Study On Mechanical Fault Diagnosis Method Of High Voltage Circuit Breaker Operation Mechanism Based On Signal Processing

Most of the high voltage circuit breakers work outdoors and suffer from the external environment all year around,so they are very prone to failure.According to statistics,most of the faults are mechanical faults of their operating mechanism.When different faults occur,the time-frequency characteristics of the vibration signal and coil current signal of the circuit breaker during opening and closing will show different characteristics.So analysis the characteristics of vibration signal and coil current signal can classifying the type of circuit breaker operating mechanism,which is very important to improve the efficiency of circuit breaker maintenance.In this context,firstly,collecting vibration signal and coil current signal,then processing and extracting feature characteristics and combining with the artificial intelligence classification algorithm,studied the method of the mechanical fault diagnosis of circuit breaker operating mechanism.Firstly,in order to obtain the vibration signal and coil current signal,this paper designs a data collecting scheme.For the vibration signal,acceleration sensor is selected to collect vibration signal,which is installed on the pedestal of breaker operating mechanism,and used waveform monitor to collect vibration signal.For the coil current signal,selected hall current sensor to collect coil current signal.In addition,this paper simulated three common mechanical fault states of circuit breaker operating mechanism: pedestal screw loosed,iron core jam and operation mechanism jam,and repeated tripping and closing operations to collecting and preserving vibration signal and coil current signal of each state,One part is used as training data,the other part is used as testing data,as the data support of this paper.Vibration signal is non-linear and non-stationary signal.In order to analyze it in full frequency domain,using wavelet packet to decompose vibration signal,then get the envelope of the decomposed signals in each frequency band.Then the signals of each frequency band were divided into equal time periods in the time domain and acquisition the energy of each time period,so as to obtain the time-frequency energy characteristic spectrum of the vibration signal.Analysis the difference of time-frequency energy spectrum of each state.Establish the fault diagnosis model by using Deep Auto-Encoder Networks and verifying this model.In case study,analyzed the influence on the fault diagnosis accuracy of different decomposition layers of wavelet packet and the number of time segments of each frequency,and choosing the optimal layer of wavelet packet decomposition and the number of equal time division.In addition,analyzed the influence of the Deep Auto-Encoder Networks on the fault diagnosis accuracy in different input nodes number,the number of network layer and different node modes.Finally,verified the diagnosis model and the result shows that the model could achieve a good fault diagnosis accuracy.Trip and close coil current are the important monitoring data to reflecting the working state of high voltage circuit breaker operating mechanism,it's feature extraction can completely in the time domain.Firstly,description the working character of the iron core and the typical trip and close coil current,selecting the time and the current feature point,and denoising to close coil current signal collecting in experiment,extracting the seven important characteristics of close coil current and establish them with a feature vector,and compared and analyzed the variation of the close coil current waveform and feature vectors between three fault states and the normal state;Secondly,established the mechanical fault diagnosis model of breaker operating mechanism with KNN algorithm,and contrast fault diagnosis accuracy according to different k values.Finally,improved the method of extracting characteristic of the close coil current by adding three characteristic points,and combined them to establish an improved feature vector,then verified the diagnosis accuracy of new method by KNN classification algorithm.The result showed that the improved feature vector can significantly improve the accuracy of diagnosis.This paper studies the characteristics of the vibration signal and the coil current signal in the mechanical fault of the breaker operating mechanism,and combines different diagnostic methods to diagnose the mechanical fault of the breaker operating mechanism.The research and method shows that when the circuit breaker operating mechanism occurs different mechanical fault,the time-frequency characteristics of vibration signals and coil current also will be different,the method could achieve a good diagnosis result of the fault states of circuit breaker operating mechanism.The proposal research can provide a theoretical basis for power grid operation and the decision of maintenance.