid across the blades.
  • Surrogate Modeling (Kriging): High-fidelity CFD simulations are computationally expensive. To speed things up, the team used Latin Hypercube Sampling (LHS) to run a limited set of simulations and then built a Kriging model—essentially a mathematical "cheat sheet" that predicts performance for designs the CFD hasn't seen yet.
  • Optimization (NSGA-II): They employed the Non-dominated Sorting Genetic Algorithm II, a multi-objective evolutionary algorithm, to find the "Pareto frontier"—the set of designs that offer the best possible trade-offs between efficiency and moisture reduction.

Datasets

The study validated its methods using two primary models:

  1. Dykas Planar Cascade: A well-known benchmark in the field of steam turbine research, used here to verify the accuracy of the CFD and optimization logic against experimental data.
  2. 3D Turbine Cascade: A newly developed 3D model of a high-power-density marine turbine stage, including both static and moving blades (root and top sections).

Key Insights and Ideas

  • The Overload Advantage: One of the most striking findings was that the optimization yielded even greater benefits during "overload" conditions than at "rated" conditions. This suggests that standard designs are particularly poorly suited for high-load scenarios.
  • Sensitivity Analysis: The researchers discovered that the mounting angle of the blade had the most significant impact on humidity, while the Bezier control parameters and trailing-edge wedge angles were more influential on overall isentropic efficiency.
  • Shock Wave Mitigation: The optimized designs successfully reduced the strength of "dovetail-type" condensation shock waves at the trailing edge. By smoothing the flow transition, they lowered the local expansion rate, which directly inhibits the sudden "nucleation" (burst of droplet formation) that causes efficiency drops.

Key Results

The quantitative improvements were significant across the board:

For the Dykas Planar Cascade: * Isentropic Efficiency: Increased by 0.69% (rated) and 0.93% (overload). * Outlet Humidity: Decreased by 6.06% (rated) and 8.89% (overload). * Droplet Size: The maximum droplet diameter (MDD) decreased by up to 15.77%, drastically reducing the risk of blade erosion.

For the 3D Turbine Stage: * Stage Efficiency: Improved by 0.9% (rated) and 0.74% (overload). * Stage Power: Increased by 0.18% and 0.19% respectively, while maintaining a constant mass flow.

Limitations and Future Work

While the results are promising, the authors acknowledge the massive computational hurdle: training the Kriging model required a 400-core supercomputer. This level of resource might not be available to all design teams.

Future Work could involve:

  • Multi-physics Integration: Incorporating structural stress analysis directly into the optimization loop (rather than just as a constraint) to ensure the thinner, optimized blades can handle the centrifugal loads of a nuclear turbine.
  • Broader Operating Envelopes: Expanding the optimization to include "low-load" or "maneuvering" conditions to create a truly universal marine turbine blade.

Summary updated

Abstract

<title>Abstract</title> In order to improve the aerothermodynamics performance of the marine nuclear-powered turbine and control the exhaust humidity of the blade grid, thereby improving the operating efficiency and power of the turbine and ensuring safety. We propose a method for parameterized reconstruction of steam turbine blade profiles based on their geometric parameters using a coordinate equation developed based on the third-order Bezier curve. By combining the blade parameterized reconstruction method with a Kriging approximation model and a multi-objective genetic algorithm (GA), we developed an optimized system for thermodynamic performance in turbines. The optimization objective was the cascade core thermal parameters of steam turbine under multiple operating conditions. The design parameters were the geometric parameters of the parameterized blade profile. Based on the calculation results of wet steam non-equilibrium condensate flow of steam turbine, the optimization method and process of multi-objective thermodynamic performance of steam turbine blades based on Kriging model were proposed. Then, we executed parameterized reconstruction of a Dykas planar cascade and a steam turbine 3D cascade to achieve multi-parameter, multi-condition design optimization of planar and 3D cascades. Using the proposed novel method, the isentropic efficiency and stage power of steam turbine cascades at variable operating conditions were enhanced, thermodynamic parameters (e.g., velocity, temperature, and pressure) were more homogeneously distributed, and overload condition sites showed more significant improvements. Thus, the proposed method achieved multi-condition, multi-constraint thermodynamic performance design optimization of wet steam turbine blades, thereby providing insights for intelligent design optimization and operation of wet steam turbine cascades.
BibTeX
@inproceedings{zhang2025research,
  author = {Lei Zhang and G.B. Chen and Yindong Shi and Luotao Xie},
  title = {{Research on the parametric design of marine nuclear-powered turbine and the multi-objective intelligent optimization of aerothermodynamics performance under variable working conditions}},
  booktitle = {Research Square},
  year = {2025},
  doi = {10.21203/rs.3.rs-6236801/v1},
}
Figure