论文部分内容阅读
针对传统方法识别高铁工况存在特征提取不完备和识别性能不精确的问题,提出一种多视图分类集成的高铁工况识别方法(MVCE)。该方法结合多视图特征提取和分类集成技术,从信号本身特性、频域和时频域三个角度提取小波能量、频谱系数、聚合经验模态分解模糊熵,并使用Fisher比率对其频域特征进行特征选择,从而构建高铁振动信号三个视图的特征。使用最小二乘支持向量机和K最近邻分类器分别对每个视图的特征进行初步识别。最后采用分类熵投票策略对多个分类器输出结果进行集成。试验结果表明:该方法对仿真数据和实验室数据的平均识别率分别达到89.18%和90.87%。同时对比结果说明了该方法提取特征的完备性和具有多样性集成模型的有效性。
Aiming at the problem of incomplete extraction of feature and inaccurate recognition performance of traditional methods for identifying high-speed rail, a multi-view classification integrated high-speed rail condition recognition (MVCE) method is proposed. Combining with multi-view feature extraction and classification integration technology, this method extracts wavelet energy, spectral coefficients and aggregated empirical mode decomposition entropy from three aspects of signal characteristics, frequency domain and time-frequency domain, and uses Fisher’s ratio for its frequency domain features Feature selection, to build high-speed railway vibration signal three views of the characteristics. Least Squares Support Vector Machines and K nearest neighbor classifiers are used to preliminarily identify the features of each view. Finally, the classification entropy voting strategy is used to integrate the output of multiple classifiers. The experimental results show that the average recognition rate of the proposed method for simulated data and laboratory data is 89.18% and 90.87% respectively. At the same time, the comparison results show the completeness of the extracted features and the effectiveness of the integrated model with diversity.