改进的道岔智能故障诊断系统建模研究

Research on Modeling Improved Intelligent Switch Fault Diagnostic System

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。

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.