Abstract : The energy stored in the Large Hadron Collider is unprecedented. The impact of the beam particles can cause severe damage on the superconductive magnets, resulting in significant downtime for repairing. The Beam Loss Monitors System (BLMS) detects the secondary particles shower of the lost beam particles and initiates the extraction of the beam before any serious damage to the equipment can occur. This thesis defines the BLMS specifications in term of reliability. The main goal is the design of a system minimizing both the probability to not detect a dangerous loss and the number of false alarms generated. The reliability theory and techniques utilized are described. The prediction of the hazard rates, the testing procedures, the Failure Modes Effects and Criticalities Analysis and the Fault Tree Analysis have been used to provide an estimation of the probability to damage a magnet, of the number of false alarms and of the number of generated warnings. The weakest components in the BLMS have been pointed out. The reliability figures of the BLMS have been calculated using a commercial software package (Isograph™). The effect of the variation of the parameters on the obtained results has been evaluated with a sensitivity analysis. The reliability model has been extended by the results of radiation tests. Design improvements, like redundant optical transmission, have been implemented in an iterative process. The proposed system is compliant with the reliability requirements. The model uncertainties are given by the limited knowledge of the thresholds levels of the superconductive magnets and of the locations of the losses along the ring. The implemented model allows modifications of the system, following the measuring of the hazard rates during the LHC life. It can also provide reference numbers to other accelerators which will implement similar technologies.