Study On Health Management For Intelligent Switching-mode Power Supply

With the development of space technology, spacecraft structure and application requirements have become more complex today. In order to improve the safety and reliability of the spacecraft, and make routine maintenance costs as low as possible, health management technology has been used in western countries. AHMS for main engines of Block II type space shuttle and ‘Prognostics and Health Management System’ for the F-35 Joint Strike Fighter are very representative. Overall, PHM has become a key technology to develop a new generation of weaponry and autonomous security.Switch-mode power supply is a common space part, the reliability and safety of power supply has a very important influence on the performance of the whole device. Traditional design of switch-mode power supply only has the voltage conversion function without the real-time monitoring and fault diagnosis, even the condition monitoring and protection. Therefore, the health management research for the Switch-mode power supply has important significance.The study is based on the switch-mode power supply device. The main content includes real-time monitoring, fault diagnosis, fault prediction, and timely warning function. The main works are shown as follows:1. ‘the selection of the fault feature’ : The output voltage of the switch-mode power supply is studied, when the parameters of components chang.Fault characteristics of the electronic system are established by using the method of least squares support vector machine(SVM) regression model,.2. ‘how to define the fault’ : The Mahalanobis distance of the output voltage is used to establish the control limit diagram and get the healthy threshold. The health of the system is evaluated by monitoring changes of power supply using Mahalanobis distance.3. ‘how to predict the fault’ : Historical data is used for fault prediction to estimate the status and trends, which can effectively prevent the occurrence of catastrophic failure and make loss to a minimum. The LS-VSM and BP neural network are used to forecast fault, and the results are analyzed.