铁道标准设计
鐵道標準設計
철도표준설계
RAILWAY STANDARD DESIGN
2015年
7期
147-151,152
,共6页
故障诊断%贝叶斯网络%模糊数学%决策理论
故障診斷%貝葉斯網絡%模糊數學%決策理論
고장진단%패협사망락%모호수학%결책이론
Fault diagnosis%Bayesian Network%Fuzzy math%Decision-making theory
高速列车长时间工作运行不可避免发生故障,其中的车载设备故障发生具有不确定性和相关性特性。分析故障发生和查找故障致因是一项复杂而重要的过程和工作。贝叶斯网络在解决不确定性和相关性问题有其独特的优势,可以利用贝叶斯网络查找故障成因和故障点并且加以控制,从而提高列控车载设备运行的安全性。在实际理论和工程实践中运用贝叶斯推理时往往过分强调贝叶斯推理功能而忽略先验概率确定的问题,模糊贝叶斯决策是在经典贝叶斯理论上综合运用决策理论、模糊数学、贝叶斯方法和期望效用理论,建立一个在多属性指标下故障态势的模糊贝叶斯决策模型,克服了单一贝叶斯推理的不足,增强了模型的适用性。
高速列車長時間工作運行不可避免髮生故障,其中的車載設備故障髮生具有不確定性和相關性特性。分析故障髮生和查找故障緻因是一項複雜而重要的過程和工作。貝葉斯網絡在解決不確定性和相關性問題有其獨特的優勢,可以利用貝葉斯網絡查找故障成因和故障點併且加以控製,從而提高列控車載設備運行的安全性。在實際理論和工程實踐中運用貝葉斯推理時往往過分彊調貝葉斯推理功能而忽略先驗概率確定的問題,模糊貝葉斯決策是在經典貝葉斯理論上綜閤運用決策理論、模糊數學、貝葉斯方法和期望效用理論,建立一箇在多屬性指標下故障態勢的模糊貝葉斯決策模型,剋服瞭單一貝葉斯推理的不足,增彊瞭模型的適用性。
고속열차장시간공작운행불가피면발생고장,기중적차재설비고장발생구유불학정성화상관성특성。분석고장발생화사조고장치인시일항복잡이중요적과정화공작。패협사망락재해결불학정성화상관성문제유기독특적우세,가이이용패협사망락사조고장성인화고장점병차가이공제,종이제고렬공차재설비운행적안전성。재실제이론화공정실천중운용패협사추리시왕왕과분강조패협사추리공능이홀략선험개솔학정적문제,모호패협사결책시재경전패협사이론상종합운용결책이론、모호수학、패협사방법화기망효용이론,건립일개재다속성지표하고장태세적모호패협사결책모형,극복료단일패협사추리적불족,증강료모형적괄용성。
Long time operation of the high-speed train is likely to encounter some failures, and the failures of on-board equipment are of uncertainty and correlated. Analysis of the failures and there causes is a complicated and important process. Bayesian network has unique advantages to deal with the uncertainty and correlation. Bayesian network can be used to look for and control failure causes and failure points to improve the safety of train operations. The practical theory of and engineering practice in using Bayesian for inference often put too much emphasis on functions of Bayesian inference and ignore the prior probability. Fuzzy Bayesian decision-making is based on classical Bayesian theory integrates the use of decision theory, fuzzy mathematics, the Bayesian method and expected utility theory to establish a fuzzy Bayesian decision-making model with fault situation under multiple attribute index, which overcomes the deficiency of the single Bayesian inference, and enhances the applicability of the model.