Abstract:Compared to the classically performed elastic limit rule in structural lightweight optimization design, the shakedown analysis can help determine the non-failure external load region which is greater than the elastic limit but less than the plastic limit, without accounting for the loading history. By sacrificing some abundant load-bearing capacity, a lightweight design of the structure can be achieved. Combining the shakedown analysis with parameter optimization can help resolve parameter optimization intervals while considering structural fatigue in complex loading conditions. This approach ultimately translates into solving a max-min optimization problem. However, due to the nonlinear relationship between the shakedown limit and massive design parameters, a rugged 'performance landscape' which is considered to be extremely unfriendly will be led through a traditional surrogate model, sometimes it is not even possible to generate. Moreover, the total time-consuming of simulation data acquisition is unacceptable. The current research utilizes Generative Adversarial Networks (GANs) to enrich the surrogate model with shakedown strength boundary constraints. Benefiting from the unsupervised learning feature and the similarity between the generated data and the real data, GANs are able to build relationships in the above optimization problem quickly, which is an innovative and executable technique to the lightweight design under structural strength constraints. The research first presents the shakedown theory and its numerical implementation. Then, a shakedown strength constrained for mass minimization form and the GANs flowchart for solving this problem are built, and a benchmark example is presented to validate the suggested technique. By using this effective workflow combining shakedown theory and GANs to optimize lightweight frame structures, the structure was able to quickly achieve a lighter configuration design while avoiding failure forms like incremental collapse, ratcheting, and alternate plasticity. Furthermore, the method is used to determine an optimal lightweight design for a manned airtight module. The parametric study presented in the paper demonstrated the effectiveness of the proposed method in reducing the total mass of the structure. Moreover, the study highlights the improvement in design performance by allowing redundant shakedown load capacity and reducing material usage.

Generative adversarial network for optimization of operational parameters based on shield posture requirements

Optimal Decision-making System of Shield Tunneling Parameters Based on LSTM and MNSGA-II

A Method of Adaptive Hyperparameter Optimization for Deep Generative Models

Training Generative Adversarial Networks in One Stage

Evaluation of GAN-Based Model for Adversarial Training

Soft Generative Adversarial Network: Combating Mode Collapse in Generative Adversarial Network Training Via Dynamic Borderline Softening Mechanism

Research on Data Generation Based on the Combination of Growing-Pruning Gan and Intelligent Parameter Optimization

MO-PaDGAN: Reparameterizing Engineering Designs for augmented multi-objective optimization

Rob-GAN: Generator, Discriminator, and Adversarial Attacker

Generative Adversarial Networks Guided Lightweight Design Based on Shakedown Strength Constraint

Machine learning-based automatic control of tunneling posture of shield machine

Direct Adversarial Training for GANs

A Generation Adversarial Network Based on Multi-Condition Confrontation and Gradient Optimization

Improving posture recognition among construction workers through data augmentation with generative adversarial network

Assessment of Tool Wear With Insufficient and Unbalanced Data Using Improved Conditional Generative Adversarial Net and High-Quality Optimization Algorithm

Model-Informed Generative Adversarial Network (MI-GAN) for Learning Optimal Power Flow

Training Generative Adversarial Networks by Auxiliary Adversarial Example Regulator

Improving OPT-GAN by Smooth Scale Mapping and Adaptive Exploration

Real-time Analysis and Prediction of Shield Cutterhead Torque Using Optimized Gated Recurrent Unit Neural Network

Improving GAN Training via Feature Space Shrinkage

Augmenting Character Designers Creativity Using Generative Adversarial Networks