深部煤层采煤机关键传动部件混叠故障解耦诊断理论研究

Harsh working environment in deep coal seam accelerates the wear process of coal cutters. A simple failure in the coal cutter would knock off the whole mining production line for a couple of days. Hence, it is imperative to monitor the machine condition to prevent break-downs. One of the challenging tasks is to diagnose faults that occur simultaneously in the same/different components of a machine. To address this issue, this project aims to propose a novel approach to decouple compound faults and develop new technologies for fault type detection and fault severity identification. .To achieve the research goal, this project adopts the following four phases: (1) In order to identify the fake fault impulse components aroused by cutting coals in deep mining seam, a new approach using nonlinearity measurement will be developed in the first phase. The identification of the occurrence of hybrid faults would be investigated using nonlinearity measurement approach..(2) Then, in the second phase, a novel approach for correlated vibration source separation will be presented. The vibration signals recorded in real word essentially are convolved from different vibration sources and in fact these sources somehow correlate with each other. To solve this problem, a new generation technology will be developed to deal with the correlated source separation issue in this project. The new technology, convolutive bounded component analysis (Convolutive BCA), is capable of separating not only the desired independent sources but also the sources that are dependent/correlated in both component (space) and sample (time) dimensions from real word vibration signals. .(3) Whereafter, a novel convolutive bounded component analysis with reference (Convolutive BCA-R) will be proposed for hybrid faults decoupling, where each vibration source corresponding to a fault type/location in the hybrid faults could be extracted from the sensor observations. This is achieved based on the BCA framework and an intrinsic frequency tracking reference construction algorithm. Herein the challenge is how to construct the reference signal for BCA to decouple the hybrid faults signal. Given typically nonlinear and nonstationary of a machine vibration signal, it is always difficult to capture accurate instantaneous frequency (IF) which depict the fault characteristics. The IF is essential for a meaningful interpretation of reference construction. To that end, a modified multivariate empirical mode decomposition (MEMD) will be employed to exploit the IFs associated with machine faults. Based on the IFs, the characteristics of typical single faults will be fully investigated and an intrinsic frequency tracking reference algorithm will be established. .(4) In the last phase, a manifold learning approach for fault severity assessment will be developed. The intrinsic fault features will be extracted to develop a fault severity indicator for the coal cutters.

群故障研究是机械故障诊断领域的难点。本项目提出深部煤层采煤机关键传动部件混叠故障解耦诊断的学术构想。研究将以准确提取故障源信号为手段，以实现将混叠故障振动信号分解成单一故障子信号集簇为目的，结合故障程度评估模型，构建混叠故障解耦诊断理论。.研究思路为：1.考虑到截割冲击等环境影响，利用非线性度量技术，辨识伪故障；2.考虑到实际信号振动源存在相关性，且以卷积形式混合，提出以Bounded Component Analysis(BCA)为理论基础的相关源分离方法，并在频域中解卷积；3.研究噪声辅助多元经验模式分解的故障瞬时频率变化规律分析方法，以构造故障源参考信号，提出参考约束卷积BCA(Convolutive BCA-R)的混叠故障解耦方法；4.利用流形学习新方法提取解耦信号的固有特征，建立故障程度评估模型。.通过本项目的研究，可望创新出混叠故障解耦诊断方法，为采煤机群故障研究提供理论基础。

群故障研究是机械故障诊断领域的难点。项目通过研究采煤机动力传动关键部件的故障机理，建立了考虑多源激励的齿轮系统动态响应数学模型；设计了典型复合故障齿轮试验测试与数据采集，分析了混叠故障解耦诊断方案，提出了基于振动信号分解的齿轮混叠故障解耦诊断技术；设计了考虑温度-负载因素的轴承实验台，通过理论计算、有限元模拟和试验验证，确定了轴承配置方案并完成样机制造；设计了典型复合故障轴承试验测试与数据采集，提出了基于轴承混叠故障解耦诊断技术。同时，项目研究了故障解耦信号定量表征，利用混沌理论获取故障子信号不同时频域特征信息，提出基于混沌因子的子故障定量识别技术。最后，项目有效结合振动信号分解与混叠指标，创新性地提出了一种盲源分离技术，解决了非线性信号以及信息关联识别难点，实现了齿轮系统混叠故障解耦诊断。.整个项目的研究成果集成了机械系统混叠故障诊断基础理论研究所涉及到的多学科关键技术，为关键机械设备的健康管理奠定了良好基础。项目执行期间项目负责人在领域主流期刊《Structural Health Monitoring》《Nonlinear Dynamics》《Renewable Energy》《Journal of Sound and Vibration》《Measurement Science and Technology》《Measurement》等以第一作者发表学术论文9篇、通讯作者4篇，全部被SCI收录，其中ESI热点论文1篇，ESI高被引论文2篇。项目执行期间与西安交通大学合作研制了轴承故障仿真实验台，项目负责人受邀担任国际会议学术委员会委员1次，担任4本行业重要SCI期刊编辑/编委。