| 阎军,刘洪沅,王骏杰,蒋官儒,严梓睿,聂英豪,阎琨.DeepSeek辅助计算结构力学研究的可行性分析及探索[J].计算力学学报,2025,42(2):189~195 |
|
| 
码上扫一扫！ |
| DeepSeek辅助计算结构力学研究的可行性分析及探索 |
| Feasibility analysis and exploration of DeepSeek-assisted computational structural mechanics research |
| 投稿时间：2025-02-26 修订日期：2025-03-22 |
| DOI：10.7511/jslx20250226001 |
| 中文关键词: DeepSeek 计算结构力学 人工智能 点阵结构 有限元聚类分析 |
| 英文关键词:DeepSeek computational structural mechanics artificial intelligence lattice structures FCA |
| 基金项目:国家自然科学基金(12432005;12472116)资助项目. |
|
| 摘要点击次数: 22 |
| 全文下载次数: 18 |
| 中文摘要: |
| 计算结构力学是现代工程科学的核心支撑学科之一,但其专业门槛高、程序实现复杂的特点,导致研究者需投入大量时间进行前期知识储备与算法开发训练。大语言模型的语义解析、知识整合、代码生成与算法推导等能力,为破解上述问题提供了一种范式革新的技术路径。基于此,本文针对DeepSeek-R1大语言模型辅助计算结构力学研究的可行性与局限性开展了评估分析。首先以有限元聚类分析这一新兴计算结构力学方法为研究对象,测试了基于大语言模型完成其关键环节程序开发的可行性,分析了语义交互策略、专业术语完备等因素对Deepseek-R1输出有效性的影响,并对比了相同输入信息下不同大语言模型的性能。其次,本文基于若干典型计算结构力学问题测试了不同大语言模型输出结果的有效性。最后,本文基于数值实例经验,浅析了以DeepSeek为代表的新一代人工智能模型辅助计算结构力学研究的可行性与局限性。测试结果显示,DeepSeek-R1等大语言模型在降低计算结构力学研究门槛、提升算法开发效率方面展现出优势,但用于复杂问题时仍需理论基础和程序经验支撑。 |
| 英文摘要: |
| Computational structural mechanics is one of the core supporting disciplines of modern engineering science.However,due to its high professional threshold and complex program implementation,researchers need to invest a lot of time in pre-requisite knowledge reserve and algorithm development training.The capabilities of semantic analysis,knowledge integration,code generation and algorithm derivation of the large language model(LLM) provide a technical path for paradigm innovation to solve the above problems.Based on this,this paper evaluates and analyzes the feasibility and limitations of DeepSeek-R1 LLM-assisted computational structural mechanics research.Firstly,taking the Finite Element Method-Cluster based Analysis(FCA),a new computational structural mechanics method,as the research object,the feasibility of completing the key link program development based on the LLM is tested.The influence of semantic interaction strategy,terminology completeness and other factors on the effectiveness of Deepseek-R1 output is analyzed,and the performance of different large language models under the same input information is compared.Secondly,this paper tests the effectiveness of the output results of different LLMs based on several typical computational structural mechanics problems.Finally,based on the experience of numerical examples,this paper analyzes the feasibility and limitations of the new generation of artificial intelligence(AI) models represented by DeepSeek to assist the study of computational structural mechanics.The test results show that LLMs such as DeepSeek-R1 have shown lowering in lowering the research threshold of computational structural mechanics and improving the efficiency of algorithm development,but they still need theoretical basis and program experience support for complex problems. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
|
