结构可靠性拓扑优化的深度学习协同方法
PDF下载 (93)阮相当,许孟辉,李 萍,蒋章宇,贺元强,刘天旭,陶 磊.结构可靠性拓扑优化的深度学习协同方法[J].宁波大学学报(理工版),2024,37(4):103-109.DOI:10.20098/j.cnki.1001-5132.2024.0138
RUAN Xiangdang,XU Menghui,LI Ping,JIANG Zhangyu,HE Yuanqiang,LIU Tianxu,TAO Lei.Reliability-based topology optimization collaborated with deep learning[J].Journal of Ningbo University(Natural Science & Engineering Edition),2024,37(4):103-109.DOI:10.20098/j.cnki.1001-5132.2024.0138
| Title: | Reliability-based topology optimization collaborated with deep learning |
| 作者: | 阮相当, 许孟辉, 李 萍, 蒋章宇, 贺元强, 刘天旭, 陶 磊 |
| Author(s): | RUAN Xiangdang, XU Menghui, LI Ping, JIANG Zhangyu, HE Yuanqiang, LIU Tianxu, TAO Lei |
| 关键词: | 拓扑优化; 可靠性优化; 深度神经网络; 有限元分析 |
| Keywords: | topology optimization; reliability-based optimization; deep neural network; finite element analysis |
| 分类号: | O34; O39 |
| DOI: | 10.20098/j.cnki.1001-5132.2024.0138 |
| 文献标识码: | A |
| 摘要: | 在给定约束条件下, 结构拓扑优化以最优化性能获得设计域内材料的最佳布局. 传统拓扑优化面临两大挑战: 其一, 未定量计入参数不确定性的影响, 导致优化后结构在服役中常常不满足性能约束; 其二, 基于有限元模型的拓扑优化分辨率严重依赖于有限元网格. 本文以随机参数定量计及固有不确定性, 基于顺序优化与可靠性分析实现可靠性优化模型的确定性转化. 进一步, 基于网络训练与拓扑优化的过程相似性, 以深度神经网络表达单元相对密度与网络参数之间的映射, 通过构造计及可靠性的损失函数, 将拓扑优化过程转换为网络训练过程. 算例验证了所提方法获得的拓扑优化结构能够在满足性能可靠性要求的前提下实现较高的拓扑分辨率. |
| Abstract: | Structural topology optimization obtains the best layout of materials in the design domain with optimal performance under given constraints. Two challenges faced by the traditional topology optimization include: (1) the effects of parameter uncertainties are not quantitatively accounted for, resulting in optimized structures that often do not satisfy performance constraints in service; (2) the resolution of topology optimization based on finite element models is heavily dependent on the finite element mesh. In this paper, the deterministic transformation of reliability-based topology optimization (RBTO) model is realized based on sequential optimization and reliability assessment method by quantitatively accounting for the inherent uncertainty with stochastic parameters. Further, based on the process similarity between the neural network training and the topology optimization, the network parameters (including weights and bias) are mapped to the relative density of all elements by a deep neural network. As a result, the topology optimization process is converted into a network training process by constructing a loss function that accounts for reliability. Finally, it is demonstrated that the proposed method can achieve higher resolution of the optimized structure while meeting the reliability requirement. |
| 参考文献 /References: | [1] MICHELL A G M. LVIII. The limits of economy of material in frame-structures[J]. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 1904, 8(47):589-597. [2] BENDSOE M P, KIKUCHI N. Generating optimal topologies in structural design using a homogenization method[J]. Computer Methods in Applied Mechanics and Engineering, 1988, 71(2):197-224. [3] ROZVANY G. The SIMP method in topology optimization: theoretical background, advantages and new applications[C]//8th Symposium on Multidisciplinary Analysis and Optimization. Long Beach, CA: AIAA, 2000:4738. [4] SOSNOVIK I, OSELEDETS I. Neural networks for topology optimization[J]. Russian Journal of Numerical Analysis and Mathematical Modelling, 2019, 34(4):215- 223. [5] YU Y, HUR T, JUNG J, et al. Deep learning for determining a near-optimal topological design without any iteration[J]. Structural and Multidisciplinary Optimization, 2019, 59(3):787-799. [6] CHI H, ZHANG Y Y, TANG T L E, et al. Universal machine learning for topology optimization[J]. Computer Methods in Applied Mechanics and Engineering, 2021, 375:112739. [7] WANG D L, XIANG C, PAN Y, et al. A deep convolutional neural network for topology optimization with perceptible generalization ability[J]. Engineering Optimization, 2022, 54(6):973-988. [8] YUE T L, YANG H, DU Z L, et al. A mechanistic-based data-driven approach to accelerate structural topology optimization through finite element convolutional neural network (FE-CNN)[EB/OL]. [2023-12-14]. http://arxiv. org/abs/2106.13652. [9] CHANDRASEKHAR A, SURESH K. TOuNN: topology optimization using neural networks[J]. Structural and Multidisciplinary Optimization, 2021, 63(3):1135-1149. [10] JUNG H S, CHO S. Reliability-based topology optimization of geometrically nonlinear structures with loading and material uncertainties[J]. Finite Elements in Analysis and Design, 2004, 41(3):311-331. [11] JALALPOUR M, TOOTKABONI M. An efficient approach to reliability-based topology optimization for continua under material uncertainty[J]. Structural and Multidisciplinary Optimization, 2016, 53(4):759-772. [12] PHAM T, HOYLE C. Reliability-based topology optimization under random-field material model[EB/OL]. [2023-12-14]. http://arxiv.org/abs/2112.14802. [13] MENG Z, PANG Y S, PU Y X, et al. New hybrid reliability-based topology optimization method combining fuzzy and probabilistic models for handling epistemic and aleatory uncertainties[J]. Computer Methods in Applied Mechanics and Engineering, 2020, 363:112886. [14] BRUGGI M, ISMAIL H, LÓGÓ J. Topology optimization with graded infill accounting for loading uncertainty[J]. Composite Structures, 2023, 311:116807. [15] CHENG C Z, YANG B, WANG X, et al. Reliability-based topology optimization using the response surface method for stress-constrained problems considering load uncertainty [J]. Engineering Optimization, 2023, 55(11):1923-1939. [16] DU X P, CHEN W. Sequential optimization and reliability assessment method for efficient probabilistic design[J]. Journal of Mechanical Design, 2004, 126(2): 225-233. [17] LOPEZ R H, BECK A T. Reliability-based design optimization strategies based on FORM: a review[J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 2012, 34(4):506-514. [18] LU L, SHIN Y, SU Y H, et al. Dying ReLU and initialization: theory and numerical examples[EB/OL]. [2023-12-14]. http://arxiv.org/abs/1903.06733. [19] SMITH L N. A disciplined approach to neural network hyper-parameters: part 1: learning rate, batch size, momentum, and weight decay[EB/OL]. [2023-12-14]. http://arxiv.org/abs/1803.09820. [20] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323:533-536. [21] PASZKE A, GROSS S, MASSA F, et al. PyTorch: an imperative style, high-performance deep learning library [EB/OL]. [2023-12-14]. http://arxiv.org/abs/1912.01703. [22] PASCANU R, MIKOLOV T, BENGIO Y. On the difficulty of training recurrent neural networks[EB/OL]. [2023-12-14]. http://arxiv.org/abs/1211.5063. |
| 备注/Memo: | 收稿日期: 2024−01−26. 宁波大学学报（理工版）网址: http://journallg.nbu.edu.cn/ 基金项目: 宁波大学省属高校基本业务费项目（SJLY2022003）; 新疆维吾尔自治区天山青年计划（2020Q083）; 新疆维吾尔自治区自然科学基金（2021D01B46）. 第一作者: 阮相当, 硕士研究生, 主要研究方向: 结构可靠性拓扑优化. E-mail: nanjiang_nbu@163.com *通信作者: 许孟辉, 副教授, 主要研究方向: 结构拓扑优化、可靠性分析. E-mail: xumenghui@nbu.edu.cn 宁波大学学报（理工版）网址：http://journallg.nbu.edu.cn/ |
