在汽车线控转向优化控制的研究中,汽车传感器容错技术模型精确度低和易受扰动影响等问题、造成汽车的安全性能受到影响.针对传统解析关系模型精度低,采用了邻域粗糙集模型对传感器信息进行预处理,用以精确找出与容错对象存在解析关系的相关联传感器信息;为了消除观测器的扰动影响,利用了神经网络组建容错对象的冗余信息,将关联传感器信号作为径向基神经网络的输入,容错对象的信号用作进行监督训练.利用神经网络的估计输出和容错对象的输出差值,即残差是否超出门限来实现故障判别,在残差超过门限后进行输出控制,屏蔽故障传感器输出,可用神经网络的估计输出来完成信号补偿.通过仿真表明,改进设计具有较好的抗噪性和逼近能力,能很好的完成故障检测和信号补偿,达到容错控制的目的.%In the view of the disadvantage of the traditional fault-tolerant technique, this paper presented an intel-ligent fault-tolerance design for sensors of steer-by-wire aulomohile to replace the inaccurate model and observer which is vulnerable to disturbance affects. A pretreatment method based on rough set was utilized to find sensors which are related to the object sensor. The sensors were taken as the inputs of a radial basis function neural network, and the information of the object sensor as the standard output of the network when it worked well. The redundancy information of the object sensor was included in the well trained network, and the error was generated by the discrep-ancy between the estimated value and the sensor output. Then the fault was determined by the threshold of the error. The estimated value was utilized to realize the signal compensation when the error occured. The design was simulated with Simulink. Simulation results show that the proposed design provides a better level of fault-tolerance and higher accuracy.
展开▼
