AI 服务器硬件故障模式识别与可靠性评估研究

邓杰谦

信息科学与工程研究 ›› 2026, Vol. 7 ›› Issue (7) : 41 -44.

PDF (1305KB)
信息科学与工程研究 ›› 2026, Vol. 7 ›› Issue (7) : 41 -44. DOI: 10.12349/iser.v7i7.11403

AI 服务器硬件故障模式识别与可靠性评估研究

作者信息 +

Research on Hardware Failure Mode Recognition and Reliability Assessment for AI Servers

Author information +
文章历史 +
PDF (1335K)

摘要

针对AI服务器硬件故障频发、运维被动响应的现状,本文构建基于FMEA的硬件故障模式识别框架,归纳GPU、显存、散热、电源、互联通信五类故障模式,通过计算风险优先数(RPN)对各类故障进行量化风险排序;利用真实监测数据统计各组件实际故障率,建立串并联可靠性评估模型,以RPN作为权重修正系数,进一步优化维护决策优先级.实验结果表明,所提框架能有效识别高风险故障模式,可靠性评估结果与实测数据吻合良好,为AI服务器主动运维提供定量依据.

Abstract

In response to the frequent hardware failures and passive maintenance practices in AI servers, this paper proposes a failure mode recognition framework based on FMEA. It identifies five major failure modes-GPU, video memory, thermal managementheat dissipation, power supply, and interconnect communication-and quantifies their risk levels using Risk Priority Number(RPN) for prioritization. By analyzing real-time monitoring data, actual failure rates of each component are statistically determined, and a reliability assessment model combining series and parallel configurations is established. The RPN values are used as weighting correction factors to further optimize maintenance priority decisions. Experimental results demonstrate that the proposed framework effectively identifies high-risk failure modes, and the reliability assessment outcomes closely match real-world measurements, providing quantitative support for proactive AI server maintenance.

关键词

AI服务器 / 硬件故障模式 / FMEA / 可靠性评估 / MTBF / RPN

Key words

AI server / hardware failure mode / FMEA / reliability assessment / MTBF / RPN

引用本文

引用格式 ▾
邓杰谦. AI 服务器硬件故障模式识别与可靠性评估研究[J]. 信息科学与工程研究, 2026, 7(7): 41-44 DOI:10.12349/iser.v7i7.11403

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

张文波, 贾孟君, 明亮. 基于EfficientNet的服务器硬盘故障识别[J]. 广播电视信息, 2025, 32(10): 106-109.

[2]

李逍, 张亚强, 史宏志. 服务器故障注入技术综述:实现、优化与评估[J]. 计算机应用, 2025, 45(S1): 144-153.

[3]

王宁. 云计算数据中心服务器检测维修模块化教学的实践路径[J]. 信息与电脑, 2025, 37(04): 203-205.

[4]

李盛新, 叶丰华, 李道童, 等. 基于IQPSO-GA优化ANFIS模型的服务器故障预警方法[J]. 计算机测量与控制, 2024, 32(04): 37-45.

[5]

刘伟峰. 基于机器学习的大规模并行计算机系统硬件故障检测方法[J]. 技术与市场, 2022, 29(06): 51-52+55.

AI Summary AI Mindmap
PDF (1305KB)

7

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉