Research and Application of Online Fault Prediction Model Based on LS-SVM

Abstract

According to the characteristics of fault prediction algorithm, SVM (support vector machine) is not suitable for mass training and online training. Therefore, the SVM-based forecasting process is not a fast and online process. Utilizing the basic principles of the OS-LSM (online sparse least square support vector machine) algorithm, an online prediction model based on OS-LSSVM is established. Through monitoring of a UAV gyroscope voltage, online prediction and fault prediction on the trend of device status are exactly implemented. By proving, it is fast and effective to use the OS-LSSVM based model for online monitoring and fault prediction.