【作者】 王禹桥；

【导师】 李威；

【作者基本信息】 中国矿业大学 ， 机械电子工程， 2012， 博士

【摘要】 我国的城市轨道交通目前正处于大规模建设的高峰时期，研究地铁杂散电流分布规律并实现地铁杂散电流腐蚀监测是城市轨道交通建设和运营中的关键课题，对城市轨道交通健康发展具有重要的理论意义和实际应用价值。论文研究的主要内容包括：单机车不同运行工况和多机车同时运行两种复杂工况下杂散电流分布建模，回流系统参数对杂散电流分布的耦合约束界定；研究杂散电流的腐蚀机制及检测方法，采用时频联合分析算法实现腐蚀信号的特征提取及处理；基于径向基函数神经网络的杂散电流腐蚀状态预测；基于WEB的杂散电流腐蚀智能监测系统的工程实践。在传统地铁单机车杂散电流分布静态模型的基础上，建立单机车不同运行工况的杂散电流动态解析模型。在完备的边界条件下，以机车运行位置为分界点，将两牵引变电所组成的供电区间在分界点上分成互连的不同分析域；确定机车取流值与不同变电所回流值之间的映射机制，进行地铁机车牵引计算；模拟单机车运行的不同工况，并基于解析模型进行杂散电流分布规律仿真研究。结果表明动态解析模型能够有效反应地铁现场实测结果。建立多机车同时运行的杂散电流分布有限单元模型。基于分布模型中导纳矩阵的对称性和正定性，提出采用不完全乔列斯共轭梯度算法进行解算，改善了导纳矩阵求解的收敛速度；对地铁机车等效电导进行恒功率迭代，在满足机车运行功率与牵引变电所输出功率匹配下，研究双机车同时处于牵引状态及同时处于制动状态、单机车处于牵引状态及单机车处于制动状态、牵引变电所外存在机车取流等不同工况下的杂散电流分布。研究地铁杂散电流腐蚀机制及检测方法，并对杂散电流腐蚀信号进行平稳性检验；基于短时傅立叶变换、连续小波变换及S变换，提取不含噪声的杂散电流腐蚀信号的时间及频率成分，对非平稳性杂散电流腐蚀信号进行时频联合分析。结果表明，S变换的时间分辨率及频率分辨率优于其他两种方法；将S变换用于含噪声的杂散电流腐蚀信号分析，同样能够在噪声中准确的分辨出异常信号的时间及频率成分。以埋地金属结构的极化电位偏移值作为杂散电流腐蚀表征参数，实验研究埋地金属结构与钢轨之间的水平净距、深度、周围土壤电阻率、电源正负极间距及电源电压和极化电位偏移值的关系，提出基于径向基函数神经网络的杂散电流腐蚀预测模型。采用杂散电流腐蚀模拟实验数据作为预测模型的输入输出样本集，基于次胜者受罚的竞争学习算法对输入输出样本集聚类，聚类数目作为预测模型隐层节点的数目，针对传统径向基函数神经网络结构参数学习训练存在的不足，利用改进粒子群算法和自适应遗传算法实现预测网络结构参数的优化，建立腐蚀预测模型性能评价体系。预测结果表明，与自适应遗传算法优化过的径向基函数神经网络和传统径向基函数神经网络相比较，经改进粒子群算法优化过的径向基函数神经网络在收敛精度和预测性能上更优，能够有效的实现杂散电流腐蚀状态的预测。论文构建基于WEB的杂散电流腐蚀智能监测系统，介绍了该系统在地铁现场的应用实践。本论文有图74幅，表28个，参考文献201篇。更多还原

【Abstract】 Nowadays Urban rail transit is in large-scale construction in our country. As metrostray current distribution and stray current corrosion monitoring is a key subject in theconstruction and operation of urban rail transit, this dissertation researches on metrostray current distribution and stray current corrosion monitoring which is importanttheoretically and practically for healthy development of urban rail transit.The content of this dissertation is presented as following. Stray current distributionmodeling in two conditions:1) different operation of single metro locomotive,2)simultaneous running of multi-metro locomotives; the coupling constraint definition tothe stray current of reflux systems parameters; the corrosion principle and the detectingmethods of stray current studying, as well as corrosion abnormal signal of stray currentbased on time-frequency joint analyzing, meanwhile, stray current corrosion degree byRadial Basis Function(RBF) predicting, and intelligent monitoring system of straycurrent corrosion based on WEB application.Based on stray current distribution static model of single metro locomotive，firstlythis dissertation built stray current distribution dynamic analytical model in differentoperation conditions of single metro locomotive, taking the locomotive running positionas the cut-off point, we divided the supply sector into different analysis domaininterlocking at the cut-off point under complete boundary condition, and computed themetro locomotive traction after determined relationship between locomotive takingcurrent and the backflow values. Then based on analytical model we simulated straycurrent distribution under single metro locomotive. Research results shows that thedynamic analytical model is consistent with actual measured value..Secondly this dissertation established finite element model of stray currentdistribution in simultaneous running of multi-metro locomotives. Based on symmetryand positive definiteness of admittance matrix, incomplete Joe Arguelles conjugategradient algorithm was put forward, which improves convergence rate of the model.Conducting constant power iteration of the locomotive equivalent conductance to satisfymatching requirements between the locomotive running power and traction substationoutput, stray current distribution was studied in different operatingconditions:1)simultaneous traction state and braking state of double metrolocomotive,2)the traction and braking state of single locomotive,3)presence oflocomotive taking current outside traction substation. Thirdly this dissertation researched metro stray current corrosion principle anddetecting methods, and proposed stationary test of stray current corrosion signal; Basedon short-time Fourier transformation, continuous wavelet transformation and Stransformation, time and frequency components of the stray current corrosion signalwithout noise was extracted, simultaneously time-frequency joint analysis of straycurrent corrosion abnormal signal was completed. Research results show that frequencyand time resolution of S transformation are superior to other two methods, timeresolution of S transformation accurately distinguishes the time and frequencycomponents of abnormal signal from noise using S transformation.Next this dissertation proposed a stray current corrosion model to predictpolarization potential offset value of buried metal based on RBF neural network, as theexperimental results show polarization potential offset value of buried metal is effectedby the following parameters: horizontal distance from buried metal to the rail, depth, thesurrounding soil resistivity, positive and negative pole spacing of power supply andsupply voltage. Taking simulation experiment data of the stray current corrosion as theinput/output samples in the prediction model, we gathered input/output samples based onrival penalized competitive learning method, taking the number of clustering asconcealed layer node number in the predictive model, combining deficiency existed inthe traditional RBF neural network structure parameter study, and then putt forwardimproved particle swarm optimization and adaptive genetic algorithm to implementoptimization of predictive network structure parameters, finally performance evaluationsystem of corrosion prediction model is established. The prediction results show thatRBF neural network optimized by improved particle swarm optimization in convergenceaccuracy and predictive performance is superior to the adaptive genetic algorithm andtraditional RBF neural network, which is effective in the prediction of the stray currentcorrosion.From above, this dissertation constructed the intelligent monitoring system of straycurrent corrosion based on WEB and introduced application of the system in metro.There are74figures,28tables,201references in the dissertation.更多还原