The aim of this work was to propose a fault detection method on the high pressure direct injection system (called Common Rail system) set up on Diesel vehicles. The importance of the fault detection procedure implementation was highlighted thanks to the description of the stakes (lowers consumption, reduction in the pollutant emissions and sound, increase of performances) and constraints dependent on Common Rail (high pressure, high frequency, gasoillubrication, high precision machining, standards EURO respect, ...) but also through a listing of failures which can occur on Common Rail. A synthesis on the different diagnosis methods of systems contributed to select a fault detection method with expected performances (detection of fault beginning, detection speed, isolation and characterization of detected fault and minimizing false alarm and bad detections). After a detailed study (properties, sequential formulations and sensitivity study) of the selected detection method (finite memory observers) and a modeling ofthe Common Rail various bodies behavior, the algorithm of detection was tested on three differentmodels of the system Common Rail. Moreover, the comparison between the finite memory observerand a Luenberger observer and a Kalman filter allow to appreciate the residual robustness degree. Obtained results allow to conclude on good detection of actuator and sensor faults.
