Early Fault Detection And Performance Degradation Prediction Of A High-Voltage Power Supply

High-voltage power supply plays a role of power supply in the radar system.The performance of high-voltage power supply directly determine the operation quality of radar system,which is the basis for ensuring the normal and safe operation of all electronic equipment.Therefore,the reliability of high-voltage power supply is particularly important.Fault detection and prediction technology of high-voltage power supply system is the key technology for the maintenance and support of the equipment.Early fault means that the current performance of the circuit deviates from the allowable degradation range,it is the early stage of soft fault.Detecting the early faults of the system can discover the faults in time at the early stage,and take maintenance measures in time to avoid the further deterioration of the faults.Performance degradation prediction is to extrapolate the trend of the performance degradation function of the circuit according to the current relevant operating state information,through reasoning to find out whether the tested object needs maintenance.It has practical value to reduce the maintenance cost of equipment and avoid disaster.In order to solve the above problems,this thesis takes a high-voltage power supply as the research object,and deeply studies its early fault detection and performance degradation prediction methods.The research contents are divided into the following aspects.(1)The function of each module of the high-voltage power supply and the possible degradation sources are analyzed.Combined with the results of sensitivity analysis,the key components causing the degradation of the circuit performance are determined as sources of degradation,the degradation is analyzed and the model is selected.The degradation data of the circuit is obtained through simulation experiments.(2)The degradation characteristics of the high-voltage power supply are selected.For the control part of the circuit,the output parameters are directly selected as the degradation characteristics because of the obvious degradation in the time domain.For the transformer part of the circuit,the methods of wavelet decomposition,wavelet packet decomposition,principal component analysis,information entropy and deep belief network are studied respectively.The effects of each method are compared and the parameters that can obviously reflect the degradation of the output signal are selected as the degradation characteristics.(3)The early fault detection method of high-voltage power supply is studied.The normal envelope model based on Gaussian process regression is established,Then the model is utilized to detect the circuit and the detection rate is calculated.The influence of data error on the detection result is analyzed also.Finally,hardware experiments are carried out to verify the effectiveness of the method.(4)The performance degradation prediction method of high-voltage power supply is studied.The prediction model based on particle filter is established and the data is predicted.Then the prediction effect is evaluated,and the influence of noise on the prediction result is analyzed.Next,compared with the prediction model established by directly using the output features or another method,the simulation results show that the performance degradation prediction method proposed in this paper has a good effect.Finally,hardware experiments are carried out to verify the effectiveness of the method.