The problem of fault detection and isolation has been introduced into automatic control as a paradigm for the design of algorithms capable of detecting the occurrence of faults and isolating their causes. Many methods have been developed for the class of integer systems. In contrast, few methods have been developed for the class of fractional order systems. In this paper, a new fault detection and isolation method, based on the flatness property, is introduced for the class of fractional order linear systems to detect and isolate faults on sensors and actuators. In this method, the measurement of the so-called fractional flat output vector is used to construct the redundant components of the system and to generate fault indicators called residual signals. In addition, new definitions of fault detection and isolation are introduced. The effectiveness of this method is validated by simulations on a fractional order system, the thermal bi-dimensional system.