证据推理法
可靠性(半导体)
组分(热力学)
数据挖掘
计算机科学
可靠性工程
登普斯特-沙弗理论
变量(数学)
过程(计算)
人工智能
工程类
决策支持系统
数学
功率(物理)
商业决策图
物理
量子力学
操作系统
热力学
数学分析
摘要
Abstract Expert knowledge is an important information source for reliability assessment of those systems with limited time‐to‐failure data. For a better understanding of the degradation profiles of systems, multiple experts are oftentimes invited to express their judgments and the associated uncertainties on the reliability‐related measures of the system. Evidential variables, as an alternative of uncertainty quantification, have been extensively used in expert systems to quantify the epistemic uncertainty of the elicited expertise. When eliciting the reliability‐related information by evidential variables, experts only need to express the possible ranges of reliability‐related measures and their associated probabilities. Such a type of information well caters to the experts’ elicitation process. In this article, an evidential network (EN)‐based reliability assessment method is put forth by fusing multi‐source evidential information. The proposed method mainly contains three steps. In the first place, the multi‐source evidential information related to all components is elicited from experts in the form of evidential variables. Next, the evidential variable of the component reliability is assessed via a constrained optimization model by treating all pieces of multi‐source evidential information as constraints. The component reliability results are transformed into pieces of mass functions of the components’ states under the theory of belief functions. The system reliability‐box and the reliability‐box over time are, therefore, calculated by the EN model by inputting all mass functions of components’ states. A pipeline system and a chip cutting system are exemplified to examine the effectiveness of the proposed method.
科研通智能强力驱动
Strongly Powered by AbleSci AI