Chengtao Zhang

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. Defining the design objectives help set the goals to be completed and optimized, which guides the generative design process in evaluating the best solution. For example, when choosing the optimal design for a building using generative design, if the design objective is minimizing construction cost, the solution will lean towards smaller floor areas and common geometries. If the design objective is to minimize energy consumption, the solution will lean towards compact shapes with the least surface areas and external walls. If the design objective is to maximize natural sunglighting, the solution will lean towards shapes with more windows and higher surface area. Different design objectives highlight different performance criteria that would produce different design variations.
  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? Identifying target taxonomies enables data balance that helps prevent overrepresentation of a particular morphology and avoids bias in ML models. An organized taxonomy also allows for easier and more accurate dataset management in query, filterin, and labeling for specific scenarios. Target taxonomies also mean there will be a diverse coverage of design spaces and shapes (such as polygons, irregular shapes, L-shapes).
  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? Some benefits are faster speed in generating a lot of design instances with little manual input, more scalability across different projects and fields by adding and adjusting the parameters, as well as repeatability across projects with version tracking and consistent workflow. Some challenges are certain projects with high complexity and computational load that takes large amount of resources (time and space) to process, as well as ensuring that each generated solution meets the design constraints (repetitive to check for each one in case of computer error). Modularity and scalability can be achieved through creating custom nodes and group functions in Dynamo that can be applied to multiple projects, organizing workflow into discrete and testable modules, as well as storing data in JSON or CSV format for ML compatibility.
  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?

Iterative processes enable continuous refinement of objectives and constraints based on different stages and feedbacks of a project. Tweaking and repeating studies help narrow down the designs to be more feasible and high-quality over time, as well as efficiently and programmatically eliminate designs that aren’t fit for the project.

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?

The general workflow of creating a generative design study is

  1. Define the objective
  2. Establish design variables and objectives
  3. Set parameters in Dynamo
  4. Generate Design Options
  5. Label and Evaluate
  6. Review and Refine

This process can rapidly explore a wide range of design options and identify solutions that best balance the design criteria, while also maintaining automation and repeatability.

  1. 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 create a synthetic dataset for a specified building morphology study, here’s an outline of how I would approach it:

  1. Define the objectives and provided dataset on what should be achieved
  2. Implement parametric models in Dynamo and create the basic geometry model parametrically
  3. Categorize building forms and define taxonomy
  4. Use Revit’s generative design process to generate solutions through different variables
  5. Run simulations to generate analytical models and obtain evaluator results
  6. Export meaningful data for machine learning to train and learn
  1. 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?

The four solvers are:

  1. Randomize solver: randomly samples the design space and explores a wide variety of forms. Impacts the design options by offering high diversity of options with less structure.
  2. Cross Product solver: explores every combination of parameters and produces a comprehensive dataset. Impacts the design options by providing a complete coverage of solutions.
  3. Like-this solver: makes adjustments based on the design selected, which is good for fine-tuning and sensitivity analysis. Impacts the design options by narrowing down the design focus and emphasizing local exploration.
  4. Optimize solver: uses evolutionary algorithms to improve design, which is useful for performance oriented form-finding and design optimization. Impacts the design options by focusing on high-performing designs for narrowed optimization.
  1. 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 relationship between solver choice and design diversity:

  • Randomize solver produces a wide variety of forms, which is great for exploring designs but miss out on edge cases or optimal solutions
  • Cross Product solver covers all combinations possible, which is great for thorough analysis and training ML models with a balanced dataset
  • Like-this solver produces results based on a single selected base design, which is great for building on top of a solid foundation
  • Optimize solver covers on a few high-performance forms, which is good for finding the optimal solutions at the cost of diversity

Based on these insights, for a practical parametric design project using Dynamo I would first use Randomize and Cross Product solver to generate a diverse range of designs, then use Like-this solver to refine on the designs that are promising, and lastly use Optimize solver to narrow down on the designs that best optimize performance goals.