Stator Early Fault Diagnosis And Identification Research On Induction Motor

Induction motor has a simple and fixed structure,with reliable operation and easily maintenance,which price is pretty low,also has other characteristics,widely used in power plants,steel mills,shipbuilding industry and national defense.However,due to the complex changes in the assembly,operating characteristics,structure,working methods and loads in motors,the failure is inevitable.In order to ensure safe and efficient operation of manufacturing,domestic and foreign scholars did a lot of research in induction motor fault diagnosis.Through condition monitoring and fault diagnosis of the motor,we can achieve early detection of faults,while avoid production lost caused by motor failure.Affected by aging,wear and tear,overheating,vibration and other factors,the stator winding in-turn short circuit is one of the common early motor fault,and may lead to the coil short circuit or phase-to-phase short circuit,that is to say,motor's serious fault.The early signal of the fault is relatively weak,vulnerable to external factors,which may affecting the accuracy of fault diagnosis.Under complex conditions,such as load fluctuations and supply voltage unbalance,the steady-state analysis of the motor stator current signal is difficult to accomplish reliable fault diagnosis.On the other hand,it is needed to explore an effective method to distinguish the stator early failure,high impedance faults and supply voltage imbalance,to determine the cause of different faults with the same fault symptoms,in order to take reasonable measures to avoid more serious failure.In this context,this paper studies the multi-type fault diagnosis method based on analytical mathematical models and signal processing&analysis,established a theoretical framework of induction motor stator early fault diagnosis and identification,systematically and comprehensively.Thus improve and supplement the existing stator early fault diagnosis technology.The main completed work as follows:Firstly,the theoretical mechanism analysis of induction motor stator early fault explaining the principle of the fault features essence,establishing the transient and steady-state fault model,laid the theoretical foundation for stator early fault diagnosis.Secondly,a new method for stator early fault diagnosis is proposed based on the analytical mathematical model,stator early fault model has been improved in this paper,with a new fault diagnostic indicators.Using the combining fault diagnosis method of self-adaptive state observer and parameter identification,corresponding to the real-time running status of motor,the state observer was built.Input the voltage measurements,we can get the error estimate value from the comparison of the state observer's stator current and measured values.Analyse the steady-state model of stator,we get the positive and negative sequence component of stator current error estimates,obtained negative sequence component error estimates through reverse coordinate transformation,excluded the impact of uncertain interference factors on fault diagnosis.The negative sequence component,as a reliable indicator of fault diagnosis,using stator current error estimates,is proportional to the number of turns and the fault current,has no direct correlation with parameter error,can run the stator early fault diagnosis during load changes and different speeds.Thirdly,a simple and effective method to stator early fault diagnosis is the detection of negative sequence current,while its drawback is the negative sequence current are from multiple information sources,obtained by interaction of the motor stator in-turn short circuit fault,supply voltage unbalance,and the.motor naturally nonlinear.In order to achieve the separation of multiple information sources,the stator early fault diagnosis method was proposed based on sequence component extraction.This method,obtained positive and negative sequence components from two-dimensional dq system in the time domain at first,then determining the motor negative sequence current characteristic parameters by recursive least squares method,eliminating the interference of negative sequence current component in normal state,made up the accurately conclusion of whether the motor early stators fault exist or not,and obtained the accurately number of shorted turns in stator winding,according to short-circuit fault negative sequence current values.This method greatly improves the accuracy of the non-invasive fault diagnosis in stator early fault.Then,there is something in common between the stator in-turn short circuit fault and high impedance faults,while still have their own characteristics.In order to avoid misclassification of the two faults types,experienced deep theoretical analysis,the use of stator current negative sequence component Isn zero-sequence component Usz was proposed as the identification indicators to the two types of fault.Amplitude and phase angle of the 3-phase current and positive sequence current were used to support the classification algorithm.Under different fault and load conditions,this paper achieved the simple and effective recognition between the stator in-turn fault and high resistance,accomplished fault diagnosis,fault location and fault severity estimates.Depending on type and severity of the fault,we determine the repairs needed accordingly,make the running and maintenance of motor more flexible and more efficient.Thus,the content of the induction motor stator early fault diagnosis technology has been extended.Finally,the integration of the signal processing and support vector machine was raised,as the multiple intelligences fault diagnosis approach.Facing changing load and run state of the motor,the stator voltage imbalance and inter-turn short circuit fault is easily confused,through the establishment of state identification/classification model,achieved motor stator early fault diagnosis.In-depth study of the support vector machines theory and it's problems in practical application was made,that is to say,the improved genetic algorithm was integrated into support vector machine(SVM),as the improved form of support vector machine,and combined with the dual tree complex wavelet transform,we found multiple type intelligences fault diagnosis method,and thus improved the existing stator early fault diagnosis technology.