【作者】 安国庆；

【导师】 刘教民；

【作者基本信息】 河北工业大学 ， 电气工程， 2013， 博士

【摘要】 异步电动机故障早期，故障特征非常微弱。且出于调速和节能要求，许多场合下的异步电动机由变频器驱动，变频器中剧增的谐波成分，加上外围机械设备所产生的噪声，使异步电动机早期故障特征的检测更为困难。以异步电动机转子断条、定子匝间短路以及轴承故障为对象，研究了强噪声背景下早期故障特征的提取与检测技术。针对异步电动机转子早期故障时，定子电流中新增的故障特征分量被基波信息湮没而难以识别的问题，提出了改进型相关算法来检测转子断条故障。根据定子电压与电流基波频率相等的特点，构造出与定子电压同频的参考信号，利用改进型相关算法提取定子电流基波信号的幅值与相位，将基波分量准确滤除来突出转子断条故障信息。由变频电源驱动的异步电动机转子故障时，在变频器输入端的电流频谱中会出现以载波频率为中心，按某特定频率间隔分布其左右的故障边频。指出该故障特征容易被附近的频率分量湮没，且当低频或轻载运行时，故障特征的提取变得更加困难。提出根据变频器输入、输出以及载波频率构造参考信号，利用相关性消去法滤除湮没故障特征的频率分量的方法，使在变频器噪声背景下的转子故障特征在频谱图上凸显。异步电动机发生定子匝间短路早期故障后，尤其在变频器供电下，电流信号中存在大量谐波及噪声严重影响了定子故障诊断结果。针对该问题利用三相改进型相关算法准确获取故障电机三相定子电流基频信息，提出通过逆序同步速坐标变换将基频正序分量转换成二倍频交流量，负序分量转换成直流量。采用均值法提取直流分量，导出了负序分量在逆序同步速旋转坐标系下合成矢量幅值，并考虑电机先天不平衡因素定义了灵敏度因子以表征匝间短路故障程度。异步电动机轴承早期故障特征容易被机器运转时的背景噪声湮没而难以识别。提出一种提取轴承故障早期微弱信号的峭度滤波器。该方法对振动信号进行短时傅立叶变换以得到信号的峭度，并自动构建一个受控于故障信噪比的滤波器滤除信号中无关的噪声成分，对滤波后的信号进行包络分析，可凸显故障信息。以32位的S3C2440A内核微处理器作为电动机参数的采集核心，以Labview软件为数据分析及故障诊断的平台，构建出一套异步电动机早期故障检测系统。最后验证了该系统在定子匝间短路、转子断条和轴承故障诊断的可行性。更多还原

【Abstract】 The incipient fault characteristic of the induction motor is usually very weak. Forspeed regulation and energy saving requirements, the induction motor is driven byfrequency converter. The soaring harmonic component in converter and the noiseproduced by peripheral equipment make it more difficult for the incipient fault detection.Aiming at the broken rotor bar, inter-turn short circuit, and bearing fault of the inductionmotor, this dissertation focuses on incipient fault characteristic extraction and detection inheavy noise.When induction motor has the incipient broken bar fault, the new added faultcharacteristic component is difficult to be discriminated in the spectrum analysis because ofbeing blanketed by fundamental component. The improved correlation algorithm isproposed to solve this problem. According to the truth that the voltage and the fundamentalcomponent of current have the same frequency, reference signals whose frequency areequal to the one of stator voltage are constructed. By the improved correlation algorithm,the amplitude and phase of the fundamental component in stator current can be obtained, sothe rotor broken bar fault signal will be emerged after the fundamental component isremoved accurately. When broken rotor bar fault occurs in induction motor with variablefrequency power, the fault characteristic centred on the carrier frequency is distributed inthe input-side current spectrum of inverter with specific frequency interval. But it is easy tobe covered with neighbouring components, and the condition gets worse as the inverterruns in low frequency or under light load. The reference signals are constructed accordingto input, output and carrier frequency of inverter. With improved correlation algorithm, thecomponents confusing the diagnosis result is removed，and the fault characteristic canemerge from the spectrum.Since there are a lot of harmonics and noise in the stator current signal, especiallywhen induction motor is fed with variable converter, the inter-turn short circuit faultdiagnosis should be greatly influenced. With the three-phase improved correlationalgorithm, the fundamental component in stator current can be got accurately. This paperpresents the detection method based on the anti-synchronous-speed coordinatetransformation. Via this transformation, the positive sequence component is turned into the second harmonic, and the negative one is transformed into DC component which can beextracted with the mean algorithm, then the amplitude of synthetic vector of the negativesequence component in new coordinate can be derived. A sensitivity factor is defined toevaluate the severity extension of the inter-turn-short-circuit faults, which takes intoaccount the manufactured asymmetry of the induction motor.The incipient bearing fault characteristic is so weak that it is very difficult to bediscriminated because of being obscured by noises of the peripheral equipment. A filterused to extract weak signal in heavy noise is presented to extract weak fault signal based onspectral kurtosis. Via short-time Fourier transform to the vibration signals, the estimation ofthe spectral kurtosis is given. According to the spectral kurtosis, a filter controlled bysignal-to-noise ratio is constructed automatically to remove the noises and harmoniccomponents. Then the incipient bearing fault characteristic information could be veryobvious by means of the envelope analysis.Taking32-bit S3C2440A microprocessor as a core of motor parameters collection，viathe Labview software platform for data analysis and fault diagnosis，an incipient faultdiagnosis system for the induction motor is constructed. Experiment results indicate that thesystem is feasible in the broken rotor bar, inter-turn short circuit, and bearing faultdiagnosis.更多还原