用物理仿真生成包装运输数据，自动评估堆叠稳定性
Physics-Based 3D Simulation for Synthetic Data Generation and Failure Analysis in Packaging Stability Assessment
- 构建3D物理引擎模拟托盘动态行为，支持多种包装布局和材料
- 通过神经网络预测堆叠崩溃风险，准确率提升显著
- 减少实物测试，适合物流与包装安全研究者使用
托盘堆叠设计与分析对保障货物运输安全至关重要。随着物流业需求上升，采用先进技术的自动化系统开发日益关键。同时，塑料缠绕膜的广泛应用促使研究者探索环保替代方案，同时满足安全标准。本文提出一个完全可控且高精度的物理仿真系统，可真实再现移动托盘的行为。该系统基于3D图形虚拟环境，支持多种配置，包括可变包装布局、不同缠绕材料及多样动态条件。此方法大幅减少实物测试需求，降低费用与环境影响，同时提升堆叠动力学分析的测量精度。此外，我们训练了一个深度神经网络，用于评估模拟生成的视频，作为托盘配置的碰撞测试预测器，进一步增强系统在安全分析中的应用价值。
原文摘要 · Abstract (English)
The design and analysis of pallet setups are essential for ensuring safety of packages transportation. With rising demands in the logistics sector, the development of automated systems utilizing advanced technologies has become increasingly crucial. Moreover, the widespread use of plastic wrapping has motivated researchers to investigate eco-friendly alternatives that still adhere to safety standards. We present a fully controllable and accurate physical simulation system capable of replicating the behavior of moving pallets. It features a 3D graphics-based virtual environment that supports a wide range of configurations, including variable package layouts, different wrapping materials, and diverse dynamic conditions. This innovative approach reduces the need for physical testing, cutting costs and environmental impact while improving measurement accuracy for analyzing pallet dynamics. Additionally, we train a deep neural network to evaluate the rendered videos generated by our simulator, as a crash-test predictor for pallet configurations, further enhancing the system's utility in safety analysis.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。