Module 8 Questions:
- How does defining the design objectives influence the outcomes in a parametric design study using Generative Design in Revit? Provide examples of how different objectives might result in varied design options.
- Discuss the importance of identifying target taxonomies when generating synthetic datasets in architectural design. How can this help in managing large datasets and ensuring diversity and accuracy in your designs?
- What are the potential benefits and challenges of using automated workflows for generating diverse synthetic datasets in parametric design? How can modularity and scalability be achieved in such workflows?
- Explore the role of iterative processes in optimizing design options within the generative design framework. Why might it be necessary to tweak and repeat studies, and how can this approach lead to better design outcomes?
The design objectives can reduce objectives to different kinds of outputs. The objectives decide how the generative design optimizing the solution in specified direction.
For example, given a mass building similar to the class example, set objectives to maximize floor areas and minimize floor areas will gives totally different design options. For objective of maximizing the floor areas of the building, the generative design options will try to enlarge the radius or perimeter of the building floor or try to create more building floors. In opposition, if the design objective is to minimizing the floor areas, the generative design options will try to cut down the radius and perimeter of the each floors and increase the story height at the same time. By comparing generative designs of these objectives, the building shape will become much more different that maximum floor area building will have a fatter shape and minimum floor area building will get a slender shape.
Similar to defining the design objectives, identifying target taxonomies when generating synthetic datasets can help reduce the unnecessary design outputs and pinpoint the required synesthetic datasets by the user.
By identifying target taxonomies, it can help reducing the operating time when managing large dataset. Without identifying categories, the computational process need to go through all the datasets which is quiet time consuming. Additionally, it can also help identifying more solution in the category by using the same amount of time without identifying category. From perspective of accuracy, the output stuff produced based on specific category will provides higher accuracy compared to the random produced stuff.
Potential benefit of using automated workflows are higher efficiency procedure. Compared to manual workflows, the automated workflows can generate much more datasets which provide more possibilities for the user.
Challenges of using automated work is that it may hard to troubleshooting the problem if there are any mistakes need to fix inside a long and complexed automated workflows. Addition to that, some automated work may need even longer operation time compared to manual work for some small scale diverse synthetic dataset generation because automated work usually need complete cycle of data algorithms which takes long time for operating.
The modularity can be achieved by encapsulate each function in the workflows. Such kind of modularity will be helpful for debug in a large workflow or reusing of the module in the future for other flows. For scalability, the workflow need to have a complete and logic computation algorithm to deal with data and situation with different kinds of scales of data. At beginning, the function can be tested use a small scale of data. Such measure can save a large amount of time in the preliminary stage of testing functions.
The iterative process is the basic in optimizing design options. This is because the requirement of optimization usually can not be achieved at the first loop. The process of optimization need to have several rounds of optimization and making adjustment based on the latest optimization results.
This approach can get better design outcomes by optimizing the design system through loops of optimization logarithm. If the design objectives are set correctly, each time of optimization will make the result become better. On another hand, if you find your result is deviate from the original good measurement, it represents there are places in your algorithm that need to be revised.
Questions Related to the Autodesk Class:
- Describe the general workflow of creating a generative design study in Revit, as presented in the lecture. What are the key steps involved, and how do they contribute to the generation of optimized design options?
- Given the properties of a good synthetic dataset outlined in the class, such as being large, labeled, parametric, expandable, diverse, and balanced, how would you apply these principles to create a dataset for a specific building morphology study using Dynamo? Outline your approach.
- Identify and discuss the four different solvers mentioned in the lecture that can be used to generate building masses in the Generative Design tool. How do these solvers impact the sample space and variety of design options produced?
There are mainly four steps for the general workflow of creating a generative design study. First step is to define problem of the study by defining the input and output. Next, the study need to build a dataset of the study by implement generative algorithm, generate models and run simulations based on such algorithm. Then, we need to fit surrogate model to the dataset that we want to investigate. Lastly, we need to validate the model accuracy after several testing.
One key step include is to define the input and output of the study. By defining such variables, your system will become able to optimize based on such parameters. That data are good source and limit for optimization. The step 2 to 4 is a good process to create such algorithm to fit the data which want to investigated. Then such algorithms can be used to validate and pursue the optimized design options.
For the first step, I will define study problem as maximize the solar insolation of the building with lowest cost and the cost limit is 5 millions. After than, I will set my inputs as top height, top rotation and shapes of the building footprint. For output, I will define the solar insolation, cost per floor areas. Then, I will create generative algorithm and run it to see whether the result of simulation satisfies my requirement of cost limit. Finally, I will compare my simulation result to the requirement of the problem. If that doesn’t work, I will try to adjust the lower and upper limits of my input for achieving the design goals of the study.
Four different solvers mentioned in the lecture are randomize solver, cross product solver, like-this solver and optimize solver.
For the randomize solver, it will randomly generate different kinds of design options without any specified characters. Sample spaces for randomize solver will be helpful for design initialization. The user can catch the special character they want and applied to future generative design studies. The variation of design options will be large compared to other solvers.
For the cross-product solver, it will generate design options based on the specified feature. Sample spaces for cross-product solver will generate designs in similar format. In the example given by Autodesk, all generated designs have similar shape of four corners’ box. The variation of design options will be limited to the specified feature.
For the like-this solver, it will generate design options based on the selected shape. Sample spaces for like-this solver will generate designs with similar features to the selected shape. The variation of design options will be similar to the given designs with minor changes in different perspectives.
For the optimize solver, it will optimize design options based on performance of generated options we already have. Sample spaces for optimize solver will like a flow chart which shows the evolution of the design based on performance of previous designs. The variation of design options in each iteration of optimization will become smaller with more specified characters of the design.