用微调的大型语言模型进行故障事件采矿:分析美国核电厂故障事件报告的案例研究
Sai Zhang1, Shahidur Rahoman Sohag2, Min Xian2
1Regulatory Research & Plant Optimization Division, Idaho National Laboratory, Idaho Falls, Idaho, USA.
Risk analysis : an official publication of the Society for Risk Analysis
|February 20, 2026
概括
本研究引入了一种新的大型语言模型 (LLM) 方法,用于从故障事件报告中自动提取因果关系. 该方法有效地识别了因果关系,改善了高可靠性行业的故障分析.
科学领域:
- 工程 工程师 工程师 工程师
- 计算机科学 计算机科学
- 数据科学数据科学数据科学
背景情况:
- 失败事件叙述为失败的启动和传播提供了关键的见解.
- 分析事件因果关系有助于理解故障物理,并利用非故障数据.
- 由于复杂的语言和有限的注释数据集,自动因果关系提取面临挑战.
研究的目的:
- 开发一种基于大型语言模型 (LLM) 的新方法,用于从文本中自动提取因果关系.
- 为了利用LLM的能力来理解复杂的语言模式和远程背景.
- 从故障事件报告中准确地提取因果对.
主要方法:
- 提出了一个双重框架,包括因果关系的检测和因果关系的提取.
- 训练了一个深度学习模型来识别含有因果关系的文本.
- 一个基于T5的LLM (T5-CE) 已经开发出来,用于提取因果对.
- 美国核电厂故障事件报告的大型注释数据集被用于培训和评估.
主要成果:
- 提出的基于LLM的方法证明了有效地检测隐含和嵌入的因果关系.
- 这些模型在识别文本中的因果对方面取得了强的表现.
- 包括精度,回忆和F1分数在内的评估指标被用于评估模型性能.
结论:
- 这种新的LLM方法显著提高了从故障事件叙述中自动提取因果关系.
- 这种方法可以通过准确识别因果关系来改善故障分析.
- 该方法为具有有限故障数据的高可靠性行业提供了有价值的工具.
相关概念视频
Nuclear Fission
9.5K
Many heavier elements with smaller binding energies per nucleon can decompose into more stable elements that have intermediate mass numbers and larger binding energies per nucleon—that is, mass numbers and binding energies per nucleon that are closer to the “peak” of the binding energy graph near 56. Sometimes neutrons are also produced. This decomposition of a large nucleus into smaller pieces is called fission. The breaking is rather random with the formation of a large...
9.5K
Nuclear Power
7.5K
Controlled nuclear fission reactions are used to generate electricity. Any nuclear reactor that produces power via the fission of uranium or plutonium by bombardment with neutrons has six components: nuclear fuel consisting of fissionable material, a nuclear moderator, a neutron source, control rods, reactor coolant, and a shield and containment system.
Nuclear Fuels
Nuclear fuel consists of a fissile isotope, such as uranium-235, which must be present in sufficient quantity to provide a...
Nuclear Fuels
Nuclear fuel consists of a fissile isotope, such as uranium-235, which must be present in sufficient quantity to provide a...
7.5K
Nuclear Transmutation
12.9K
Nuclear transmutation is the conversion of one nuclide into another. It can occur by the radioactive decay of a nucleus, or the reaction of a nucleus with another particle. The first manmade nucleus was produced in Ernest Rutherford’s laboratory in 1919 by a transmutation reaction, the bombardment of one type of nuclei with other nuclei or with neutrons. Rutherford bombarded nitrogen-14 atoms with high-speed α particles from a natural radioactive isotope of radium and observed...
12.9K
Mismatch Repair
5.4K
Organisms are capable of detecting and fixing nucleotide mismatches that occur during DNA replication. This sophisticated process requires identifying the new strand and replacing the erroneous bases with correct nucleotides. Mismatch repair is coordinated by many proteins in both prokaryotes and eukaryotes.
The Mutator Protein Family Plays a Key Role in DNA Mismatch Repair
The human genome has more than 3 billion base pairs of DNA per cell. Prior to cell division, that vast amount of genetic...
The Mutator Protein Family Plays a Key Role in DNA Mismatch Repair
The human genome has more than 3 billion base pairs of DNA per cell. Prior to cell division, that vast amount of genetic...
5.4K
Types of Errors: Detection and Minimization
8.7K
Error is the deviation of the obtained result from the true, expected value or the estimated central value. Errors are expressed in absolute or relative terms.
Absolute error in a measurement is the numerical difference from the true or central value. Relative error is the ratio between absolute error and the true or central value, expressed as a percentage.
Errors can be classified by source, magnitude, and sign. There are three types of errors: systematic, random, and gross.
Systematic or...
Absolute error in a measurement is the numerical difference from the true or central value. Relative error is the ratio between absolute error and the true or central value, expressed as a percentage.
Errors can be classified by source, magnitude, and sign. There are three types of errors: systematic, random, and gross.
Systematic or...
8.7K
Investigation of Disease Outbreaks
74
Multistate foodborne outbreaks pose significant public health risks and require meticulous investigation to identify sources and implement control measures. The Centers for Disease Control and Prevention (CDC) utilizes a dynamic seven-step process for these investigations, integrating data from laboratories, interviews, and environmental assessments to protect public health.Outbreak Detection: The detection of multistate outbreaks typically begins with PulseNet, the CDC's national laboratory...
74


