MODELING WIND SPEED PROBABILITY DISTRIBUTIONS AT THE PEDESTRIAN LEVEL AROUND AN ISOLATED BUILDING BASED ON MIXTURE DISTRIBUTION FUNCTION

DOI
  • GAO Yishuai
    Interdisciplinary Graduate School of Engineering Sciences, Kyushu University
  • WANG Wei
    Faculty of Engineering Sciences, Kyushu University
  • IKEGAYA Naoki
    Faculty of Engineering Sciences, Kyushu University

Description

The pedestrian-level wind environment plays an important role for human safety and comfort. Although previous studies proposed several statistical models to determine probability distribution function (PDF) of wind speed based on statistics, the accuracy still needs to be improved. This study proposed an updated framework for modelling the PDF based on the mixture distribution function. The database of an isolated building case was generated by large-eddy simulations (LES) and validated with wind tunnel experiment. Three empirical distribution functions (i.e., two-parameter and three-parameter Weibull distribution (2W and 3W), and a mixture of two 2W (2W2W)) were analyzed. Two distribution parameter estimation methods (i.e., moment method (MM) and maximum likelihood method (ML)) were evaluated. It was found ML and MM methods within the 2W2W framework closely match the probability density functions obtained from LES, while the 2W and 3W methods provide useful insights but with slightly less accuracy. Both ML and MM methods achieve high accuracy at most locations, but the ML method is more stable. Conversely, if feasibility and lightweight implementation are prioritized, the MM method is recommended over the ML method within the 2W2W framework due to the extensive time-series data required by the ML method. These findings offer valuable insights for improving building design and urban planning to better handle and mitigate wind-related hazards.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top