[Paper Review] Predicting wind pressures around circular cylinders using machine learning techniques

Gang Hu, K.C.S. Kwok|arXiv (Cornell University)|Jan 21, 2019

Wind and Air Flow Studies77 references4 citations

TL;DR

This study applies gradient boosting regression trees (GBRT) to predict mean and fluctuating wind pressures around smooth circular cylinders using Reynolds number (Re), turbulence intensity (Ti), and circumferential angle as inputs. GBRT models achieved high accuracy across Re = 10⁴–10⁶ and Ti = 0–15%, offering a cost-effective alternative to wind tunnel testing and CFD simulations for engineering applications.

ABSTRACT

Numerous studies have been carried out to measure wind pressures around circular cylinders since the early 20th century due to its engineering significance. Consequently, a large amount of wind pressure data sets have accumulated, which presents an excellent opportunity for using machine learning (ML) techniques to train models to predict wind pressures around circular cylinders. Wind pressures around smooth circular cylinders are a function of mainly the Reynolds number (Re), turbulence intensity (Ti) of the incident wind, and circumferential angle of the cylinder. Considering these three parameters as the inputs, this study trained two ML models to predict mean and fluctuating pressures respectively. Three machine learning algorithms including decision tree regressor, random forest, and gradient boosting regression trees (GBRT) were tested. The GBRT models exhibited the best performance for predicting both mean and fluctuating pressures, and they are capable of making accurate predictions for Re ranging from 10^4 to 10^6 and Ti ranging from 0% to 15%. It is believed that the GBRT models provide very efficient and economical alternative to traditional wind tunnel tests and computational fluid dynamic simulations for determining wind pressures around smooth circular cylinders within the studied Re and Ti range.

Motivation & Objective

• To develop machine learning models that predict wind pressures around circular cylinders using existing experimental data.
• To evaluate the performance of multiple machine learning algorithms in capturing mean and fluctuating pressure distributions.
• To identify the most accurate and generalizable model for predicting wind pressures across a wide range of Reynolds numbers and turbulence intensities.
• To provide a computationally efficient alternative to traditional wind tunnel testing and CFD simulations for structural design.

Proposed method

• Three machine learning algorithms—decision tree regressor, random forest, and gradient boosting regression trees (GBRT)—were trained on experimental wind pressure data.
• Input features included Reynolds number (Re), turbulence intensity (Ti), and circumferential angle of the cylinder.
• The models were trained to predict both mean and fluctuating pressure coefficients separately.
• Model performance was evaluated using standard regression metrics across a wide range of Re (10⁴–10⁶) and Ti (0–15%).
• Hyperparameter tuning was applied to optimize GBRT performance.
• Generalization capability was assessed across the full range of Re and Ti in the dataset.

Experimental results

Research questions

• RQ1 Can machine learning models accurately predict mean and fluctuating wind pressures on circular cylinders using Re, Ti, and circumferential angle as inputs?
• RQ2 How do different machine learning algorithms compare in predicting wind pressure distributions?
• RQ3 What is the generalization performance of the best-performing model across the full range of Re and Ti?
• RQ4 Can the best model serve as a reliable alternative to wind tunnel testing and CFD simulations?

Key findings

• Gradient boosting regression trees (GBRT) outperformed both decision tree regressor and random forest in predicting both mean and fluctuating wind pressures.
• GBRT models achieved high accuracy across Reynolds numbers from 10⁴ to 10⁶ and turbulence intensities from 0% to 15%.
• The GBRT model demonstrated strong generalization capability, maintaining accuracy across the entire tested parameter range.
• The study confirms that GBRT models can serve as an efficient and economical alternative to traditional wind tunnel testing and CFD simulations.
• The trained models captured complex pressure distribution patterns, including separation points and peak pressures, with high fidelity.