Elissa Irwanto

Journal Entry For
Module 8 - Gen Des and ML
ACC Folder Link
Link to Student

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.

Defining the design objectives in Revit’s Generative Design directly unveils algorithmic prioritization, producing diverse geometric outcomes. It also influences based on the objective how the algorithms evaluates design options. It is important to clearly identify the problem to set up an effective design study. By clearly setting up an objective, it frames the alternative generation and the trade-offs that are associated with these objectives.

For example, minimizing shading decreases window-to-wall ratios and east-to-west orientations to reduce solar heat gain. Meanwhile, maximizing view quality, it increases window-to-wall ratios and aesthetics, but it risks higher energy consumption, overheating, and decreases occupant thermal comfort. The tradeoffs results in certain places getting a better view, daylighting, and/or shade than others. The conflicting objectives are resolved using solvers such as Cross Product or Optimize to generate optimal solutions. By combining objectives, designers can explore tradeoffs and the design alternatives that best fits the project goals.

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?

Identifying target taxonomies datasets when generating synthetic datasets in architectural design is important because taxonomies classifies current architectural elements and conditions to enhance dataset management and quality. By sorting large datasets into clear categories, it can simplify analysis and reduce computational load allowing for an output to be produced even faster. This helps in ensuring dataset diversity by providing a variety of inputs and typologies that are all equally represented to prevent algorithmic bias. Lastly, having a large and unbiased dataset can help in ensuring accuracy in our design as structured taxonomies provides precise parametric control. This can make it easier for designers to compare and explore the high-quality data and make informed decisions.

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?

Benefits

  • A potential benefit of using automated workflows is that it can generate multiple design outputs rapidly and consistently as compared to inputting the data in a manual model.
  • Another benefit is that these solves can explore solutions and tradeoffs between a combination of outputs normally not discussed.

Challenges

  • A potential challenge of using automated workflow is that it is computationally demanding as high-level softwares are running.
  • Another challenge is ensuring target taxonomies are well-defined and not repeated so that results can significantly contribute to decision making.

Modularity can be achieved in such workflows by creating independent Dynamo scripts from each taxonomy. This keeps the workflow organized and will allow for it to be easier to spot mistakes. They can also reuse previous scripts as needed.

Scalability can be achieved by using the multi-objective Revit Generative Design solvers as they can auto-generate a multitude of design options based on varying input variables. Cloud computing can also be leveraged to provide the optimal solution based on large datasets using machine learning.

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 role of iterative processes is a vital step in optimizing design options within the generative design framework because it allows for the exploration of multiple solutions. It also allows for the inputs to be continuously refined based on feedback from previous iterations. The results of the previous iterations have flexible and adjustable inputs so that a vast amount of solutions can be explored. It is a feedback loop and provides improved outcomes from all the previous iterations. Iterative testing reveals overlooked constraints and expands the solution to wider possibilities. By repeating this process, the user is presented with a multitude of optimal, balanced, and unbiased solutions that meets the project’s objectives.

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?

General workflow of creating a generative study in Revit is:

  • Step 1: Define the design variables
  • Step 2: Define the design objectives and problem
  • Step 3: Assign variables as inputs (ensure diversity and scalability of inputs)
  • Step 4: Assign objectives as ranks/outputs and establish design objectives
  • Step 5: Choose an appropriate solver, which is an algorithm that automates the sampling of the input design variables. Moreover, the algorithm should move smartly through the design space to satisfy objectives and support different types of exploration
  • Step 6: Generate design options!
  • Step 7: Explore generated options against the specified objectives
  • Step 8 or 9: Export the desired option or tweak the study by adjusting the parameters and repeat this process iteratively.
  • This workflow allows for a deep dive into exploring the design alternatives. Designers can select an optimized solution that satisfies desired objectives and understands the trade-offs. This saves them a lot of time, money, and rework.

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 the target taxonomies
  • Write Dynamo scripts for generating each taxonomy category independently and parametrically
  • Define variables (such as height, length, width, indentation in the horizontal dimension, and indentation in the vertical dimension), ensuring that the dataset are large and have distinct geometries and typologies between two categories
  • Automate a categorical variable by giving a sampling of single set of inputs to represent all category combinations using solvers like Optimize or Randomize in Generative Design for Revit. It can also automate stimulations based on all outputs to generate a large and diverse dataset.
  • Ensure balance by maintaining an equal number of equivalent sampling across categories to avoid overrepresentation and potential biases. This can make sure that the dataset is well-suited for training models
  • Ensure that the logic is set up to be expandable so that new taxonomies and inputs can be easily added in the future
  • By linking parametric models to softwares, it includes ground truth labels relevant to the training task and can produce outputs to help evaluate the different design alternatives

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?

  • Randomize
    • Develops a large and diverse datasets with a wide variety of design option by assigning random values to input parameters across the different taxonomies
  • Cross Product
    • Creates a feature search with all parameter values. It is highly exhaustive and generates a comprehensive combinatorial analysis. It is key to creating a dataset that is unbiased.
  • Like-This
    • Explores a variation of form finding around a design that a user already likes. It is useful in sensitivity analysis datasets and has low diversity.
  • Optimize
    • Implements design optimization based on performance objectives. It continuously iterates and refines designs based on quantifiable performance goals to find the most optimal solution.

Reflect on the examples of building masses generated with different solvers from the class handout. What insights can you gain about the relationship between solver choice and design diversity? How would you leverage this understanding in a practical parametric design project using Dynamo?

The Randomize solver produces a diverse, broad, and unpredictable results. Cross Product yields diverse results based on its objective and specific parameter range. Like-This generates many variations of an already built design that the user approves of. Outputs are often similar to the design inputted. Optimize creates fewer solutions, but provides the highest performing solutions. Because the iteration uses input configuration from the previous generation to optimize the new design options, the dataset has limited diversity

I would leverage this understanding in a practical parametric design project by using a combination of these solvers at different stages of design. I would utilize the Randomize solver in the early design stages to explore the different possible options. Cross-Product solver can explore each and every possible combination available. By generating combinatorial datasets, I can further understand the design space. After I narrow down my design options and select one, I would utilize Like-This to adjust my current input configuration and explore the different variations possible. Then, I would use the Optimize solver to iteratively improve on solutions. Lastly, I would select the solution that most optimally satisfies the project’s objective.