小波分析和神经网络结合的变速箱状态识别

针对变速箱故障信号的非平稳和时变特性，提出了小波分析和神经网络结合的变速箱状态识别方法。为了验证该方法的有效性，试验模拟了某型车辆变速箱正常、7216轴承滚动体点蚀及3挡被动齿轮严重磨损3种状态，以箱体振动信号作为分析信号，首先对信号应用小波阈值法降噪减少干扰，接着将小波分解系数单子带重构得到不同频带的信号分量，提取各频带能量作为特征向量输人神经网络进行状态识别，结果表明该方法能有效识别变速箱的3种状态。

Condition Recognition of Gear-box with Combination of Wavelet Analysis and Neural Network

Aiming at the non-stationary and time-variation characteristic of the gear-box fault signal, a condition recognition method for gear-box combining wavelet analysis and neural network is proposed. To validate it, three conditions of gear-box are simulated by experiment, which involves normal condition, 7216 bearing pitting and the third speed driven gear severe wear. Taking the vibration of gear-box case as the analysis signal, de-nosing the signal via wavelet thresholding is firstly carried out to reduce the disturbance, and then wavelet coefficients branch is reconstructed from to gain the signal component of each frequency band. The signal energy of each frequency band is extracted as the feature vector and input to neural network for condition recognition, the result indicates that the method can recognize the 3 conditions of the gear-box effectively.