Application of data-driven models to predict the dimensions of flow separation zone

Area of Science:

• Hydraulics and Fluid Mechanics
• Computational Fluid Dynamics
• Water Resources Engineering

Background:

• Flow separation zones (DFSZ) in intake channels can negatively impact hydraulic structure efficiency and sediment transport.
• Submerged vanes are a potential solution to mitigate adverse flow phenomena.
• Optimizing vane design and placement is crucial for effective DFSZ control.

Purpose of the Study:

• To investigate the impact of submerged multiple-vane systems on the dimensions of flow separation zones (DFSZ).
• To develop and compare data-driven models for predicting DFSZ.
• To identify optimal vane configurations for maximum DFSZ reduction.

Main Methods:

• Experimental measurements of DFSZ dimensions using 192 datasets.
• Vane design variations including curvature (β) and angle of attack (θ).
• Development and evaluation of Gene Expression Programming (GEP), Support Vector Regression (SVR), and a hybrid SVR-ACO model.

Main Results:

• A maximum DFSZ reduction of 78% (length) and 76% (width) was achieved with specific vane configurations (θ=30°, β=34°, δl=10 cm).
• The hybrid SVR(RBF)-ACO model demonstrated superior accuracy in predicting DFSZ, with Total Grades (TG) of 6.75 (length) and 5.8 (width).
• GEP-generated formulas for DFSZ showed good agreement with experimental data (0-10% error).

Conclusions:

• Submerged multiple-vane systems are effective in reducing DFSZ dimensions.
• The hybrid SVR-ACO model provides a precise and reliable tool for predicting DFSZ.
• Optimized vane design and placement are critical for hydraulic efficiency in intake channels.