Evolutionary Design
Bentley Design Evolution (DE) is a plug-in prototype for Bentley Generative Components (GC). It enables Evolutionary Design, an innovative approach for achieving cost-effective and environmentally-friendly solutions by computationally emulating the process of natural evolution and the key principles of genetic reproduction. Design Evolution employs the Darwin Optimization Framework (DOF), a generic parallel optimization platform based on evolutionary computation, automatically generates and evaluates tens of thousands of the alternative solutions, thus the quality of the design solution is expected to be improved and consequently cost efficiency is maximized. It is effective at handling various design optimization tasks with and without the criteria of the structural finite element analysis and/or building energy analysis. With DE, users are able to select the desirable parameters as the decision variables, connect the implemented GC transactions (dedicated as objectives) with fitness functions and the GC transactions (dedicated as constraints) with the constraint. The optimization run is conducted by invoking DOF and multiple top- or near-optimal solutions are saved for users to further evaluate in GC design environment. Bentley DE plug-in prototype is to be productized with Generative Component and Multidisciplinary Design Optimization projects.
Software prototype:
- Download Design Evolution on Be Community
Multidisciplinary Design Optimization
With growing concerns on the building energy consumption, for instance, buildings collectively consume about 40% of the total energy and contribute more than 30% of USA’s carbon footprint. It is imperative to design, construct and operate energy-efficient buildings. Commercial building is a complex system, with the energy use and performance of any one part of the system affecting the energy use of the building as a whole through a complex cascade of interactions. However, the typical design process for commercial buildings is a linear, sequential process that precludes the analysis and design of the buildings as an integrated system. In order to achieve deep savings in energy use, an integrated and iterative design process, involving all members of the design team, is required. An effective process of the integrated design is to optimize all of the design variables that affect one another so that the overall performance is maximized. The performance can be measured in many criteria such as structural cost, energy cost and/or carbon footprint, which are not always complementary but often conflict each other. A generic optimization model is developed for simultaneously minimizing structure cost and energy consumption while satisfying all the design constraints. The method is capable of optimizing any combination of user specified structural design variables, such as cross section areas, member sizes, truss topology and connectivity, as well as energy performance related design variables e.g. windows, insulations. Darwin Optimization Framework is applied to generate and evolve the design solutions that are analyzed by performing structure analysis and energy analysis. Thus it results in a Multidisciplinary Design Optimization (MDO) framework, which is productized into Bentley software product for multidisciplinary design optimization.
Technical papers:
- Parallel Optimization of Structural Design and Building Energy Performance, presented at ASCE Structure Congress 2011, Las Vagas, NV, USA
Geometry Design Optimization
Geometry modeling has become an increasingly powerful approach for architectural and structural design. It is an effective approach especially when a geometry model is constructed by making use of the associative parameters, so-called algorithm-based parametric design. However, the parametric associative model provides limited function for effectively exploring design alternatives, users have to manually tune the design by adjusting the parameters (e.g. predefined graph variables) to arrive with a final design. Only a handful of design alternatives can be generated and evaluated, the design space is not fully explored to arrive with the high-quality design. A new approach is developed to leverage the use of parametric model to optimize a geometry design by using genetic algorithm (GA), a search method based on the principles of natural evolution and genetic reproduction. Using the GA and geometry modeling tool Bentley Generative Component, a design can be represented by encoding design variables onto a binary string or genotype, design alternatives are evolved by mimicking crossover, mutation and natural selection principle of Darwin’s survival-of-fittest. The tool allows user to select any combination of parametric graph variables and geometry parameters to optimize the design. A design solution is evaluated by using the fitness score that can be any user-defined criteria. One fitness score is assigned to each of new designs. The fitter the solutions are, the more likely the solutions are selected to reproduce next generation of design solutions. Thus solutions are optimized generation after generation via emulating natural evolution. The method developed in this research is productized into the software products Bentley Generative Component and Multidisciplinary Design Optimization.
Software prototype:
- Download Design Evolution on Be Community
- Design Evolution Demo:
Technical report: