【摘要】 针对非平稳工况下轴承振动信号多源耦合干扰难题，提出一种基于阶频-稀疏特征映射理论框架的协同诊断方法。通过阶频谱相关(Order-Frequency Spectral Correlation, OFSC)技术构建振动信号的角度-频域双域表征模型，解析故障冲击的循环平稳调制特性，并生成时频耦合矩阵揭示其内在规律。在此基础上，设计稀疏非负矩阵分解(Sparse Non-negative Matrix Factorization,SNMF)模型，在频域基向量与阶次特征向量的双域空间中实现噪声及多源干扰成分的协同解耦，强化故障冲击的稀疏表达。进一步结合基向量优选与子空间重构策略，建立故障敏感频带的自适应提取机制，融合增强包络阶次谱(Enhanced Envelope Order Spectrum,EEOS)算法，可显著提升低信噪比环境中微弱故障特征的可辨识性。仿真与实验分析结果表明，该方法能够有效分离复杂干扰背景下的故障特征，由此验证了其在时变工况中的鲁棒性与工程适用性。更多还原

【Abstract】 To address the challenge of multi-source coupled interference in bearing vibration signals under nonstationary operating conditions, a collaborative fault diagnosis method based on order-frequency sparse feature mapping theoretical framework was proposes. By using Order-Frequency Spectral Correlation(OFSC) technique, an angularfrequency dual-domain representation model was constructed to characterize the vibration signals, analyze the cyclostationary modulation characteristics of fault-induced impulses and generate time-frequency coupling matrices to reveal their intrinsic law. On this basis, a Sparse Non-negative Matrix Factorization(SNMF) model was designed to realize collaborative decoupling of noise and multi-source interference components in the dual-domain space of spectral frequency base-vector and cyclic order feature vector, thereby enhancing the sparse representation of fault-related impacts.Furthermore, an adaptive extraction mechanism for fault-sensitive frequency bands was established by the combination of optimal base-vector selection with subspace reconstruction strategies. Integrating the adaptive extraction mechanism with the Enhanced Envelope Order Spectrum(EEOS) algorithm, the detectability of weak fault features in low signal-to-noise ratio environments could be significantly improved. Simulation and experimental analyses demonstrate that the proposed method can effectively extract fault characteristics under complex interference backgrounds, which validates its robustness and practical applicability in time-varying operating conditions.更多还原