[1]刘志卓,吕柏林,李天祥,等. 融合大语言模型与MCP协议的离心泵智能优化平台研究[J].机械与电子,2026,44(01):1-8.
　LIU Zhizhuo,LYU Bailin,LI Tianxiang,et al. Research on an Intelligent Optimization Platform for Centrifugal Pumps Integrating Large Language Models and MCP Protocol[J].Machinery & Electronics,2026,44(01):1-8.
点击复制

 融合大语言模型与MCP协议的离心泵智能优化平台研究()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年01期
页码:
1-8
栏目:
研究与设计
出版日期:
2026-01-27

文章信息/Info

Title:
 Research on an Intelligent Optimization Platform for Centrifugal Pumps Integrating Large Language Models and MCP Protocol
文章编号:
1001-2257(2026)01-0001-08
作者:
 刘志卓1; 吕柏林1; 李天祥1; 孙旭东2; 韦自建1
 1.辽宁石油化工大学机械工程学院,辽宁 抚顺 113001;2.阜新市万达铸业有限公司,辽宁 阜新 123004
Author(s):
 LIU Zhizhuo1; LYU Bailin1; LI Tianxiang1; SUN Xudong2; WEI Zijian1
 1.School of Mechanical Engineering,Liaoning Petrochemical University,Fushun 113001,China;2.Fuxin Wanda Casting Co.,Ltd.,Fuxin 123004,China
关键词:
离心泵; 大语言模型; MCP协议; 多目标优化
Keywords:
centrifugal pump; large language model; MCP protocol; multi objective optimization
分类号:
TH311;TP181
文献标志码:
A
摘要:
 针对传统离心泵设计优化方法存在多目标协同能力弱、智能化程度低和用户门槛高等问题,提出一种融合大语言模型(LLM)与模型上下文协议(MCP)的离心泵智能优化平台。该平台以GPT 4等LLM 为核心,实现了自然语言需求解析、结构参数自动映射与知识推理,极大提升了用户交互的友好性与平台智能化水平。结合NSGA II多目标遗传算法和高斯过程回归(GPR)性能预测,平台能够实现多目标协同优化,并利用GPR置信区间对优化结果进行不确定性量化和风险判别。通过MCP协议,系统与企业级数据库、CFD仿真平台等异构资源实现了标准化安全集成。通过典型工业泵工程案例验证,所开发平台可在极少人工干预条件下,自动获得Pareto最优解集,优化效率提升超过60%,性能预测最大误差低于1%。这一成果充分显示出平台在工业应用中的高效性、可靠性及广阔的推广前景。
Abstract:
 Traditional design optimization methods for centrifugal pump often suffer from weak multi objective coordination,low levels of intelligence,and high user barriers.To address these issues,this study proposes an intelligent optimization platform for centrifugal pumps that integrates Large Language Models(LLMs) with the Model Context Protocol (MCP).Utilizing advanced LLMs such as GPT 4 as its core,the platform enables natural language requirement parsing,automatic mapping of structural parameters,and knowledge reasoning,significantly enhancing both user interaction and platform intelligence.By integrating the NSGA II multi objective genetic algorithm and Gaussian Process Regression (GPR) for performance prediction,the platform achieves collaborative multi objective optimization,while the GPR confidence intervals are used to quantify uncertainty and assess risks in the optimization results.Through the MCP protocol,the system realizes standardized and secure integration with heterogeneous resources,including enterprise level databases and CFD simulation platforms.A typical industrial pump case is used to validation,and the results demonstrates that the developed platform can automatically obtain the Pareto optimal solution sets with minimal manual intervention.The optimization efficiency is improved by over 60%,and the maximum error in performance prediction remains below 1%.These results fully demonstrate the platform’s high efficiency,reliability,and broad application potential in industrial settings.

参考文献/References:

[1] GÜLICH J F.Centrifugal pumps[M].3rd ed.Berlin:Springer,2014.
[2] KARASSIK I J,MESSINA J P,COOPER P,et al.Pumphandbook[M].4th ed.New York:McGraw Hill Education,
2012.
[3] BENTURKI M,DIZENE R,GHENAIET A.Multi objectiveoptimization of two stage centrifugal pump using NSGA II algorithm[J].Journal of applied fluid
mechanics,2018,11(4):929-943.
[4] 刘栋.倪子建,黄凯,等.基于CFD DEM 的离心泵叶轮磨损特性研究和优化设计[J].流体机械,2022,50(9):26-32.
[5] AHMAD H M.Experimental and computational fluid dynamics (CFD) analysis of erosion in centrifugal pumps operating under HVAC conditions[J].IRE Journals,2022,5(9):529-538.
[6] DEB K,PRATAP A,AGARWAL S,et al.A fast and elitist multiobjective genetic algorithm:NSGA II[J].IEEE Transactions on evolutionary computation,2002,6(2):182-197.
[7] DAVE G S,PANDHARE A P,KULKARNI A P,et al.Innovative data techniques for centrifugal pump optimization with machine learning and AI model[J].
PLOS one,2025,20(6):e0325952.
[8] VISWANATHAN P C,VENKATESH S N,DHANASEKARAN S,et al.Deep learning for enhanced fault diagnosis of monoblock centrifugal pumps:Spectrogram based analysis[J].Machines,2023,11(9):874.
[9] AHMAD S,AHMAD Z,KIM J M,et al.A centrifugal pump fault diagnosis framework based on supervised contrastive learning[J].Sensors,2022,22(17):6448.
[10] SUN Z N,ZHENG S H,ZHANG W W,et al.A data driven segmented model based on variance information for centrifugal pump efficiency prediction[J].Engineering applications of artificial intelligence,2024,136:108992.
[11] ZHAO X,ZHANG D,ZHANG R,et al.A comparative study of Gaussian process regression with other three machine learning approaches in the performance prediction of centrifugal pump[J].Journal of mechanical engineering science,2022,236(8):3938-3949.
[12] DEB K,JAIN H.An evolutionary many objective optimization algorithm using reference point based nondominated sorting approach[J].IEEE Transactions on evolutionary computation,2014,18(4):577-601.
[13] MA C B,YANG Z R,JIAO K,et al.Multi objective optimization of centrifugal compressor impeller through coupling SVM with NSGA III[J].International journal of green energy,2021,18(13):1383-1395.
[14] NISHI Y,KONNO K,ONO S,et al.Multi objective optimization design method for single blade centrifugal pump impellers to improve performance and reduce radial thrust [J].European journal of mechanics B/fluids,2025,114:204311.
[15] WANG C N,YANG F C,NGUYEN V T T,et al.CFD analysis and optimum design for a centrifugal pump using an effectively artificial intelligent algorithm[J],Micromachines,2022,13(8):1208.
[16] 陈云飞.基于机器学习的离心泵叶轮智能优化设计平台[D].镇江:江苏大学,2022.
[17] BROWN T B,MANN B,RYDER N,et al.Language models are few shot learners[C]∥ 34th International Conference on Neural Information Processing Systems,2020:1877-1901.
[18] OpenAI.GPT 4 technical report[EB/OL].[2025-04-03].https:∥cdn.openai.com/papers/gpt 4.pdf.
[19] GÖPFERT J,WEINAND J M,KUCKERTZ D,et al.Opportunities for large language models and discourse in engineering design[J].Energy and AI,2024,
17:100383.
[20] RIOS T,MENZEL S,SENDHOFF B,et al.Large language and text to 3D models for engineering design optimization[C]∥ 2023 IEEE Symposium Series on Computational Intelligence (SSCI).New York:IEEE,2023:1704-1711.
[21] Anthropic.Model context protocol (MCP) documentation[EB/OL].[2025-04-03].https:∥docs.claude.com/en/docs/mcp.
[22] SHARMA A,KOSASIH E,ZHANG J,et al.Digital twins:state of the art theory and practice,challenges,and open research questions[J].Journal of industrial information integration,2022,30:100383.
[23] ZHAO W,YANG X,LYU Z H,et al.Road of large language model:source,challenge,and future perspectives[J].Research,2025,8:0655.
[24] FULLER A,FAN Z,DAY C,et al.Digital twin:enabling technologies,challenges and open research[J].IEEE Access,2020,8:108952-108971.
[25] ZHAO Y F,NIFORATOS E,CUSTIS T,et al.Guest editorial:special issue large language models in design and manufacturing[J].Journal of computing and information science in engineering, 2025, 25(2):020301.
[26] 徐源,翟春艳,王国良.基于对抗学习与深度估计的车辆检测系统[J],辽宁石油化工大学学报,2020,40(3):83-90.
[27] 刘志卓,吕柏林,韦自建.基于人工智能大语言模型的离心泵智能设计与性能优化系统:202411496875.5[P].2024-10-25
[28] 田文才,乔伟彪,周国峰,等.基于小波变换和深度学习的短期天然气负荷预测研究[J].辽宁石油化工大学学报,2021,41(5):91-96.

备注/Memo

备注/Memo:
收稿日期:2025-09-06
基金项目:辽宁省教育厅青年项目(LJKQZ20222277)
作者简介:刘志卓 (1999-),男,河北沧州人,硕士研究生,研究方向为流体机械智能优化;吕柏林 (1969-),男,吉林镇贵人,博士,副教授,研究方向为流体机械智能优化和故障检测,通信作者,E-mail:lvbailin@126.com。
更新日期/Last Update: 2026-03-09