This paper deals with a decentralized fault-tolerant control methodology based on an Active Fault Diagnosis approach. The proposed technique addresses the important problem of monitoring interconnected Large-Scale Systems (LSS). The fault diagnosis approach is made of a passive set-based fault detection method and an active fault isolation technique, able to guarantee isolability subject to local input and state constraints. The proposed scheme can be implemented locally in a decentralized way. A significant feature is the decentralized design constructed on tube-based Model Predictive Control to possibly allow the disconnection of faulty subsystems or the reconfiguration of local controllers. The Active Fault Diagnosis tool is designed to support the decision-making process for the control and monitoring of the LSS.