城轨列车轮对轴箱并发故障解耦与状态辨识研究

Urban rail trains run has its particularity: the frequent load changes, start and stop cycle is short, the source of interference and more speed fluctuations, impact load, which can easily lead to bruising round, flat scar, peeling, grinding round loss, concurrent axle box wear, cracks, gluing and other failures. Wheel axle box concurrent fault feature intermixed rendering more coupling, and other signs of fuzziness, fault identification to bring great difficulties, it carries out research to identify urban rail train wheel axle box and decoupling the concurrent fault state, to achieve the wheel axle box accurate fault location, saving maintenance time and cost of great significance. The project first analyzes the wheel axle box working mechanism and failure modes, internal and external coupling incentive to build the next round of the axle box concurrent fault dynamics model; secondly, analyzes the wheel urban rail train running on the axle box particularity under concurrent fault feature, in-depth understanding of the intrinsic relationship between the fault and the external signs, reveal the coupling mechanism of concurrent fault; Finally, experimental verification, multi-dimensional time-frequency domain to extract the wheel axle box arrangement state data entropy, attribute mapping algorithm using perturbation frequency domain permutation entropy feature vector dimensionality reduction, according to the timing algorithm concurrent fault probability decoupling and precipitation, the use of subspace clustering of sparse precipitation data classification, and wheel and axle box decoupling the concurrent fault identification.

城轨列车运行具有其特殊性：载荷变化频繁、启停周期短、干扰源多、速度波动大、冲击负荷大，这些极易导致轮对擦伤、扁疤、剥落、磨损失圆，同时并发轴箱磨损、裂纹、胶合等故障。轮对轴箱并发故障特征相互混杂呈现多耦合、模糊性等征兆，给故障辨识带来极大困难，故开展城轨列车轮对轴箱并发故障解耦与状态辨识研究，实现轮对轴箱故障的准确定位，对节约维修时间和成本具有重要意义。本项目首先分析轮对轴箱的工作机理及故障模式，构建内外耦合激励下轮对轴箱并发故障动力学模型；其次，着重分析城轨列车运行特殊性下的轮对轴箱并发故障特征，深入了解故障与外部征兆间的内在关联关系，揭示其并发故障的耦合机理；最后，通过实验验证，提取轮对轴箱状态数据的多维时频域排列熵，采用扰动属性映射算法对时频域排列熵特征向量降维，根据时序概率解耦算法对并发故障进行解耦和析出，运用稀疏子空间聚类对析出数据进行分类，实现轮对轴箱并发故障解耦与辨识。

本项目针对城轨列车运行的特殊性及轮对轴箱并发故障特征的耦合性，开展城轨列车轮对轴箱并发故障解耦与状态辨识研究。经过三年研究，取得如下成果：（1）分析轮对轴箱的工作机理及故障模式，构建内外耦合激励下轮对轴箱并发故障动力学模型，并着重分析城轨列车运行特殊性下的轮对轴箱并发故障特征，深入了解故障与外部征兆间的内在关联关系，揭示了其并发故障的耦合机理；（2）提出了基于多尺度本征模态排列熵和模拟退火优化支持向量机的列车轴承故障诊断方法，并将其成功应用于列车轴承多模式混淆信号的故障诊断中；（3）提出了基于LCD-MPE和ELM-Adaboost的故障辨识方法，实现了列车轴承故障的智能辨识；（4）提出了基于深度学习模型的故障诊断算法，并将其应用于轮对轴箱并发故障诊断中，实现了轮对轴箱并发故障的自动识别。基于以上研究成果，共发表论文23篇，其中SCI/EI检索10篇；出版学术专著2部；申请国家发明专利6项，其中2项已授权。