Abstract
Airflow Network (AFN) models are suitable candidates for the coupling with building models for the evaluations of both internal airflows and infiltrations given the reduced time required for their computation. Furthermore, it is also possible to evaluate the concentration of pollutants such as the CO2 and the concentrations of virus, as, for example, COVID-19. While the computation of pollutants and virus concentrations leads to the possibility to evaluate both Indoor Air Quality and the risk of contagion in indoor environments, the implementation of AFN models in building simulations presents some issues. Indeed, building simulations that calculate airflows and infiltrations using airflow networks can drastically increase the computational effort required for each simulation. In addition, some of the parameters required to define an airflow network model are not always known and can lead to the calculation of unrealistic airflows. For this reason, a fast and reliable algorithm is needed for the optimisation process of both building and airflow network models. Therefore, this thesis presents a new efficient algorithm for the optimisation process of computationally expensive models. The new algorithm is also suitable for use with multi-objective optimisation problems by reducing the multi-objective space to a one-dimensional space, thus ensuring a higher accuracy with respect to global approximation models. The process is carried out using a probabilistic approach coupled with a local search-based method, ensuring both accuracy and efficiency while avoiding premature convergence of the process. The main objective is to reduce the computational time required for the optimisation process to be carried out, while increasing the quality of the solutions found. The new algorithm has therefore been tested on a series of multi-objective problems, and the results have been compared with other optimisation algorithms in terms of efficiency, efficacy and solution quality using several metrics. For the comparison process, the solutions of an integer optimisation problem obtained by a brute-force approach regarding the refurbishment of three simplified reference buildings were used as a reference for the evaluation of the algorithm's results. The possibility to calibrate AFN and building models with good accuracy and within a limited amount of time, open the opportunity to apply both models for the evaluation of the risk of contagion due to COVID-19. In fact, the outbreak of COVID-19 raised general awareness of importance of proper ventilation of indoor environments to reduce risk of infection. Particular attention has been paid to certain categories of buildings with high occupancy densities and more susceptible occupants, such as schools. Robust analyses of the effectiveness of different strategies to reduce the risk of infection have been difficult to conduct, despite the attention paid to classroom ventilation. In fact, the COVID-19 pandemic is still ongoing and it is difficult to fully quantify the impact of indoor ventilation by simply analysing the available monitoring data due to many factors, such as the presence of multiple virus strains, the use of face masks, the progress of vaccination, the installation of air purifiers and other disinfection devices. In addition, ventilation-related mitigation strategies are often dynamic, increasing the complexity of the problem to be assessed. In this research, a Monte Carlo model is integrated with building performance simulation to perform risk assessment of different case studies and ventilation strategies by considering dynamic boundary conditions. Airflow network models have also been integrated into the building performance simulations, extending the capabilities of the Monte Carlo model. The evaluation of airflows and infiltrations through the airflow network allows the risk assessment process to be carried out for more than one room at a time. The Monte Carlo model has also been improved several times to include COVID-19 variants, vaccines and air purifiers. Finally, the Monte Carlo model was applied to several case studies and used to evaluate different ventilation strategies in terms of risk reduction.