IP Library › Granted Patent US 12,346,641
Granted Patent B2
US 12,346,641 · App. 17/200,546 · Granted Jul 1, 2025

Learning to simulate and design for structural engineering

Inventors: Kai-Hung Chang (El Cerrito, CA); Chin-Yi Cheng (Walnut Creek, CA); Mehdi Nourbakhsh (Richmond, CA)
Assignee: AUTODESK, INC.
G06F30/27G06F30/13G06N20/00G06F30/23
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,346,641
App. No.
17/200,546
Granted
Jul 1, 2025
Kind
B2
Abstract

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.

Claims (61)

1. A computer-implemented method for optimizing a structural engineering design, comprising:

acquiring a structural skeleton design of an entire building, wherein the structural skeleton design defines locations and connectivities of bars, wherein every bar represents a column or beam of the building;

training a simulation model based on computed structural simulation results for a synthetic dataset acquired from the structural skeleton design of the building based on a structural graph, various loads, and a structural analysis, wherein the simulation model serves as a surrogate for a finite element analysis structural analysis;

training a size optimization model for the structural skeleton design represented as the structural graph with each bar represented as a graph node and an edge connecting two graph nodes corresponding to two bars that are joined together in the structural skeleton design, wherein cross-section sizes for the bars is based on the size optimization model to satisfy a building mass objective and building constraints, and wherein the size optimization model is further based on the output from the simulation model; and

outputting the structural engineering design from the size optimization model.

2. The computer-implemented method of claim 1 , wherein the training the simulation model comprises:

embedding each graph node into the simulation model, wherein the simulation model approximates the structural simulation results;

iteratively updating the embedded graph nodes based on an aggregated message to obtain final embeddings of all graph nodes; and

decoding the final embeddings to output predicted drift ratios and a classification regarding whether a ground-truth drift ratio exceeds a drift ratio limit.

3. The computer-implemented method of claim 2 , wherein the aggregated message comprises:

forces and reaction forces between two bars.

4. The computer-implemented method of claim 2 , wherein the aggregated message comprises:

a position-aware message that provides global information to help identify load conditions.

5. The computer-implemented method of claim 2 , further comprising:

determining a multitask loss based on the predicted drift ratios and the classification; and

further training the simulation model based on the multitask loss.

6. The computer-implemented method of claim 1 , wherein the training the size optimization model comprises:

embedding each graph node into the size optimization model, wherein the size optimization model optimizes cross-section sizes;

iteratively updating the embedded graph nodes based on an aggregated message to obtain final embeddings of all graph nodes; and

mapping the final embeddings concatenated by a graph embedding to a probability over cross-sections.

7. The computer-implemented method of claim 6 , further comprising:

determining a differentiable loss based on the building mass objective and building constraints; and

further training the size optimization model based on the differentiable loss.

8. The computer-implemented method of claim 7 , wherein the building constraints comprise:

a drift ratio constraint that requires a drift ratio for each story of the entire building to be less than a limit under lateral seismic loads;

a variety constraint that sets a maximum number of different cross-section types used; and

an entropy constraint that is based on an entropy output for each bar, a maximum entropy over different cross-sections, and a defined target ratio.

9. The computer-implemented method of claim 1 , further comprising:

utilizing the simulation model and size optimization model to visualize studies for a procurement, a fabrication, and a construction of the structural engineering design.

10. A computer-implemented system for optimizing a structural engineering design, comprising:

(a) a computer having a memory;

(b) a processor executing on the computer;

(c) the memory storing a set of instructions, wherein the set of instructions, when executed by the processor cause the processor to perform operations comprising:

(1) acquiring a structural skeleton design of an entire building, wherein the structural skeleton design defines locations and connectivities of bars, wherein every bar represents a column or beam of the building;

(2) training a simulation model based on computed structural simulation results for a synthetic dataset acquired from the structural skeleton design of the building based on a structural graph, various loads, and a structural analysis, wherein the simulation model serves as a surrogate for a finite element analysis structural analysis;

(3) training a size optimization model for the structural skeleton design, represented as the structural graph with each bar represented as a graph node and an edge connecting two graph nodes corresponding to two bars that are joined together in the structural skeleton design, wherein cross-section sizes for the bars is based on the size optimization model to satisfy a building mass objective and building constraints, and wherein the size optimization model is further based on the output from the simulation model; and

(4) outputting the structural engineering design from the size optimization model.

11. The computer-implemented system of claim 10 , wherein the operations training the simulation model comprises:

embedding each graph node into the simulation model, wherein the simulation model approximates the structural simulation results;

iteratively updating the embedded graph nodes based on an aggregated message to obtain final embeddings of all graph nodes; and

decoding the final embeddings to output predicted drift ratios and a classification regarding whether a ground-truth drift ratio exceeds a drift ratio limit.

12. The computer-implemented system of claim 11 , wherein the aggregated message comprises:

forces and reaction forces between two bars.

13. The computer-implemented system of claim 11 , wherein the aggregated message comprises:

a position-aware message that provides global information to help identify load conditions.

14. The computer-implemented system of claim 11 , wherein the operations further comprise:

determining a multitask loss based on the predicted drift ratios and the classification; and

further training the simulation model based on the multitask loss.

15. The computer-implemented system of claim 11 , wherein the operations training the size optimization model comprises:

embedding each graph node into the size optimization model, wherein the size optimization model optimizes cross-section sizes;

iteratively updating the embedded graph nodes based on an aggregated message to obtain final embeddings of all graph nodes; and

mapping the final embeddings concatenated by a graph embedding to a probability over cross-sections.

16. The computer-implemented system of claim 15 , wherein the operations further comprise:

determining a differentiable loss based on the building mass objective and building constraints; and

further training the size optimization model based on the differentiable loss.

17. The computer-implemented system of claim 16 , wherein the building constraints comprise:

a drift ratio constraint that requires a drift ratio for each story of the entire building to be less than a limit under lateral seismic loads;

a variety constraint that sets a maximum number of different cross-section types used; and

an entropy constraint that is based on an entropy output for each bar, a maximum entropy over different cross-sections, and a defined target ratio.

18. The computer-implemented system of claim 11 , wherein the operations further comprise:

utilizing the simulation model and size optimization model to visualize studies for a procurement, a fabrication, and a construction of the structural engineering design.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Mar 16, 2021
From: NOURBAKHSH, MEHDI
To: AUTODESK, INC.
Reel/Frame 055611/0404 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Mar 12, 2021
From: CHANG, KAI-HUNG; CHENG, CHIN-YI
To: AUTODESK, INC.
Reel/Frame 055580/0818 →
Continuity (2)
Provisional Application 62988845 · Mar 12, 2020
Related Publication 20210287138A1 · Sep 16, 2021
References Cited (62)
US 5815394A · Adeli · 1998 [cited by examiner]
US 8825459B2 · Aish · 2014 [cited by examiner]
US 9697326B1 · Bowman · 2017 [cited by examiner]
US 9891791B1 · Khan · 2018 [cited by examiner]
US 10190792B2 · Jacobson · 2019 [cited by examiner]
US 10699043B2 · Ho · 2020 [cited by examiner]
US 11488062B1 · Earthman · 2022 [cited by examiner]
US 11900026B1 · Schubert · 2024 [cited by examiner]
US 20040044503A1 · McConaghy · 2004 [cited by examiner]
US 20110082638A1 · Khorashadi · 2011 [cited by examiner]
US 20120179430A1 · Aish · 2012 [cited by examiner]
US 20120203806A1 · Panushev · 2012 [cited by examiner]
US 20160147843A1 · Haley · 2016 [cited by examiner]
US 20180046732A1 · Bergin · 2018 [cited by examiner]
US 20210056242A1 · De Zaeytijd · 2021 [cited by examiner]
US 20210073449A1 · Segev · 2021 [cited by examiner]
US 20210150373A1 · Crouse · 2021 [cited by examiner]
US 20210287138A1 · Chang · 2021 [cited by examiner]
US 20220372068A1 · Kim · 2022 [cited by examiner]
Balogh et al. (Genetic algorithm based optimization of regular steel building structures subjected to seismic effects, WCEE, 2012, pp. 1-10) (Year: 2012). [cited by examiner]
Yeh et al. (Optimal Design of Steel cols. with Axial Load using Artificial Neural Networks, AMME 2017, pp. 189-194) (Year: 2017). [cited by examiner]
Cao et al. (Skeleton and Infill Housing Construction Delivery Process Optimization Based on the Design Structure Matrix, MDPI 2018, pp. 1-18) (Year: 2018). [cited by examiner]
Roith et al. (Supporting the building design process with graph-based methods using centrally coordinated federated databases, 2017, Springer, pp. 1-17) (Year: 2017). [cited by examiner]
Langenhan et al. (Graph-based retrieval of building information models for supporting the early design stages, 2013, Elsevier, pp. 413-427) (Year: 2013). [cited by examiner]
Achiam et al., “Constrained policy optimization”. In Proceedings of the 34th International Conference on Machine Learning—vol. 70, pp. 22-31. JMLR.org, 2017. [cited by applicant]
Balogh et al., “Genetic algorithm based optimization of regular steel building structures subjected to seismic effects”. In Proceedings 15th world conference on earthquake engineering, pp. 1-10, 2012. [cited by applicant]
Battaglia et al., “Relational inductive biases, deep learning, and graph networks”. 40 pages, arXiv preprint arXiv:1806.01261, 2018. [cited by applicant]
Bello et al., “Neural combinatorial optimization with reinforcement learning”. 15 pages, arXiv preprint arXiv:1611.09940, 2016. [cited by applicant]
International Energy Agency. World Energy Balances 2017. 21 pages, doi: https://doi.org/https://doi.org/ 10.1787/world energy bal-2017-en. URL https://www.oecd-ilibrary.org/content/ publication/world_energy_bal-2017-en. [cited by applicant]
Yeh et al., “Optimal design of steel columns with axial load using artificial neural networks”. DEStech Transactions on Engineering and Technology Research, (amme), pp. 189-194, 2017. [cited by applicant]
Cui et al., “Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecasting”. IEEE Transactions on Intelligent Transportation Systems, vol. 21, No. 11,… [cited by applicant]
Do et al., “Graph transformation policy network for chemical reaction prediction”. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 750-760, 2019. [cited by applicant]
Fout et al., “Protein interface prediction using graph convolutional networks”. In Advances in neural information processing systems, pp. 6530-6539, 2017. [cited by applicant]
Frans et al., “Unsupervised image to sequence translation with canvas-drawer networks”. arXiv preprint arXiv:1809.08340, 2018.Greco, L. Machine learning and optimization techniques for steel connections. In Proceedings … [cited by applicant]
Guo et al., “Attention based spatial-temporal graph convolutional networks for traffic flow forecasting”. In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, pp. 922-929, 2019. [cited by applicant]
Haarnoja et al., “Soft actor-critic algorithms and applications”. arXiv preprint arXiv:1812.05905, 17 pages, 2018. [cited by applicant]
Hamrick et al., “Relational inductive bias for physical construction in humans and machines”. arXiv preprint arXiv:1806.01203, 7 pages, 2018. [cited by applicant]
Hasanc ebi et al., “A neural network approach for approximate force response analyses of a bridge population”. Neural Computing and Applications, 22(3-4):755-769, 2013. [cited by applicant]
Imran et al., “Design optimization of composite submerged cylindrical pressure hull using genetic algorithm and finite element analysis”. Ocean Engineering, 190:106443, 12 pages, 2019. [cited by applicant]
Zhou et al., “Model-based deep hand pose estimation”. arXiv preprint arXiv:1606.06854, 7 pages, 2016. [cited by applicant]
Jang et al., “Categorical reparameterization with gumbel-softmax”. arXiv preprint arXiv:1611.01144, 13 pages, 2016. [cited by applicant]
Jin et al., “Junction tree varia-tional autoencoder for molecular graph generation”. arXiv preprint arXiv:1802.04364, 17 pages, 2018a. [cited by applicant]
Jin et al., “Learning multimodal graph-to-graph translation for molecular optimization”. arXiv preprint arXiv:1812.01070, 13 pages 2018b. [cited by applicant]
Kipf et al., “Neural relational inference for interacting systems”. arXiv preprint arXiv:1802.04687, 17 pages, 2018. [cited by applicant]
Kipf et al., “Semi-supervised classification with graph convolutional networks”. arXiv preprint arXiv:1609.02907, 14 pages, 2016. [cited by applicant]
Kool et al., “Attention, learn to solve routing problems!” arXiv preprint arXiv:1803.08475, 2018. [cited by applicant]
Li et al., “Combinatorial optimization with graph convolutional networks and guided tree search”. In Advances in Neural Information Processing Systems, pp. 539-548, 2018. [cited by applicant]
Prates et al., “Learning to solve np-complete problems: A graph neural network for decision tsp”. In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, pp. 4731-4738, 2019. [cited by applicant]
Rajeev et al., Genetic algorithms—based methodologies for design optimization of trusses. Journal of structural engineering, 123(3):350-358, 1997. [cited by applicant]
Sanchez-Gonzalez et al., “Graph networks as learnable physics engines for inference and control”. arXiv preprint arXiv:1806.01242, 2018. [cited by applicant]
Sanchez-Gonzalez et al., “Learning to simulate complex physics with graph networks”. arXiv preprint arXiv:2002.09405, 2020. [cited by applicant]
Tamura et al., “Machine learning for combinatorial optimization of brace placement of steel frames”. Japan Architectural Review, 1(4):419-430, 2018. [cited by applicant]
Tian et al., “Learning to infer and execute 3d shape programs”. arXiv preprint arXiv:1901.02875, 2019. [cited by applicant]
Torky et al., “A deep learning approach to automated structural engineering of prestressed members”. 2018. [cited by applicant]
Veličković et al., “Graph attention networks”. arXiv preprint arXiv:1710.10903, 2017. [cited by applicant]
Watters et al., “Visual interaction networks: Learning a physics simulator from video”. In Advances in neural Information processing systems, pp. 4539-4547, 2017. [cited by applicant]
Wu et al., “A comprehensive survey on graph neural networks”. IEEE Transactions on Neural Networks and Learning Systems, vol. 32, No. 1, 2021, pp. 4-24. [cited by applicant]
Xu et al., “How powerful are graph neural networks?” arXiv preprint arXiv:1810.00826, 17 pages, 2018. [cited by applicant]
You et al., “Graph convolutional policy network for goal-directed molecular graph generation”. In Advances in neural information processing systems, pp. 6410-6421, 2018. [cited by applicant]
You et al., “Position-aware graph neural networks”. arXiv preprint arXiv:1906.04817, 2019. [cited by applicant]
Zheng et al., “Strokenet: A neural painting environment”. 12 pages, 2018. [cited by applicant]
Zhou et al., “Graph neural networks: A review of methods and applications”. AI Open 1, pp. 57-81, arXiv preprint arXiv:1812.08434, 2020. [cited by applicant]