齿轮早期微小故障的综合诊断方法研究

项目名称： 齿轮早期微小故障的综合诊断方法研究

项目编号： No.51505353

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 机械、仪表工业

项目作者： 曲永志

作者单位： 武汉理工大学

项目金额： 21万元

中文摘要： 齿轮的故障诊断对于一些关键性场合，例如大型风力机、直升机、核电站等，具有非常重要的意义。本项目针对目前在齿轮故障诊断中难以检测的初期微小故障，利用光纤光栅传感器对应变和温度的直接测量方法，结合传统的振动分析手段，提出齿轮微小故障识别与诊断的新原理和新方法。主要研究内容为：通过齿轮建模与仿真建立故障-现象的物理关系模型，揭示不同齿轮故障对传动过程各参数的影响，构建齿轮早期微小故障诊断的理论依据；研究基于分布式光纤光栅应变与温度同时测量传感过程的解耦分离，提取多测量点的应变和温度的动态分布以及由其组成的多维时序序列，实现故障直接物理变化的定量分析；通过特征分析与数据挖掘，进行负载波动的自动补偿，应变和温度时序序列异常检测，故障点的定位分析与故障模式分析，以及故障恶化趋势分析；结合振动信号分析，对比光纤光栅的传感诊断性能，建立一套多传感器融合，直接与间接测量相互支持的可靠早期齿轮故障诊断系统。

中文关键词： 齿轮；早期故障；智能诊断；光纤光栅；多变量时序序列

英文摘要： Gears diagnostics are crucial for certain key occasions as in wind turbines, helicopters, and nuclear power plants, etc. Monitoring and early warning of the incipient faults could significantly reduce catastrophic failures. This project proposes a new principle and method to analyze incipient gear failure mode and failure symptoms. The main research topics include: First, conduct gear modeling and simulation research to establish physical-phenomena model and to further reveal the deviation of parameters due to defects, construct the theoretical basis for minor fault detection; second, study the separation of coupled strain and temperature measurements from distributed optical Fiber Bragg Grating (FBG) sensors, in order to obtain the dynamic distribution of strain and temperature in multiple locations and the corresponding samples in time series manner, accomplish quantitative analysis of direct physical changes; third, using feature analysis algorithms and data mining approaches, to automatically filter the load fluctuation effect, to perform strain and temperature time series anomaly detection, to accurately determine fault location and fault type, and to analyze the deterioration trend; finally, by comparing the pros and cons of FBG sensors to vibration sensors, design a reliable initial minor fault diagnostic system, which combines both direct measurements and indirect measurements.

英文关键词： Gear;Incipient Faults;Intelligent Diagnostics;FBG;Multivariate Time Series