Abstract: Faults detection and diagnosis systems, which are designed and implemented for the common rail direct injection diesel engine. Requiring models that are very near to the mechanical and thermodynamic reality process which is substantially non linear and time varying. A complete model is presented, the parameters identification using neuro-fuzzy approach, the around this model, we have built a faults diagnosis and detection structure, using the parity space and the observers. Finally, we propose a human machine cooperation structure in the monitoring loop, in order to have more reliability and an easier intervention for this process maintenance.