Embedded Files
Department of Architecture, Faculty of ArchitectureNaresuan University
ARCH 655: Parametric Modeling in Design
ARCH 655: Parametric Modeling in Design
Professor Dr. Wei Yan
Thanasarn Changnawa
Department of Architecture, Texas A&M University
PROJECT 1 : Smithsonian Institution, Washington, D.C.
PROJECT 1 : Smithsonian Institution, Washington, D.C.
SURFACE POINT
Project Description
Project Description
Parametric modeling in this project
First step, points are created by using SURFACE POINT component to control the design changing model.
Second step, POINT DEFORM component is used to transform the shape of the roof geometry. In this step, NUMBER SLIDER component and LIST ITEM component are used to control the design form in XY-axis and Z-axis.
Third step, MESH SURFACE component and DIAGRID STRUCTURE component are used to create diagonal grid and points for creating structure frame and glass panels.
The final step, BOX MORPH component is used to create glass panels in the roof surface.
The Original Design
The Original Design
In the original design, building elements such as roof, glass panels, column, bench and floor are created. For the roof shape, three smooth curves are located in the central of the roof.
Changing Design parameters
Changing Design parameters
By changing some parameters, the original design model can be changed to the new design. The new design is based on the design method of the original one.
Physically-based Model
Physically-based Model
KANGAROO component is used to generate the physically-based modeling. In this step, DYNAMIC WEIGHT 1D, UNARY LOAD, LENGTH (LINE) and ANCHOR are added to generate the model.
Analyses
Analyses
Curvature analysis is used to explain the differences of the length in Z-axis of each point. In this step, the circles in the 3D shape are clearly presented.
Parametric Model
Parametric Model
visual programming
visual programming
PROJECT 2 : HUMAN + MACHINE
PROJECT 2 : HUMAN + MACHINE
Project Description
Project Description
In the project 1, the parametric model was created by SURFACE POINTS component and controlled by POINT DEFORM component. In case of NURBS surface design manipulation, Z-axis parameters (NUMBER SLIDERS) were used in controlling the shape design. It is a flexible and simple method for manipulating the parameters in form finding. However, a limitation of this method is time consuming in the process of generative modeling.
For the project 2, human logics, generative design algorithm and genetic algorithm (GA) were utilized for extending the performance of form finding. Using GA enables us to achieve the design objective while using generative design algorithm can help creating unlimited design solutions and ideas. Moreover, the human logics is beneficial for us to be more creative in the process of generative design.
- Previous limitation
The limitation of using human manual manipulation is time consuming in the process of form finding.
2. Design Rules
2. Design Rules
In this project, the human design rules were set by six major curves which were spread on the roof surface. These rules were used in generating new designs.
3. Strength combination
3. Strength combination
Each designer has their own strengths. The strength of each designer (human, generative design algorithm, and GA) was used in creating a creative design generator.
4. Design objective
4. Design objective
In generating design model, the design objective needs to be determined at the beginning. After that, GA was used to run a prototype according the pre-specified objective.
5. Visual programing
5. Visual programing
In this project,
Galapagos was used as a genetic algorithm tool in generating design models.
Honey Bee and Lady Bug were used in creating the design objective. The objective of this project was to maximize the solar radiation.
Machine Learning (ML) was used in clustering the results. In this case, K-means clustering was applied to identify the performance in solar radiation of each design solution.
6. Input Data to ML
6. Input Data to ML
The results from solar radiation analysis were recorded and then transferred to ML.
7. K-meanS clustering
7. K-meanS clustering
The results were categorized into three groups (low=0, medium=1, high=2) by using K-means clustering.
8. Solar Radiation
8. Solar Radiation
The number of designs generating by GA has no limit. However, due to the time limitation, 837 designs were generated within 30 minutes. Each design was analyzed with solar radiation.
9. Check performance
9. Check performance
In this case, the most appropriate design that met the pre-specified objective was selected (design number 837, cluster 1). It was further used to develop in details.
Perspective
Perspective
Design number 837, cluster 1
Interior
Interior
Design number 837, cluster 1
Texas A&M UniversityDepartment of Architecture
ธนสาร ช่างนาวา ( Thanasarn Changnawa )
ธนสาร ช่างนาวา ( Thanasarn Changnawa )
Page updated
Report abuse