复杂工况下轴承微弱故障信号的非均匀稀疏采样与感知方法研究

Weak bearing fault detection under complicated conditions is a key factor to the success of bearing early faults diagnosis and to guarantee the safe and stable operation of rotating equipment. To solve problems about big monitoring data and weak fault characteristics during the condition monitoring of rolling bearing, based on the newly development compressive sensing theory and broke through the sampling limit of Shannon theorem, the present proposal built a new sparse representation framework of bearing fault vibration based on peak features, on the basis of which a weak feature enhancement method is built making use of noise, and also developed a new non-uniform sparse sampling method with adjustment ability to capture weak fault information. Then the sparse representation and sampling method are integrated to be a sensing matrix. Finally a direct detection method for weak bearing early faults based on compressed sampling is built. Compared with conventional methods, the present proposal can complete the signal sampling and data compression at the same time, promote the weak signal feature enhancements during the process of sparsification, and do fault detection with reconstruction process simultaneously. It can directly detect fault feature without signal’s recovery, which brings a new insight to weak bearing fault detection under complex condition and has important scientific significance. At the same time, it can also reduce the burden of vibration signal acquisition, save the cost of intermediate signal processing and improve the accuracy and speed of fault detection, which also has important application value.

复杂工况下轴承微弱故障的检测是影响轴承早期故障诊断成功率和保障旋转设备安全稳定运行的关键和难点。针对其中存在的监测数据量大和故障特征微弱等问题，本项目基于新近发展的压缩感知理论，突破香农采样定理限制，建立了基于轴承故障信号峰值特性的多尺度稀疏表示模型和利用噪声的微弱特征增强方法，发展了新的具有微弱信息捕捉能力和假频抑制能力的非均匀稀疏采样方法。在此基础上构建了轴承故障信号压缩感知矩阵，发展了基于压缩采样的轴承微弱故障特征直接检测方法。与常规方法相比，本项目实现了信号采样与数据压缩同时完成,稀疏表示促进弱信号特征增强，重构过程与故障检测同步进行的综合处理方法，无需完成信号重构就可以直接实现故障特征检测,不但为复杂工况轴承微弱故障检测提供了一种全新的思路，具有重要的科学意义，同时也可以减少振动信号采集负担，节约中间信号处理成本，提高故障检测精度与速度，具有重要的应用价值。

本项目以复杂工况下轴承早期微弱故障特征的有效检测为目标，针对影响轴承早期故障诊断成功率的关键和难点问题，特别是状态监测数据量大和故障特征微弱等难题，开展了基于压缩感知理论的轴承故障信号非均匀稀疏采样与感知方法研究。基于信号处理领域新近发展的压缩感知理论，建立了基于峰值变换的振动信号稀疏表示与特征增强方法，发展了基于可变品质因子的振动信号稀疏分解与特征检测方法，有效实现了振动信号噪声压制与特征增强；建立了基于峰值保持的轴承振动信号非均匀稀疏降采样方法，大幅降低了采样率要求并保留了足够有效信息；建立了基于谐波检测和可变品质因子两种故障识别方法，构建了基于压缩采样数据的轴承微弱故障特征直接检测方法。与常规方法相比，本项目实现了振动信号采样与数据压缩同时完成、稀疏表示促进弱信号增强、重构过程与故障检测同步进行的综合处理方法，突破了香农采样定理限制，减少了振动信号采集负担，节约了中间信号处理成本，提高了故障检测精度与速度，实现了轴承早期微弱故障的快速有效检测，为复杂工况轴承微弱故障检测提供了一种全新的思路，具有重要的理论意义和应用价值。