多源融合技术在转子系统载荷类型识别中的应用.docVIP

多源融合技术在转子系统载荷类型识别中的应用 加力康1，王勤贤1，杨兆建2,王园媛3 (1太原理工大学电气与动力工程学院，2太原理工大学机械工程学院，山西 太原 030024；3国家电网长治供电公司，山西 长治 046011) 摘 要：针对单一类型响应信号在载荷识别中的局限性，提出振动信号和电机电流信号相结合的多源融合技术，进行载荷类型识别。对两种信号进行奇异值分解，将奇异值矩阵中的某些元素置零处理，实现信号降噪的同时，凸显频谱图中工频以外的频率分量。使用小波包分解得到信号的频率能量分布图，引入朴素贝叶斯作为分类器，根据能量分布图选定特征属性，通过概率学计算，进行载荷类型的判别。结果表明：在稳态、线性、正弦、冲击4种类型载荷下，两种信号经过奇异值分解和小波包分解处理后，基于朴素贝叶斯的多源融合技术能够很好地实现载荷类型的识别。 关键词：转子系统；多源融合技术；振动信号；电机电流信息；奇异值分解；能量分布；朴素贝叶斯；载荷类型识别 Application of Multi-source Fusion Technology in Load Types Recognition of Rotor System JIA Li-kang，WANG Qin-xian，YANG Zhao-jian， Wang Yuan-yuan (1College of Electric and Power Engineering, 2 College of Mechanical Engineering,Taiyuan University of Technology，Shanxi Taiyuan，030024, China, 3 State Grid Changzhi power supply company，Shanxi Changzhi,046011) Abstract: Aiming at the limitation of the single type of response signal in the load identification, a multi-source fusion technique based on the combination of vibration signal and motor current signal was proposed. The singular value decomposition of the two kinds of signals were carried out and several elements of the singular value matrix were set to zero to achieve signal noise reduction and highlight the frequency components of the spectrum map except power frequency. The features of the frequency energy distribution decomposed by wavelet packet were selected and Naive Bayesian was used as a classifier to identify the load types of rotor system. The results show that under the four types of sinusoidal,steady,linear and impulse loads, after the signals of two types were processed by singular value decomposition and wavelet packet decomposition?,multi-source fusion technology based on Naive Bayes can identify four types of load . Keywords: Rotor System; Multi-source Fusion Technology; Vibration Signal; Motor Current Signal; Singular Value Decomposition; Energy Distribution; Naive Bayes; Load Type Recognition 1 引言 转子系统受到不同载荷作用时，由于传动作用，载荷特性会反映到响应信号中。现有的载荷识别多为大小的识别和位置的识别，文献[1-3]分别使用频域法、时域法、人工智能法进行了载荷大小的识