Jump to content

Computational Design Synthesis

From Appropedia

Computational design synthesis (CDS) uses computational methods to generate and evaluate alternative designs against specified requirements and constraints. Rather than only analyzing or optimizing an existing design, CDS can explore different configurations, components, structures and parameter values. Methods may be fully automated or interactive, allowing designers to examine alternatives that would be difficult or slow to develop by hand. Among the earliest documented industrial examples were programs used at Westinghouse by 1956 to design electric motors, generators and transformers.[1] By 2005, CDS had been an active research field for almost half a century, with some results entering industrial practice and many research challenges still open.[2]

How it works

[edit | edit source]

A synthesis problem starts with design goals and constraints. Cagan et al. divide the process into four activities, and argue that any automated design system includes some form of each:[2]

  • Representation: define the design space, including required functions, available components, allowable configurations, design variables and constraints.
  • Generation: create candidate designs using computational rules, algorithms or other synthesis methods.
  • Evaluation: assess candidates against requirements such as performance, cost, reliability or manufacturability, using calculation, simulation, rules or human judgment.
  • Guidance: use evaluation results to direct further exploration, improve candidates or select promising solutions.

Generation, evaluation and guidance are typically repeated. In a genetic algorithm, for example, generation mutates and crosses over existing candidates, each candidate is evaluated against the objectives and constraints, and guidance is the selection that removes less fit candidates from the search.[2] The process does not have to be fully automated.

Methods

[edit | edit source]

A 2011 review identifies three major approaches to CDS: function-based, grammar-based and analogy-based synthesis.[3]

  • Function-based synthesis generates design concepts from a representation of the functions a system must perform.
  • Grammar-based synthesis uses formal rules to build or transform designs, including their geometry, topology or component relationships. Shape grammarsW are one example.
  • Analogy-based synthesis draws on existing designs or analogous systems to generate solutions to new problems.

Search and optimization algorithms, such as evolutionary algorithms, simulated annealing and mathematical optimization, are used to explore and compare candidates. Machine learning can support generation or performance prediction, but CDS does not require it.

Relationship to other design methods

[edit | edit source]

CDS overlaps with several computational design approaches, but the terms are not interchangeable.

  • Computer-aided design (CAD) provides tools for creating and modifying digital models, without necessarily generating alternative concepts.
  • Parametric design represents a design through adjustable parameters and relationships. Changing those parameters does not necessarily constitute design synthesis.
  • Design optimization searches for designs that improve an objective, often within a fixed configuration, although it can also address topology or architecture. CDS can explore alternative configurations, component choices and system architectures, and may incorporate optimization methods. Cagan et al. contrast synthesis with traditional optimization: synthesis aims more broadly to capture, emulate or use the decisions human designers make when they create a design.[2]
  • Generative designW also uses algorithms to create design alternatives and overlaps substantially with CDS. Terminology and scope vary between disciplines.[4]

3D printing, CNC machining and other digital fabrication technologies can be used to make or test designs produced through CDS. These are fabrication methods rather than synthesis methods, although physical testing can inform further design iterations.

Example

[edit | edit source]

Shape annealing illustrates all four activities. Shea and colleagues combined a shape grammar, which represents possible truss and frame configurations, with simulated annealing to search for designs. Candidates are evaluated using finite element analysis, and the search is guided by move sets that favour larger changes early in a run and smaller ones later.[2]

Shea and Cagan used shape annealing to generate roof truss designs, comparing them with designs proposed by six designers (three architects and three civil engineers) to assess how the method could support different design intents.[5] In a separate application, Shea and Zhao used an industrial version of shape annealing called eifForm to design the Paternoster Square Noon Mark, a cantilever structure that was built in London.[6][2]

Sustainability

[edit | edit source]

For appropriate technology and sustainable product design, objectives can include local material availability, repairability, reuse, energy use and life-cycle environmental impacts. These objectives may conflict. Research on computational multi-objective optimization for circular economy product design shows how competing objectives can be represented and their trade-offs explored.[7] Environmental benefits are not automatic; they depend on how impacts are represented and evaluated, and whether the resulting designs improve them.

Benefits and limitations

[edit | edit source]

CDS can widen the range of alternatives considered, automate repetitive design work, and help designers explore trade-offs between competing objectives. It can also help when requirements or available components change.

Its effectiveness depends on how the problem is represented, the accuracy of the evaluation methods, and the search strategy. Poorly specified requirements or incomplete models can produce unsuitable designs. Complex problems can require substantial computing resources, especially when each candidate is evaluated through numerical simulation.

Generated designs still need verification and validation. A design that is computationally feasible is not necessarily safe, manufacturable or capable of meeting its intended performance requirements.

References

[edit | edit source]
  1. ↑ Simon, H. A. (1993). Anecdotes—a very early expert system. IEEE Annals of the History of Computing, 15(3), 64–68.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 2.5 Cagan, J., Campbell, M. I., Finger, S., and Tomiyama, T. (2005). A framework for computational design synthesis: Model and applications. Journal of Computing and Information Science in Engineering, 5(3), 171–181.
  3. ↑ Chakrabarti, A., Shea, K., Stone, R., Cagan, J., Campbell, M., Vargas Hernandez, N., and Wood, K. L. (2011). Computer-based design synthesis research: An overview. Journal of Computing and Information Science in Engineering, 11(2), 021003.
  4. ↑ Caetano, I., Santos, L., and Leitão, A. (2020). Computational design in architecture: Defining parametric, generative, and algorithmic design. Frontiers of Architectural Research, 9(2), 287–300.
  5. ↑ Shea, K., and Cagan, J. (1999). The design of novel roof trusses with shape annealing: Assessing the ability of a computational method in aiding structural designers with varying design intent. Design Studies, 20(1), 3–23.
  6. ↑ Shea, K., and Zhao, X. (2004). A novel noon mark cantilever support: From design generation to realization. IASS 2004: Shell and Spatial Structures from Models to Realization, Montpellier, France.
  7. ↑ Ortner, P., Tay, J. Z., and Wortmann, T. (2022). Computational optimization for circular economy product design. Journal of Cleaner Production, 362, 132340.
Page data
Authors Cameron
License CC-BY-SA-4.0
Language English (en)
Related 0 subpages, 1 pages link here
Views 34 page views (analytics)
Created February 5, 2024 by Cameron
Last edit November 11, 2026 by Cameron
Cookies help us deliver our services. By using our services, you agree to our use of cookies.