Online neuro-inverse dynamics controller for nonlinear induction furnace system: Fault hiding approach

In this paper, an online neural inverse controller is used to deal with actuator faults. In such a way that the inverse of the nonlinear induction furnace system (IFS) is used as a fault-tolerant controller (FTC) so that it can cover the fault of the actuator. The design is such that an online neural network is used to model the NIFC, the three-layer neural network is ‎converted into a four-layer RBF neural network, and the last ‎layer is the ‎nonlinear IFS, and this layer is It is unchangeable and the ‎controller and the system are connected and finally ‎form a four-layer neural network. So, an intelligent inverse ‎model of the IFS is used as  FTC to cover the actuator fault of ‎the ‎nonlinear IFC. This controller design is done in two ways: in the ‎first part, five inputs are used for training the neural network, ‎one of which is the neural network training error, but in the ‎second part, in addition to the five inputs of the first part, the ‎derivative of the error is used. And the error integral has also ‎been used in neural network training and the advantage of the ‎second plan is to reduce overshoot. Finally, a fault actuator is ‎applied to the ‎nonlinear IFS in the ۱۰th to the ۳۰th second,  ‎despite the presence of the intelligent FTC, this defect is ‎covered in less than one second, and the system continues to ‎function normally despite the operator's defect in this interval ‎of time.‎

کلیدواژه ها:

Inverse Neural Control ، Fault-tolerant control ، Induction Furnace ، RBF Neural Network