Building Structural Design Optimization Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The structural design process for buildings is time-consuming and laborious due to the large design space, iterative nature of optimization methods, and slow simulation tools, leading to suboptimal designs that over-satisfy building codes but fail to minimize material usage.
Innovation Solution
The development of an end-to-end pipeline that represents building structures as graphs and trains graph neural networks, specifically NeuralSizer and NeuralSim, to optimize cross-section sizes of columns and beams, enabling efficient and expeditious structural design optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If optimization algorithms are used to minimize material usage, then design objective performance is improved, but computation time increases to days
Solution Approach 1:
The patent creates a neural network surrogate model that copies the behavior of expensive structural simulation tools. This surrogate model can rapidly evaluate design candidates without running full structural simulations, enabling thousands of iterations per day instead of days of computation while maintaining design optimization performance.
Solution Approach 2:
The patent replaces the mechanical computation system (slow structural simulation tools) with a neural network-based computational system. The neural network learns from simulation data and provides fast predictions, substituting the slow mechanical simulation process with a fast neural inference process that scales efficiently.
2Measurement precision
If structural simulation tools are used for each optimization iteration, then design accuracy is improved, but evaluation time increases to 2-15 minutes per iteration
Solution Approach 1:
The neural network surrogate model copies the accurate evaluation capabilities of structural simulation tools while operating much faster. It is trained on simulation data to replicate the same design evaluation accuracy in a fraction of a second, enabling high-throughput optimization.
Solution Approach 2:
The patent performs preliminary training of the neural network model using simulation data before the actual optimization process. This preliminary action creates a fast evaluation tool that can be repeatedly used during optimization without requiring time-consuming simulations for each evaluation.
3Ease of operation
If trial-and-error design method is used, then design simplicity is maintained, but material usage increases due to over-design
Solution Approach 1:
The neural network optimization system performs self-service by automatically evaluating and optimizing design candidates without requiring manual trial-and-error by engineers. The system autonomously iterates through design space, evaluates performance using the neural network surrogate, and converges on optimal solutions that minimize material usage while satisfying building codes.
Solution Approach 2:
The patent implements feedback loops where the neural network continuously evaluates design candidates and provides guidance for improvement. This automated feedback mechanism allows the system to learn from previous evaluations and progressively improve designs, avoiding the over-design that occurs in manual trial-and-error processes.
Data Source
AI summary
A method and system provide the ability to optimize a structural engineering design. A dataset is synthesized by acquiring a structural skeleton design of an entire building. The skeleton defines locations and connectivities of bars that represent columns or beams. The skeleton design is represented as a structural graph with each bar represented as a graph node and edges connecting graph nodes. Structural simulation results are computed for the synthetic dataset based on the structural graph, various loads, and a structural analysis. A simulation model and a size optimization model are trained based on the structural simulation results with the size optimization model determining cross-section sizes for the bars to satisfy a building mass objective, building constraints, and output from the simulation model. The structural engineering design is output from the size optimization model.