论文部分内容阅读
盲系统辨识是仅由输出数据来获得系统特性函数的一种信号处理方法。系统特性只与自身的结构相关,一种工况就对应着一种特定的系统特性。将系统结构及工况两者结合分析,可有效应用于齿轮箱的故障诊断。首先,利用独立分量分析对获得的的信号进行预处理,提取出包含故障频率的信号作为系统模型的响应信号。其次,高阶累积量具有消除和衰减高斯噪声的特性,使用高阶累积量构建时间序列模型。最终,依据模型的系数计算得到的ARMA双谱定性分析,用量子自组织特征映射网络给出定量的判据。实验结果表明,此方法对齿轮箱故障的存在和故障类型的诊断,可以提供一些有价值的结论。
Blind system identification is a signal processing method that obtains the system characteristic function only from the output data. System features are only related to their own structure, a situation corresponds to a specific system characteristics. The system structure and working conditions combined analysis of both can be effectively applied to the gear box fault diagnosis. First, the obtained signal is preprocessed using independent component analysis, and the signal containing the fault frequency is extracted as the response signal of the system model. Second, high-order cumulants have the property of eliminating and attenuating Gaussian noise. The high-order cumulants are used to construct time series models. Finally, based on the qualitative analysis of ARMA bispectrum calculated by the coefficient of the model, the quantitative criterion is given by the network of quantum self-organizing feature. The experimental results show that this method can provide some valuable conclusions for the existence of gearbox fault and the diagnosis of fault type.