Structural optimization of multistage depressurization sleeve of axial flow control valve based on Stacking integrated learning

• Published in: Scientific reports Vol. 14; no. 1; p. 7459
• Main Authors: Li, Shuxun, Deng, Guolong, Hu, Yinggang, Yu, Mengyao, Ma, Tingqian
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 29-03-2024; Nature Publishing Group; Nature Portfolio
• Subjects: 639/166; 639/166/988; Axial flow control valve; Design; Flow characteristics; Flow control; Fluid dynamics; Fusion model; Humanities and Social Sciences; Hydrodynamics; multidisciplinary; Noise reduction; Optimization; Parameter optimization; Particle swarm optimization; Science; Science (multidisciplinary); Sound pressure; Working conditions; Fusion model; Parameter optimization; Particle swarm optimization; Axial flow control valve
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-58178-5

Due to the requirements of the working environment, the marine axial flow control valve needs to reduce the noise as much as possible while ensuring the flow capacity to meet the requirements. To improve the noise reduction effect of the axial flow control valve, this paper proposes a Stacking integrated learning combined with particle swarm optimization (PSO) method to optimize a multi-stage step-down sleeve of the axial flow control valve. The liquid dynamic noise and flow value of the axial flow control valve are predicted by computational fluid dynamics. Based on the preliminary evaluation of its performance, the structural parameters of the multi-stage pressure-reducing sleeve are parameterized by three-dimensional modeling software. The range of design variables is constrained to form the design space, and the design space is sampled by the optimal Latin hypercube method to form the sample space. An automated solution platform is built to solve noise and flow values under different structural parameters. The Stacking method is used to fuse the three base learners of decision tree regression, Kriging, and support vector regression to obtain a structural optimization fusion model with better prediction accuracy, and the accuracy of the fusion model is evaluated by three different error metrics of coefficient of determination (
R
2
), Root Mean Squared Error, and Mean Absolute Error. Then the PSO particle swarm optimization algorithm is used to optimize the fusion model to obtain the optimal structural parameter combination. The optimized multi-stage depressurization structure parameters are as follows: hole diameter
t
1
 = 3.8 mm, hole spacing
t
2
 = 1 mm, hole drawing angle
t
3
 = 6.4°, hole depth
t
4
 = 3.4 mm, and two-layer throttling sleeve spacing
t
5
 = 4 mm. The results show that the peak sound pressure level of the noise before and after optimization is 91.32 dB(A) and 78.2 dB(A), respectively, which is about 14.4% lower than that before optimization. The optimized flow characteristic curve still maintains the percentage flow characteristic and meets the requirement of flow capacity
K
v
 ≥ 60 at the maximum opening. The optimization method provides a reference for the structural optimization of the axial flow control valve.