基于自适应分段混合系统的轴承故障诊断

针对强噪声背景下轴承早期故障的诊断问题，提出一种基于自适应分段混合随机共振（adaptive piecewise hybrid stochastic resonance,APHSR）系统的检测方法。采用经验模态分解法（EMD）进行信号预处理，分别采用能量密度法和相关系数法去除高、低频噪声，自动筛选最优固有模态函数，经尺度变换后输入分段混合随机共振系统模型，提取故障信号。工程实验显示:经过APHSR系统，轴承故障特征频率的频谱幅值、频谱幅值与周围最大噪声之差和最大信噪比（SNR）均高于经验模态分解和经典随机共振方法，其中齿轮箱故障轴承信噪比分别提高了9.579 dB和7.473 dB，转子故障轴承信噪比分别提升了8.597 dB和5.695 dB，对凯斯西储大学故障轴承数据处理后的信噪比分别提升了3.369 dB和17.043 dB。数据表明APHSR方法具有高效性，提高了轴承故障信号诊断能力。 

Bearing fault diagnosis based on adaptive piecewise hybrid system

An view of the problem of early bearing fault diagnosis under high level of background noise，an adaptive hybrid piecewise stochastic resonance （APHSR） based on unsaturated stochastic resonance was proposed.According to this method，empirical mode decomposition （EMD） was used to preprocess the signal，while energy density method and correlation coefficient method were adopted to reduce high and low frequency noises respectively，obtaining the optimal intrinsic mode functions；and the model of the unsaturated stochastic resonance system was input after scale transformation，then the optimal parameters were obtained automatically and the fault signal was extracted.The engineering test data showed that the spectrum amplitude of bearing fault characteristic frequency，the difference between characteristic frequency amplitude and surrounding maximum noise，and the system output SNR （signal to noise ratio） were higher than those of EMD denoising and classical bistable stochastic resonance methods.Among them，the output SNR of gearbox fault bearing increased by 9.579 dB and 7.473 dB，respectively，the output SNR of rotor fault bearing increased by 8.597 dB and 5.695 dB，respectively，and the output SNR of outer ring bearing fault of Case Western Reserve University increased by 3.369 dB and 17.043 dB，respectively.The data showed that the APHSR method had high efficiency and strong applicability in bearing fault detection.