课题基金 / 基金详情

基于LSTM理论的高温合金服役多模式微观损伤相关本构建模方法研究

批准号:
52105139
项目类别:
青年科学基金项目(C类)
资助金额:
30.0 万元
负责人:
黄渭清
依托单位:
学科分类:
机械结构强度学
结题年份:
2024
批准年份:
2021
项目状态:
已结题
项目参与者:
黄渭清

项目摘要

结项摘要

相似基金

相关文献

中文摘要
涡轮叶片在高温严苛工作环境长时服役会引发各类微观损伤,造成力学性能恶化,影响整机安全可靠运行。因此微观结构相关高温合金力学行为建模工作已成为国内外研究热点。本项目针对涡轮叶片DZ125合金服役中出现时间相关多模式微观损伤与宏观变形行为关系不清、难建立显式数学物理描述问题,应用时间相关长短记忆神经网络(LSTM)理论,开展耦合多模式微观损伤DZ125合金本构建模方法研究。首先,基于主控强化相损伤,解耦多组织损伤,建立本构参数训练集;其次,构建耦合LSTM模型的长时服役流动方程;最后构建耦合多模式组织损伤时序演化的Chaboche统一粘塑性本构并进行精度验证。相较于传统唯象本构理论,本项目成果将能综合考虑高温合金服役期间多模式微观损伤及其时间相关演化的影响,具有较高的合理性,有望为我国发动机涡轮叶片服役期间性能评估,提供理论方法支撑。
英文摘要
The long-term service of turbine blade in high temperature and severe working environment will cause all kinds of microstructure damage, resulting in the deterioration of mechanical properties and affecting the operation of aero-engine. Therefore, the mechanical behavior modelilng of microstructure evolution related superalloys has become an international research hotspot. The project proposed here, aiming at the problems of unclear relationship between time-dependent multi-mode microstructure damage and macro deformation behavior and difficulty in establishing explicit mathematical and physical description during service stage, the constitutive modelling method of coupling multi-mode microstructure damage DZ125 superalloy is studied by using long short memory neural network (LSTM) time-dependent neural network theory. Firstly, based on the dominant precipitates damage, the multi-mode microstructure damage is decoupled, and the constitutive parameter training set is established. In addition, the flow equation coupled LSTM model describing long term evolution is constructed. Finally, the unified-viscoplastic Chaboche constitutive model coupled with multi-mode microstructure damage time related evolution is constructed, and the accuracy is verified. Compared with the traditional phenomenological constitutive model, the results of the project proposed will be able to comprehensively consider the influence of multi-mode microstructure damage and time-dependent evolution of superalloy during service, which is more reasonable. It is expected to provide theoretical support for the performance evaluation of engine turbine blades during service in our country.
涡轮叶片在高温严苛工况下长期服役易诱发多种微观损伤,导致力学性能退化,进而影响整机运行的安全性与可靠性。针对DZ125高温合金涡轮叶片服役过程中时间相关多模式微观损伤与宏观变形行为关系不明确、难以建立显式数学物理描述的技术难题,本研究基于时间相关长短记忆神经网络(LSTM)理论,开展了耦合多模式微观损伤的DZ125合金本构建模方法研究。研究首先基于主控强化相损伤机制,实现多组织损伤解耦,构建本构参数训练集;其次,建立耦合LSTM模型的长时服役流动方程;最终提出耦合多模式组织损伤时序演化的Chaboche统一粘塑性本构模型,并通过试验验证其精度。项目开发的LSTM-UMAT计算子程序显著提升了本构模型的预测精度与计算效率。与传统唯象本构理论相比,该成果充分考虑了高温合金服役期间多模式微观损伤及其时间相关演化特征,具有较高的理论合理性与工程应用价值。研究成果已成功应用于航空发动机涡轮叶片的微观组织损伤预测与性能评估,为我国发动机涡轮叶片服役性能评估提供了重要的理论方法支撑。基于本项目研究成果已发表学术论文6篇,其中中科院Top期刊3篇,并申请国家发明专利4项。
国内基金
海外基金