基于支持向量数据描述与极端梯度提升的油动机健康诊断系统设计OA

In thermal power plants,the unit operation safety during rapid load variations under deep peak-shaving condition is of great significance.The hydraulic servo motor is a key actuator in the unit,so an overall framework of health diagnosis system for it is designed and developed.A health monitoring and early warning model based on Support Vector Data Description(SVDD)and eXtreme Gradient Boosting(XGBoost)is constructed.Considering the characteristics of these two models,an algorithm for early fault warning for hydraulic servo motor is designed:in the hypersphere created by SVDD,when the distance between data location and the hypersphere center deviates from baseline value,a fault classification diagnosis conducted by XGBoost would be triggered,thereby,real-time health assessment and fault diagnosis of hydraulic servo motor during operation can be realized.Fault injection experiments are conducted to simulate seven typical faults under different valve opening degrees,such as throttle orifice blockage of the solenoid valve and internal leakage of servo motor.Normal and faulty sample data are collected,and the accuracy of this system is verified by experimental results.Adopting this system,scaled(0-100)health quantitative assessment of hydraulic servo motor can be realized in opening range of 0％～100％ of valve,and fault recognition accuracy for low-health conditions(scale＜60)can reach 93％.The research results can provide reference for fault diagnosis of hydraulic servo motor.