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. As explained in the video, setting your objectives will influence the type of solutions a generative study will produce. If we are optimizing for size vs. energy efficiency vs. cost, the node logic is going to drive different outcomes based on those objectives. That is why it is so important to set objectives early (as explained in the video’s methodology) to set the study up for success in terms of outcomes aligned with user/designer needs. More specifically, if I am designing a study that I want to optimize for cost alone, my study will be biased to generate solutions that are smaller, use less material, and have simplier structural/material demands. On the other hand, if I am designing a study to optimize for energy efficiency, my designs are going to minimize surface solar exposure, increase envelope insulation, and maximize roof area to meet the desired outcome criteria.
- 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? Targeted taxonomies are essential for synthetic datasets because it allows us to streamline the inputs/outputs necessary for different types of buildings. For instance, an L-shaped building is going to have width/width/height inputs for all 6 sides of the footprint. More neo-classical designs like domes are going to have data for radii, heght, as well as a base lenth/width for the footprint. Organizing inputs/outputs based on similar designs allows us to have more variety in geometry while also simplifying the flow of data within the model to be somewhat uniform and more easily accessible within the study logic. It can also help narrow designs that are better fits for the respective project (i.e., some geometries will be infeasible/undesireable, so they should be explicitly excluded from the study).
- 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? Some of the benefits for automated workflows include the ability to generate more diverse synthetic data sets include more diversity in outcomes and a better-trained model that can respond more effectively to a breadth of prompts. However, complex workflows can increase the processing time in generating designs and may struggle to respond to the fast-paced needs of users in the earlier stages of design. Modularity is helpful in organizing workflows and making them reproducable for similar tasks (i.e., replicating similar logic for different taxonomies). Similarly, scalability is an important metric to consider it limiting the scope of workflows so that they do not overwhelm the algorithm and lead to unreasonably long processing times. Considering these factors can help constrain automated workflows so that they both meet the diverse needs/possibilities of a range of building typologies and deliver effective solutions within the timeframe required by users/designers.
- 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? As I learned from last week’s assignment, the initial settings of a generative study may not lead to the desired outcomes. For instance, they may lead to implausible or unrealistic design options that prompt you to edit the original code to make sure that those unanticipated bounds are not crossed again. This can lead us to better design alternatives within a data set, not only from tweaking bugs alone, but also from fine-tuning the study setitngs (i.e., # of iterations/paramater variations) to find the ideal solutions/# of comparisons to give use valuable insights.
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?
The general workflow starts with setting design variables or inputs that will change throughout the study while establishing a rubirc/value-system for how their performance will be judged. This could be minimizing cost, maximizing energy efficiency, etc. The next key step is creating an algorithm that will solve for these idealized conditions with the given inputs. The process then repeats itself as studies are generated and appropirate adjustments are made. With every iteration of the study, the logic gets more and more refined so that the outcomes are further optimized to the objectives. It is unexpected, and even discouraged, for the study to be “perfect” the first time since the iterative process is part of finding the most efficient and effective mode of analysis.
- 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. To apply these principles to a particular study, I would first identify the geometries essential in understanding the morphology (rectangular, triangular, radial, polygon, etc) and think about the types and # of inputs required to generate the morphology. Then, I would sort and sequence these geometries (as needed) to develop a logic that would take the necessary inputs to create potential output to meet project objectives. I would test the logic in multiple iterations to look for bugs and diversify the synthetic data set as needed. Are my initial runs generating the outcomes I am expecting? How might they differ? What can I change to get more robust, diverse, and applicable outcomes? I would use these guiding questions as a part of the methodology to revise and refine my logic to ensure that my study can be aligned with its goals.
- 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? As defined in the lecture, solvers are “tools that can run automatically a script many times.” In Generative Design Revit, the 4 solvers we can choose from are:
- Randomize
- Cross Product
- Like-this
- Optimize
Randomize will generate a fixed number of design alternatives based on a random combination of inputs within the set parameters. This solver would help create robust and diverse data sets, but not as targeted for particular outcomes.
On the other hand, cross-product will allow use to create an exaustive result for a design space within a range of inputs, which is most helpful when trying to find solutions witihin combinations of eachother. This is helpful in creating a robust data set, but there will likely be some repeat/similar solutions.
Like-this creates similar versions of a set input which is attuned to sentitivity-type analyses to find the best alternatives within slight variations of one another. This would help in creating a targeted data set, but the alternatives created will likely be very similar.
Finally, optimize creates designs based on outputs, using each previous iterations’s inputs as information to find a further optimized solution, which is most helpful when the Dynamo function can identify a quantifiable metric that we want to be optimzied/solved for (i.e., solar performance). This method would likely create a robust data set through many iterations, with more diverse alternatives in the earlier stages, and more targeted options later on. This might also be the most computationally extensive solver and could potentially take longer/more power.