基于自适应参数优化RSSD-CYCBD的行星齿轮箱复合故障诊断

Compound fault diagnosis of planetary gearbox based on RSSD-CYCBD by adaptive parameter optimization

针对行星齿轮箱多振源耦合导致故障源辨识困难、较弱故障特征容易被噪声和较强故障特征掩盖,以及由传播路径引起的信号衰减导致的故障特征微弱等问题,提出一种自适应参数优化的共振稀疏分解(RSSD)和最大二阶循环平稳盲解卷积(CYCBD)的行星齿轮箱多故障耦合信号分离及诊断算法。根据轴承和齿轮故障的不同共振属性,用RSSD算法将多故障耦合信号分解为包含齿轮故障特征的高共振分量和主要包含轴承故障特征的低共振分量后,通过CYCBD算法分别对高、低分量进行解卷积,消除传播路径影响和噪声干扰,实现微弱故障特征的增强和提取。特别地,针对RSSD和CYCBD中参数优化困难、依赖人工经验和自适应差等问题,使用基于松鼠算法(SSA)对参数进行自适应优化选取,设计了融合包络谱峭度、自相关函数最大值均方根和特征频率比在内的复合指标作为优化目标。对解卷积后的信号进行包络解调提取故障特征频率,识别不同故障源。通过行星齿轮箱多故障模拟信号和实测信号验证了所提算法的有效性和可行性,进一步地,将所提算法集成在边缘计算设备中,为行星齿轮箱等旋转机械的状态检测诊断及远程运维提供解决方案。

The coupling of multiple vibration sources of planetary gearboxes results in difficulty in identifying fault sources,and weak fault features are easily masked by noise and strong fault features.In addition,signal attenuation caused by the propagation path causes weak fault features.To address these issues,a multi-fault coupled signal separation and diagnosis method for planetary gearboxes utilizing resonance-based sparse signal decomposition(RSSD)by adaptive parameter optimization and maximum second order cyclostationary blind deconvolution(CYCBD)was proposed.According to the different resonance properties of bearing faults and gear faults,the multi-fault coupled signal was divided into high resonance components containing gear fault features and low resonance components mainly containing bearing fault features by RSSD.Then,the two components were treated by the CYCBD to eliminate the influence of the propagation path and noise interference,so as to enhance and extract weak fault features.In particular,to solve the problems of difficulty in parameter optimization,dependence on artificial experience,and poor adaptation in RSSD and CYCBD,an adaptive parameter optimization method based on the squirrel search algorithm(SSA)was proposed,and a composite index integrating kurtosis of envelope spectrum,root mean square of autocorrelation function maximum,and characteristic frequency ratio was designed as an optimization objective.Finally,envelope demodulation was performed on the deconvolved signal to extract the fault feature frequency and identify different fault sources.The effectiveness and feasibility of the proposed algorithm were verified by the multi-fault simulation signal and the measured signal of the planetary gearbox.Moreover,the proposed method was integrated into edge computing equipment to provide solutions for state detection and diagnosis,as well as remote operation and maintenance of rotating machinery such as planetary gearboxes.