计算机科学
控制理论(社会学)
峰度
断层(地质)
解调
瞬态(计算机编程)
平滑的
干扰(通信)
特征(语言学)
算法
方位(导航)
特征选择
包络线(雷达)
时频分析
基带
解析信号
谐波
谐波
电子工程
非线性系统
模式识别(心理学)
故障检测与隔离
小波
频带
作者
Zujie Yang,Hong-Na Dong
标识
DOI:10.1088/2631-8695/ae85c9
摘要
Abstract Spectral kurtosis has been widely used in bearing fault diagnosis because of its efficient resonance-band selection capability for impulsive feature extraction. However, its kurtosis-based criterion is highly sensitive to outliers and can be dominated by high-energy random transient impulses under complex operating conditions involving multiple impulsive sources and strong non-stationary disturbances. As a result, the selected frequency band may deviate from the actual fault-related resonance region, leading to degraded diagnostic performance. Although several improved methods have been developed using nonlinear mappings or statistical smoothing strategies, they often weaken fault-induced transient structures and compromise feature interpretability. To overcome these limitations, this study proposes an adaptive demodulation method based on the dual-tree complex wavelet packet transform (DTCWPT) and a KPS (Kurtosis–Periodicity–Stability) joint criterion. First, DTCWPT is employed to provide a redundant and approximately shift-invariant time–frequency representation, enabling more reliable characterization of impulsive components in the presence of random disturbances and resonance-frequency variations. Second, the proposed KPS criterion jointly evaluates impulsiveness, periodicity, and stability by combining kurtosis, envelope autocorrelation, and statistical consistency measures. This design enhances fault-related periodic impacts while suppressing random transient disturbances. Based on the KPS criterion, adaptive multi-scale subband evaluation and selection are performed to avoid resonance-band decisions being dominated by non-fault high-energy impulses. Simulation and experimental results demonstrate that the proposed method can reliably extract bearing fault characteristic frequencies and their harmonics under low signal-to-noise-ratio conditions and severe transient interference. The results confirm its effectiveness, robustness, and practical applicability for bearing fault diagnosis.
科研通智能强力驱动
Strongly Powered by AbleSci AI