基于电流信号的转子系统故障诊断与采煤机截割工况识别

发布时间：2018-01-12 13:40

本文关键词：基于电流信号的转子系统故障诊断与采煤机截割工况识别　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着中国制造2025和工业4.0的提出,机械设备作为生产制造企业的核心装备,发挥着举足轻重的作用。为保证设备安全、可靠、高效地运行,避免恶性事故的发生和经济的损失,开展机械设备的故障诊断和运行状态监测,具有非常重要的意义。近年来,电机电流特征分析法作为一种新兴检测技术逐渐受到广大学者青睐,通过监测电机电流信号进行机械故障诊断和状态识别已经成为一个研究热点,本文在此基础上对电流信号的特征提取方法、转子系统故障诊断和采煤机截割工况识别方法进行了探索研究,主要工作内容如下:1、从理论角度分析了负载扭矩变化对电机电流信号的影响,负载扭矩波动体现在电机电流信号频谱上会产生频率调制现象,即电流基频e0f两侧出现ieff?0的频率分量。通过在Matlab/Simulink中建立电机模型,仿真验证了理论的正确性。2、针对电机电流信号特征提取困难,特征频率易被工频湮没的问题,将总体平均经验模态分解(EEMD)引入电流信号的处理中,利用改进小波阈值去噪、EEMD及互相关分析相结合方法对电流信号进行处理,通过在转子试验台上施加正弦扭矩激励来模拟扭矩变化,采集电机电流信号进行处理。试验结果表明,利用互相关分析筛选IMF分量的方法,能够快速有效地进行IMF分量的选取并抑制50Hz工频及其谐波的干扰,提取扭矩波动的频率,从而证明了该方法在实际应用中的可行性。3、针对转子系统的不平衡、不对中故障,利用EEMD-PCA的方法提取电机电流信号的幅值域和时频域特征参数,在转子系统故障模拟试验台上采集电机电流信号,利用BP神经网络和支持向量机对故障进行识别。试验结果表明利用EEMD-PCA进行特征提取能够有效提高识别效果,且EEMD-PCA-SVM识别准确率达到了93.4%,高于EEMD-PCA-BP的80.0%。4、针对采煤机截割过程中的煤岩工况识别问题,利用小波包能量法对电机电流信号进行特征提取,得到特征向量,再从特征向量和支持向量机参数两个方面对识别算法进行优化,试验结果表明优化后的PSO-SVM算法对不同滚筒转速、不同截割高度下的煤岩截割工况识别率均达到了90%以上,效果比较理想。
[Abstract]:With the development of manufacture in China 2025 and 4.0, mechanical equipment, as the core equipment of manufacturing enterprises, plays an important role in order to ensure the safety, reliability and efficient operation of the equipment. In recent years, it is very important to avoid the occurrence of malignant accidents and economic losses, and to carry out fault diagnosis and operation state monitoring of machinery and equipment. As a new detection technology, the motor current characteristic analysis method has gradually been favored by the majority of scholars. Mechanical fault diagnosis and state identification by monitoring the motor current signal has become a research hotspot. In this paper, the current signal feature extraction method, rotor system fault diagnosis and shearer cutting condition identification methods are explored and studied. The main work is as follows: 1. The influence of load torque variation on motor current signal is analyzed theoretically. Load torque fluctuation is reflected in the frequency modulation phenomenon in the frequency spectrum of motor current signal, that is, ieffs appear on both sides of current base frequency e0f. By establishing the motor model in Matlab/Simulink, the correctness of the theory is verified by simulation. It is difficult to extract the characteristics of the motor current signal. The characteristic frequency is easy to be annihilated by power frequency. The total average empirical mode decomposition (EEMD) is introduced into the current signal processing, and the improved wavelet threshold is used to de-noise. The current signal is processed by EEMD and cross-correlation analysis. The torque change is simulated by applying sinusoidal torque excitation on the rotor test-bed, and the motor current signal is collected for processing. The test results show that. By using the method of cross-correlation analysis to select IMF components, the selection of IMF components can be carried out quickly and effectively, and the interference of 50Hz power frequency and its harmonics can be suppressed, and the frequency of torque fluctuation can be extracted. It is proved that the method is feasible in practical application. The method is aimed at the unbalance of rotor system and misalignment fault. The amplitude range and time-frequency characteristic parameters of motor current signal are extracted by EEMD-PCA method, and the motor current signal is collected on the rotor system fault simulation test platform. BP neural network and support vector machine are used to identify the fault. The experimental results show that the feature extraction using EEMD-PCA can effectively improve the recognition effect. The accuracy of EEMD-PCA-SVM recognition is 93.4, which is higher than that of EEMD-PCA-BP (80.0.4). The wavelet packet energy method is used to extract the feature of the motor current signal and the eigenvector is obtained. Then the recognition algorithm is optimized from two aspects: the eigenvector and the support vector machine parameters. The experimental results show that the optimized PSO-SVM algorithm can recognize the cutting conditions of coal and rock at different drum speed and cutting height, and the recognition rate of coal and rock cutting conditions is more than 90%, and the effect is satisfactory.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD421.6

【参考文献】