Continuous monitoring and proper diagnosis of production systems are daily concerns that involve many manufacturers. In this context, this paper proposes a feasible and effective diagnostic methodology. It is based on a recurrent dynamic neural model application, in industrial anomaly detection, with a high identification rate. The general context of this approach is summarized in the improvement of the detection and control mechanisms using intelligent systems. These tools can collaborate objectively in industrial processes diagnosis, then in anomalies detection and classification to intervene correctly. The final purpose of this paper consists in guaranteeing the operational safety for processes, ensuring their reliability and affirming the production continuity