Modelling of Turbulent Flow and Heat Transfer for Building Ventilation

Sammanfattning: This thesis contributes to studies on the assessment of building ventilation performance and the development of turbulence models accounting for Low-Reynolds-number (LRN) effects and buoyant convection with heat transfer. Assessments of building ventilation performance are discussed with respect to indoor air distribution and passive contaminant dispersion. Different concepts and methods for analyzing and assessing ventilation flow systems are addressed and re-examined. Several new ventilation scales have been developed, including the local purging effectiveness, the expected contaminant dispersion index and the local specific contaminant-accumulating index. Approaches for numerically exploring these scales are presented. The purging flow rate is re-formulated in several expressions different from its original and previous descriptions. Some scales defined from this quantity are discussed. Using stochastic theory in conjunction with the compartmental method, a Markov chain model is proposed to determine the transfer probability needed to compute the regional purging flow rate. This model contains extra and useful information that is not included in previous deterministic analyses. The new scales and methods are expected to be applicable for diagnosing problems and optimizing designs of ventilation systems. The development of turbulence models based on the eddy viscosity concept is considered. For simulating turbulent recirculating flows, a comparison is made of the two-equation k-.epsilon. model, k-.omega. model and k-.tau. model. It is found that both the k-.omega. model and the k-.tau. model have relatively poor performance. Modifications are made for the k-.omega. model in which the model constants are re-established and the turbulent transport term in the .omega.-equation is re-modelled. On the basis of these modifications, a new LRN k.omega. model is developed in which the damping functions are re-devised and the near-wall asymptotic behaviour is emphasized. The mechanism for simulating transition is preserved in the modified model. The LRN formulation is further extended for analyzing buoyant-driven flows in enclosures at moderate Rayleigh numbers. The model behaviour accounting for transition onset in the boundary layer along the vertical side wall is discussed, and some remarks are made for the LRN formulation. The new LRN model shows promising improvements in the predictions. Large eddy simulation (LES) is implemented for turbulent convection flows with heat transfer. A modified subgrid-scale (SGS) buoyancy model is proposed, where the buoyant effect is explicitly accommodated in the SGS eddy viscosity/diffusivity formulation. The modification enables the model to avoid entailing no-real solutions for simulating thermal convection flows such as occurs in the original buoyancy model. Furthermore, the proposed model is able to account for some energy backscatter for flows with positive and significant thermal stratification. Comparisons and evaluations are made of several SGS models when applied to statistically stratified and unstratified buoyant flows. The performance of the SGS models is analyzed for natural convection boundary layer flows at moderate Rayleigh numbers, where laminar, transitional and fully developed turbulent flow features subsequently arise in the boundary layer. The behaviour of SGS models in accounting for energy backscatter is argued to be an essential ingredient for predicting natural transitional boundary layer flows. The failure and success of SGS models for handling this type of flow are analyzed and discussed.