Article Preview
TopIntroduction
Most cities in the world suffer from air pollution, due several many factors such as burning fuel, industry and release of chemicals (Kurt, 2016, Li, 2012, Li, 2017). Many studies have focused on reducing emissions of pollutants, with significant progress being made. So far, large part of the population in urban areas breathe air, that does not meet European standards nor the World Health Organisation Air Quality Guidelines (Kelly, 2015). Currently, there is no ready-to-use technology available for a sustainable removal of particulate matter (PM), Nitric Oxides (NOx), nor volatile organic compounds (VOCs), in an urban environment. The photocatalytic oxidation (PCO) has been the focus of increasing attention in recent years, to abate pollutants, with possible applications in several areas, including environmental and energy related areas. The Titanium dioxide (TiO2) used as photocatalysts, is almost the only material suitable in industry at present and also probably in the future (Paz, 2010; Mamaghani, 2017). The choice of TiO2 is based on the highest stability, low cost, and transparency to visible light and a highly efficient photoactivity (Ribeiro, 2013). PCO is particularly useful for volatile organic compounds (VOC’s), but according to the literature, the NOx can also be degraded to a lesser extent (to nitrogen). Furthermore, TiO2 is also known to degrade the organic fraction of particulate matter (black carbon, soot). The latter is proven by many papers evidencing the self-cleaning properties of TiO2 (Bianchi, 2015).
In the last decades, thanks to advances made in computational resources, numerical simulation approaches have become increasingly popular. Nowadays, simulations with Computational Fluid Dynamics (CFD) is frequently used to assess urban microclimate.
Several research in artificial neural networks (ANNs) show that ANNs have powerful pattern classification and pattern recognition capabilities and they are used in many fields. They have become well established as viable, multipurpose, robust computational methodologies with solid theoretic support and with strong potential to be effective in any discipline (Dayhoff, 2001). Inspired by the biological system, especially the sophisticated functionality of human brains where hundreds of billions of interconnected neurons process information in parallel (Wang, 2003). ANNs algorithms are able to learn and generalize from examples and experiences as they have the ability to capture functional relationships among the data, even if the relationships are hard to describe or they are unknown. The advantage of using ANNs is that they minimize the error compared to other forecasting methods, and they provide results that are approximately close to analytical values.