Angelina Lee

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

Module 8 Questions:

  1. 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.
    • objectives define what success looks like in the dataset and directly impact which solutions are promoted by solvers like “optimize”
    • clarifying design objectives helps guide the generative process toward meaningful solutions
    • different objectives prioritize different design features
    • ex) prioritizing compact footprint within a restricted site may lead to tall, narrow buildings, while maximizing solar access may generate stepped or terraced forms with a larger building perimeter
  2. 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?
    1. image
    2. identifying target taxonomies (like above) structures the dataset around specific morphological goals
    3. clear taxonomies ensure coverage across design variations, avoiding overrepresentation of a single form
    4. helps manage large datasets by organizing options into defined families, making comparison and filtering easier
    5. supports more accurate training and testing if datasets are used for AI and ML applications
  3. 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:
      • saves time and reduce human error in generating thousands of design options
      • enables exploration of vast design spaces and support unbiased sampling across parameters
      • does not have IP barriers and issues regarding different file formats if data sourced from various firms
    • challenges:
      • maintaining parameter logic
      • debugging node flows in tools like Dynamo
      • controlling data volume to be fully encompassing but also not computationally too much in terms of time and money and energy
    • modularity:
      • structuring scripts into reusable components
    • scalability:
      • parametric flexibility and scripting practices that avoid hardcoding and support batch runs
  4. 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?
    • iteration allows designers to test assumptions, refine objectives, and narrow down solution space based on feedback
    • tweaking inputs or objective weights can significantly alter the generated results and lead to improved performance and aesthetics
    • iterative runs help discover edge cases, resolve conflicting goals, and gradually steer the system toward optimal outcomes
    • initial runs may miss better designs hidden in underexplored parameter zones

Questions Related to the Autodesk Class:

  1. 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?
    1. image

      Within the Dynamo and gen design sections above, the sub workflows are (note the main bullet points are taken directly from the autodesk class/handout, but I expanded on them in the sub bullet points):

    2. define the design variables
      • identify all geometric and performance related parameters that influence the building morphology (eg floor height, width, depth, number of floors, ratios, etc)
    3. define the design objectives
      • clarify the goals of the study (eg maximize daylight access, minimize energy use, optimize floor area ratio, etc)
    4. assign variables as inputs
      • in Dynamo, create sliders or inputs for each variable
      • ensure they’re exposed to the gen design interface using “Is Input” nodes
    5. assign objectives as outputs
      • define performance metrics in Dynamo (eg total surface area, solar exposure, walk distances)
      • expose these outputs to the gen design using “Is Output” nodes
    6. choose an appropriate solver
      • select from random, cross-product, optimize, or generative solver based on your priorities for diversity or performance
      • solver determines how input combinations are sampled and how the solution space is explored
    7. generate design options
      • run the study using the selected solver
      • let Revit and Dynamo work together to produce a wide range of building massing options
    8. explore generated options
      • analyze design variants using gen design’s built in filters, scatter plots, and sliders
      • look for tradeoffs and high performing solutions within the dataset
    9. export the desired option
      • select and send the preferred design variant back to the Revit model for refinement
    10. tweak the study and repeat iteratively
      • adjust input ranges, output definitions, or solver choice to refine results
      • rerun the study as needed to converge on more optimal or context-sensitive solutions
  2. 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.
    • large
      • automate generation of hundreds or thousands of massing options via Dynamo script runs
    • labeled
      • tag each design variant with calculated performance metrics (eg daylight, surface area, etc)
    • parametric
      • ensure all geometry is controlled by adjustable parameters (height, width, setbacks, etc)
    • expandable
      • design script to allow easy addition of new parameters or metrics as the study evolves
    • diverse
      • use wide parameter ranges and multiple solvers to cover a broad design spectrum
    • balanced
      • avoid overrepresentation of certain outcomes by using randomized input generation and balanced sampling techniques
  3. 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
      • assigns random values to inputs to generate diverse options
      • good for broad exploration and early-stage optioneering
      • image
    • cross product
      • combines all possible input values
      • ideal for full design space coverage and combinatorial studies
      • can create very large datasets which is also computationally expensive
      • image
    • like-this
      • makes small variations around a chosen design
      • useful for sensitivity analysis and local refinement
      • limits diversity but deepens insight into a specific region
      • image
    • optimize
      • iteratively improves designs based on defined goals
      • best when objectives are measurable in Dynamo
      • produces focused, performance-driven options
      • image
  4. 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?
    • solver influence and design diversity insights
      • random produces unpredictable and diverse results
      • cross product ensures completeness but often has very similar results iwth minor tweaks
      • like-this creates a lot of options that build off of an approved selection
      • optimize generates fewer but high-performing variants
    • practical leverage in Dynamo projects
      • use random or cross-product in earlier design stages to explore possibilities
      • transition to optimize or like-this when refining toward specific performance goals
      • mix solver strategies in phases to manage tradeoffs between creativity and efficiency
      sovler
      influence on design output
      design diversity
      use stage
      cross product
      explores all possible combos
      highest
      early exploration, dataset creation
      randomize
      generates many unique options through random sampling
      high
      optioneering, mid stage exploration
      optimize
      iteratively refines designs towards defined goals
      moderate to low (focused)
      performance tuning , later stages
      like-this
      varies one design to explore immediate neighborhood
      low (localized)
      refinement, sensitivity testing, latest stages