一种基于功率谱重心的自适应特征信息提取方法

Self-Adaptive Feature Extraction Method Based on Power Spectral Centroid

提出一种基于功率谱重心的自适应特征信息提取方法。该方法通过对不同状态转换过程中的灵敏度分布进行分析获得特征功率谱,然后计算特征功率谱的重心频率以及自适应加权系数,对特征功率谱进行自适应加权计算获得特征信息。文中给出了将自适应特征信息提取方法应用于磨煤机噪声信号特征提取的实例,对比分析了采用不同特征信息提取方法的效果。仿真结果表明,所提出的自适应特征信息提取方法能够准确剔除无效信息,自主适应运行工况变化,具有较好的灵敏度和线性度,为提高软测量模型的预测精度和泛化性能提供了保证。

A self-adaptive feature extraction method based on power spectral centroid is proposed. The spectral frequency centroid and the self-adaptive weights are calculated based on the characteristic spectra by analyzing the sensitivity distributions of working condition transitions. And then the features of the noise can be extracted. Furthermore, the proposed method is compared with two general feature extraction methods in the application of the ball mill noise feature extraction. Simulation results verify that the method can remove independent information and is adaptive for the working condition transitions. In addition, the method has better sensitivity and linearity, thus improving the estimated precision and generalization ability of the soft sensing model.