Faults Tolerant Control Application Using Neural Networks

In this work it is presented a fault tolerant control application using neural networks-based compensation schemes. The design consists of supervising the process possible faults using an observer that allows determining the present fault and its direction and then it will be used a classification neural network which will activate the appropriate controller according to the identified fault type. The plant to be controlled was the “Feedback Basic Process Rig 38-100”, a completely self-contained pipes circuit which drives the water contained in the inferior tank to a smaller dimension and double compartment superior tank, by means of the motor pumping located in the inferior deposit. In this work it is controlled the superior tank water level. Key-Words: Artificial Neural Networks, Fault Detection, Fault Tolerance Systems, Fault-tolerant Control,