数据驱动的金属疲劳寿命模型研究进展  

Advances in data-driven models for fatigue life prediction of metallic materials

作　　者：甘磊 吴昊[2] 仲政 GAN Lei;WU Hao;ZHONG Zheng(School of Science,Harbin Institute of Technology,Shenzhen 518055,China;School of Aerospace Engineering and Applied Mechanics,Tongji University,Shanghai 200092,China)

机构地区：[1]哈尔滨工业大学(深圳)理学院,广东深圳518055 [2]同济大学航空航天与力学学院,上海200092

出　　处：《力学进展》2025年第1期30-79,共50页Advances in Mechanics

基　　金：国家自然科学基金(11932005,12372081);深圳市深港合作项目(SGDX20230116091247009);深圳市高校稳定支持计划(GXWD20231130100351002);深圳市战略性新兴产业发展专项资金扶持计划(XMHT20220103004);广东省高校创新团队(2021KCXTD006)资助

摘　　要：金属疲劳寿命模型是开展工程结构完整性和可靠性评估的基础.传统的知识驱动模型关注疲劳机理和数理逻辑,一般具有明确的物理意义,并且可高度概括疲劳失效过程.然而,随着对结构安全性要求的日益提高以及新兴工程材料的不断涌现,传统模型在预测能力、应用场景、工程适用性等方面都逐渐显现出局限性.近年来,由人工智能赋能的数据驱动模型在金属疲劳寿命研究领域受到了广泛关注,相关研究成果正逐步应用于解决包括单轴疲劳、多轴疲劳、变幅疲劳在内的各类经典疲劳问题.数据驱动模型能够在最小化人因误差的情况下,从多变量作用中解析出对疲劳寿命的最优显\隐式表达,可揭示传统方法难以发现的失效规律,已然成为领域内新的研究热点.本文综述了当前数据驱动模型在金属疲劳寿命预测方面的研究进展,首先总结了纯数据驱动模型的一般应用流程及其应用现状,其次归纳了各类知识-数据混合驱动模型的实现方式及应用优势,最后对未来潜在研究方向及挑战进行了探讨与展望.Fatigue life models are fundamental when assessing the integrity and reliability of engineering components made of metallic materials.Hence,a plethora of domain knowledge-driven models have been developed over the past centuries,pursuing the consistency with fatigue failure mechanisms and the rationality of mathematical expressions.They generally demonstrate physical significance and can describe the complex processes of fatigue damage evolution explicitly and comprehensively.However,with the increasing demand for the operational safety of critical components and high-performance structural materials emerging constantly,they are facing limitations in the aspects of predictive capability,application scope,and engineering practicality.As an alternative,data-driven models,under the impetus of Artificial Intelligence tide,have attained growing attention and found increasing applications in life-prediction issues under various loading patterns.Data-driven models feature their powerful ability to derive optimal explicit/implicit relationships between fatigue life with numerous influential factors,without suffering from human errors.Moreover,they can quickly discover the physical laws governing fatigue failure which are difficult to be clarified by domain knowledge-driven models.Nowadays,data-driven models are recognized as opening a new pathway for fatigue damage analysis and life prediction,being a hot spot in fatigue research.This paper reviews the progress of research in developing data-driven models for predicting the fatigue life of metallic materials.Different types of data-driven models,including pure data-driven models and knowledge informed data-driven models,are summarized,along with their distinct construction methodologies and application advantages.The future prospects and challenges in this field are also discussed.

关 键 词：疲劳寿命预测 金属 数据驱动模型 知识-数据混合驱动模型 

分 类 号：O346.2[理学—固体力学]