Intelligent Facades Generation Scheme Considering Lighting and View Requirements
– Based on CNN and MLP Algorithms
xxx, xxx, xxx
Zhong Ximing, Li Kun
For the shops along the street on the second floor, lighting and view demands are both important factors affecting the quality of interior space.
Therefore, when designing the facade of the commercial street, it is necessary to consider the impact of the facade on indoor lighting and view. However, we found that the uniformly added grille facade after the renovation of the old vegetable market block, while ensuring the integrity of the street, did not meet the specific needs of each Second-floor shop for lighting and view, and appeared rigid and impractical. This design aims to use a convolutional neural network and multi-layer perceptron to help optimize the grille facade of the second-floor shops in the old vegetable market block driven by lighting and view requirements. On the basis of establishing the unified form of the grille, three iconic scenes outside the window are extracted through CNN (i) sky; (ii) City walls; (iii) Characteristic values of trees. First, a single shop is selected, and MLP is used to find the connection between the grille style and the sight perception and lighting condition. And try to use this method and MLP backpropagation aid to design grilles in similar blocks. The method is evaluated by calculating the lighting condition and line of sight evaluation of the generated grid. We believe that the method has strong generalization and robustness, and has strong application space in the reconstruction of historical blocks in Xi ‘an and other cities.
This project is an intelligent facade generation scheme that considers lighting and view requirements based on CNN and MLP algorithms. The design aims to use convolutional neural networks and multilayer perceptrons to help optimize the façade, driven by lighting and viewing requirements. The pilot site is Jianguanmen Old Vegetable Market, Xi’an, China.
The sampling method and analysis mode of the case for the landscape outside the window provide an important reference for our facade design based on sight requirements. They also analyze the landscape and investigate the possibility of its reaction on the building’s facade based on the landscape sampling outside the window. The method has strong generality and robustness and has a strong scope for application in the redevelopment of historic districts in Xi’an and other cities.
Methods and Evaluations
Based on the two basic requirements of the field of vision and daylighting, combined with the cost restrictions of each shop in the old vegetable market, a GH model is established to generate a field of vision evaluation, cost evaluation, and daylighting evaluation according to the location of control point, the radius of control point, grid size and grid Angle. The corresponding evaluation is generated by adjusting the above values to obtain the data set we need.
Because the construction of a curved grille is more difficult, it is obtained by multiplying the size of the grille and the radius of the control point by different weights respectively.
Field of vision evaluation:
The control points in different positions correspond to different landscapes, so the field of vision is determined by the position of the control points.
By using ladybugs and honeybees to simulate sunshine, the average illumination and evenness of daylighting in the room can be obtained. Basically all the controlling factors.
This research project is a series of outcomes from the AI architecture course launched by Archiford and Xi’an University of Architecture and Technology. The course aims to use deep learning methods to design and advance architectural design solutions.