Local Wind Control near the Wall Greening by using a Neural Network

抄録

This study focuses on generating and controlling air flow caused by temperature differences and increasing the greening rate in urban areas by means of biowalls. In this study, computational fluid dynamics (CFD) software and an artificial neural network (ANN) inverse model were used to study generating and controlling air flow. First, an ANN inverse model was trained and tested using the data obtained from the CFD simulation. Then, the trained ANN inverse model recommended greening patterns to generate the desired air flow. Finally, a model study was conducted under similar conditions on the greening patterns recommended by the ANN inverse model. The most highly recommended greening pattern was whole-greening, in which the average temperature of 35.5°C would generate ascending air flow at a rate of 0.3 m • s⁻¹. Wind velocity in the model study of a whole-greening pattern in which average temperature was 33.8°C, was 0.29 m • s⁻¹ which is close to the desired wind velocity in the ANN inverse model. This result shows that it is possible to generate and control air flow near bio-greening caused by temperature differences, and this method which used CFD simulation and ANN inverse model is applicable.