铁道标准设计
鐵道標準設計
철도표준설계
RAILWAY STANDARD DESIGN
2015年
2期
118-121
,共4页
铁路道岔%转辙机%故障诊断%遗传算法%结构学习%贝叶斯网络模型
鐵路道岔%轉轍機%故障診斷%遺傳算法%結構學習%貝葉斯網絡模型
철로도차%전철궤%고장진단%유전산법%결구학습%패협사망락모형
Rail switch%Switch machine%Fault diagnosis%Genetic algorithm%Structure learning%Bayesian network model
铁路系统转辙机维修方式仍沿用故障修模式，无法提高故障排除速度和准确性，提出利用改进遗传算法优化贝叶斯网络的方法建立故障诊断模型。利用遗传算法搜索能力强，不易陷入局部最优的特点，采用连接矩阵代替网络结构的编码方式，通过修改适应度函数、更新遗传操作方式、修正非法图等方法改进遗传算法，最终解决贝叶斯网络结构学习算法容易缩小搜索空间及易陷入局部最优的缺点。最后利用标准Asia网络验证本文算法比K2和GA算法有更好的搜索结果和更快的收敛速度，以道岔失去表示故障为例验证改进算法对转辙机故障诊断的优越性。
鐵路繫統轉轍機維脩方式仍沿用故障脩模式，無法提高故障排除速度和準確性，提齣利用改進遺傳算法優化貝葉斯網絡的方法建立故障診斷模型。利用遺傳算法搜索能力彊，不易陷入跼部最優的特點，採用連接矩陣代替網絡結構的編碼方式，通過脩改適應度函數、更新遺傳操作方式、脩正非法圖等方法改進遺傳算法，最終解決貝葉斯網絡結構學習算法容易縮小搜索空間及易陷入跼部最優的缺點。最後利用標準Asia網絡驗證本文算法比K2和GA算法有更好的搜索結果和更快的收斂速度，以道岔失去錶示故障為例驗證改進算法對轉轍機故障診斷的優越性。
철로계통전철궤유수방식잉연용고장수모식，무법제고고장배제속도화준학성，제출이용개진유전산법우화패협사망락적방법건립고장진단모형。이용유전산법수색능력강，불역함입국부최우적특점，채용련접구진대체망락결구적편마방식，통과수개괄응도함수、경신유전조작방식、수정비법도등방법개진유전산법，최종해결패협사망락결구학습산법용역축소수색공간급역함입국부최우적결점。최후이용표준Asia망락험증본문산법비K2화GA산법유경호적수색결과화경쾌적수렴속도，이도차실거표시고장위례험증개진산법대전철궤고장진단적우월성。
As the fault repair mode still used in repairing switch in railway system is unable to improve the speed and accuracy of troubleshooting, a fault diagnostic model is established with improved genetic algorithm and Bayesian network. Characterized by genetic algorithm, strong searching capability and independence of local optimal, it replaces the network structure coding with connection matrix to improve genetic algorithm by modifying fitness function, updating genetic operation mode and correcting illegal map and, consequently, overcomes the shortcomings of Bayesian network structure that learns algorithm and tends to reduce the search space and falls into local optimal. Finally, the algorithm is verified with standard Asia network to be faster in convergence speed compared with the algorithm of K2 and GA and the superiority of improved algorithm for fault diagnosis of switch is demonstrated by taking the case of losing fault indication.
