基于EEMD能量矩和改进量子粒子群神经网络的滚动轴承故障诊断

针对滚动轴承故障振动信号的非线性和周期性冲击特征, 提出一种基于集合经验模态分解(Ensemble Empirical Mode Decomposition, EEMD)能量矩和改进量子粒子群神经网络的特征提取方法. 基于小波去噪对滚动轴承原始信号进行预处理, 对重构的故障信号进行EEMD并得到多个本征模态函数分量. 利用能量矩方法计算出所需分量的能量矩并归一化, 将归一化后的能量特征参数作为量子粒子群BP神经网络的参数输入, 根据加速度传感器信号实现滚动轴承故障诊断. 分别在不同转速(载荷)下采集驱动端轴承的振动信号, 获取200个训练样本和80个测试样本, 并建立故障诊断模型. 基于文中模型对实际滚动轴承数据进行诊断, 实验结果表明, 不同转速(载荷)下测试的80个样本故障诊断准确率达到100%.

Rolling bearing fault diagnosis based on EEMD energy moment and improved quantum particle swarm optimization neural network

In light of the nonlinear and periodic shock characteristics of rolling bearing fault vibration signals, a feature extraction method based on ensemble empirical mode decomposition (EEMD) energy moment and improved quantum particle swarm neural network is proposed. Based on wavelet denoising, the original signal of the rolling bearing is preprocessed, and the reconstructed fault signal is decomposed by EEMD to obtain multiple eigenmode function components. The energy moment of the required component is calculated by the energy moment method and is normalized. The normalized energy characteristic parameter is used as the parameter input of the quantum particle group BP neural network, and the rolling bearing fault diagnosis is accomplished with the acceleration sensor signal. The vibration signals of the drive end bearings are collected at different speeds (loads), 200 training samples and 80 test samples are acquired, and a fault diagnosis model is established. With this model, the actual rolling bearing data is diagnosed for case study. The experimental results show that the accuracy of 80 samples tested under different speeds (loads) may reach 100%.