This work presents the development of a fault detection and identification module for multirotor unmanned aerial vehicles (UAVs), capable of detecting a total failure in any of its rotors. The solution is based on a supervised learning approach, for which random forest and support vector machine classifiers have been trained using simulated data, and proved to be feasible to implement in real time. To validate these models, experimental proof will be shown of a classifier running in real time onboard a particular fault tolerant hexarotor design, showing the fastest detection times in this vehicle to date.