Developing a reliable and robust finite element model of a carbon fiber-reinforced plastic (CFRP) composite structure is investigated by using the LS-DYNA solver and Python. This study tries to provide a systematic numerical approach to cover the principal impediment to adaptation of composite energy absorbers, that is the lack of a reliable predictive method. The proposed procedure aims to further the understanding of advanced composite structures’ behavior during the crash phenomenon by developing an accurate finite element model. To do so, the mechanical properties of the material were extracted from American Society for Testing and Materials (ASTM) standard test methods, followed by experimental investigation of circular CFRP tubes undergoing quasi-static loading. A numerical simulation framework was then utilized to scrutinize the effectiveness of simulation parameters on the crushing mechanism. Finally, a systematic approach based on machine learning techniques was performed to adjust non-physical modeling parameters for further calibration and validation. In this regard, a versatile Python code was developed to automate pre-processing, processing, and post-processing steps. The code also provides a groundwork to perform machine learning techniques. Interestingly, the numerical and experimental results were highly correlated with a correlation coefficient of almost 90%. Additionally, several non-physical numerical parameters were found to be inactive, while some else were identified as effective parameters, and their corresponding effectiveness was quantitatively extracted and discussed for the first time in the literature.

中文翻译：

---

碳纤维增强塑料圆管稳健数值模拟的新颖系统方法：利用机器学习技术进行校准和验证

使用 LS-DYNA 求解器和 Python 研究开发碳纤维增强塑料 (CFRP) 复合结构的可靠且稳健的有限元模型。本研究试图提供一种系统的数值方法来解决复合能量吸收器适应的主要障碍，即缺乏可靠的预测方法。所提出的程序旨在通过开发精确的有限元模型来进一步了解先进复合材料结构在碰撞现象中的行为。为此，从美国材料与试验协会 (ASTM) 标准测试方法中提取了材料的机械性能，然后对承受准静态载荷的圆形 CFRP 管进行了实验研究。然后利用数值模拟框架来检查模拟参数对破碎机制的有效性。最后，采用基于机器学习技术的系统方法来调整非物理建模参数，以进一步校准和验证。在这方面，开发了通用的Python代码来自动化预处理、处理和后处理步骤。该代码还为执行机器学习技术提供了基础。有趣的是，数值结果和实验结果高度相关，相关系数接近 90%。此外，发现一些非物理数值参数是无效的，而其他一些则被确定为有效参数，并且在文献中首次定量提取和讨论了它们相应的有效性。