Due to the complexity and harsh operating environment of planetary gear transmission system, compound fault may co-exist in the planetary gearbox, which results in different type faults coupling together and the weak fault impulse features being completely submerged by strong ambient noise, thus making it a great challenge to extract fault-related information from planetary gearbox. To address these issues, a novel compound fault features extraction technique based on parallel dual-parameter optimized resonance-based sparse signal decomposition (RSSD) and improved multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) is proposed in this contribution. Firstly, according to the oscillation properties of different type faults, the parallel dual-parameter optimized RSSD constructs wavelet basis functions matching with fault features to adaptively decouple compound fault signal into high and low resonance components. Then, the improved MOMEDA is applied to deconvolute the resonance components to eliminate the interference of complex transmission path and strong ambient noise, thereby enhancing the weak periodic fault impulses. Finally, envelope demodulation processing for the enhanced signals is carried out to extract fault characteristic frequencies and identify different type faults. The effectiveness and feasibility of the proposed method are validated using both the numerical simulated signals and practical experimental dates from two different types of planetary gearbox compound fault. Moreover, comparisons with some existing methods illustrate the superiority of the proposed method to identify weak compound fault under strong ambient noise.

由于行星齿轮传动系统的复杂性和苛刻的运行环境，复合故障可能会在行星齿轮箱中共存，从而导致不同类型的故障耦合在一起，并且弱脉冲的冲动特征完全被强烈的环境噪声淹没，从而使其从行星齿轮箱中提取与故障相关的信息是一项巨大的挑战。为了解决这些问题，提出了一种基于并行双参数优化基于共振的稀疏信号分解（RSSD）和改进的多点最优最小熵反卷积调整（MOMEDA）的新型复合故障特征提取技术。首先，根据不同类型故障的振荡特性，并行的双参数优化RSSD构造与故障特征匹配的小波基函数，以将复合故障信号自适应地解耦为高谐振分量和低谐振分量。然后，将改进后的MOMEDA应用于对谐振分量进行去卷积，以消除复杂的传输路径和强烈的环境噪声的干扰，从而增强了较弱的周期性故障脉冲。最后，对增强信号进行包络解调处理，以提取故障特征频率并识别不同类型的故障。利用数值模拟信号和来自两种不同类型行星齿轮箱复合故障的实际实验数据，验证了该方法的有效性和可行性。此外，