A Performance based approach for the prediction of indoor thermal comfort of naturally ventilated buildings in the tropics

Abstract

Natural ventilation is a key strategy for achieving energy-efficient and thermally com- fortable indoor environments, particularly in tropical climates. This thesis presents a comprehensive framework for evaluating and optimizing natural ventilation perfor- mance in buildings by integrating computational modeling, optimization techniques, and empirical validation. The research focuses on assessing airflow distribution and thermal comfort while developing a performance-based approach to guide the design of naturally ventilated buildings. The study begins by investigating the impact of various Reynolds-Averaged Navier- Stokes turbulence models on airflow distribution and thermal comfort prediction using computational fluid dynamics (CFD). Among the tested models, the Standard k-ε tur- bulence model demonstrated the highest accuracy in predicting indoor airflow fields and thermal comfort indices. Validation of computational results against wind tunnel experiments revealed that existing turbulence models have limitations in capturing transient flow phenomena, such as Kelvin-Helmholtz instabilities, highlighting the need for improved model selection strategies. The deviation in the predicted thermally comfortable area for the Renormalization Group k-ε and Standard k-ω models was within ±5.0% of the Standard k-ε model, while the Realizable k-ε model showed de- viations within the same range only in horizontal planes at y/H = 0.375 and 0.5. Addi- tionally, the incoming jet angle was found to be within 17.1°-23.7°, with the Renor- malization Group k-ε model exhibiting the highest value (23.7°) and the Rk-ε model the lowest (17.1°). Building on these findings, an optimization framework was developed using Artificial Neural Networks and Central Composite Design to refine turbulence model coeffi- cients. This novel integration significantly enhanced prediction accuracy and compu- tational efficiency. The fraction of CFD predictions falling within a factor of two of the corresponding wind tunnel measurements improved from 0.68 to 0.84 for velocity predictions and from 0.48 to 0.81 for kinetic energy predictions, confirming the effec- tiveness of the optimized model. When applied to a real residential building in Sri Lanka, the optimized turbulence model demonstrated strong agreement with validation metrics across multiple ventilation scenarios. The factor of acceptance increased to 0.92, indicating that 92% of CFD predictions fell within a factor of two of the meas- ured wind speeds. The correlation coefficient of 0.89 confirmed strong agreement be- tween simulated and experimental values, while the normalized mean squared error of 0.35 and fractional bias of -0.06 highlighted a slight underprediction tendency, ensur- ing a conservative yet reliable model. These results validate the effectiveness of the CFD framework in accurately capturing indoor airflow distribution and guiding ven- tilation design decisions. A key contribution of this research is the development of a performance-based guide- line for evaluating natural ventilation potential and thermal comfort during the building design phase. The guideline incorporates site-specific climatic data, Predicted Mean Vote based thermal comfort assessments, and systematic airflow evaluations to optimize building orientation and the placement of ventilation openings. The findings underscore the influence of wind direction, opening configurations, and building ori- entation on ventilation performance, providing actionable insights for architects and engineers. This research also addresses critical gaps in computational model validation, empha- sizing the need for distributed and three-dimensional measurement points in wind tun- nel experiments. Additionally, it highlights the importance of non-isothermal condi- tions, internal layouts, and surrounding structures in future studies to enhance the ap- plicability of the developed methodology. The proposed framework offers practical tools for designing energy-efficient, natu- rally ventilated buildings, aligning with global sustainability goals. By bridging com- putational simulations, optimization techniques, and real-world applications, this the- sis establishes a robust foundation for advancing natural ventilation studies and con- tributing to the development of thermally comfortable and sustainable built environ- ments.