基于LS-SVM软传感器的容错控制方法及应用

Fault-tolerant control method based on LS-SVM soft sensor and its application

针对一类非线性系统中传感器卡死、恒增益和恒偏差失效等易发故障,考虑控制系统多存在耦合、非线性、时变、滞后等难以建立精确的解析模型,基于数据驱动技术提出一种软传感器容错控制方法.基于LS-SVM构建了系统软传感器,并利用软传感器的预测输出与实际传感器输出之差获取残差信号;采用SPRT算法进行故障检测,当传感器发生故障时,用LS-SVM软传感器预测输出代替物理传感器的实际输出,从而以软闭环方式实现对传感器故障的容错控制.将所提出的方法应用于一阶水箱液位控制系统,实验结果表明,基于LS-SVM软传感器与SPRT的结合能够可靠及时检测非线性系统中各类传感器故障,而借助于软闭环切换还可对传感器故障实现安全容错.

Aimed at the common faults in a class of nonlinear system such as sensor sticking,constant gain,and constant deviation and taking account of the difficulty of establishing aprecise analysis model due to the coupling,nonlinearity,time-varying,retardation commonly available in control system,a soft sensor fault-tolerant control method is proposed based on data-driving technique.The soft sensor of the system is established at first based on LS-SVM and the signals of residual errors are obtained as the difference between the predicted outputs of the soft sensor and actual outputs of the physical sensor.And then the SPRT algorithm was used to carry out fault detection.Then the predicted output detected by the LS-SVM soft sensor is used to replace the actual output of the physical sensor when the fault is took place with the sensor,so that the fault-tolerant control of the sensor would be realized with the soft closed loop manner.Finally,the proposed method is applied to the first-order water tank level control system.The experimental result shows that various types of sensor faults in nonlinear systems may be detected reliably and in time by means of combined LS-SVM soft sensor with SPRT,and the safe fault-tolerant control of sensor may be achieved by soft closed loop switching.