Control of an Electrical System by Recurrent Neural Network RSOM map

This paper discusses an intelligent strategy using the Recurrent Self Organizing Map RSOM for anomaly detection over a motor pump. The proposed idea involves the most significant features of RSOM to locate and appreciate the fault origin. It concerns the topological structure of the map, the unsupervised deep learning algorithm, as well as the Unified Modeling Language (UML) activity diagram. The measurement and treatment of the coming signal from an appropriate sensor fixed on the experimental bench, was doneat a precise moment in the Automatism and Industrial Computing Laboratory. A comparative analysis of the flaw detection performance was carried out using the RSOM neural map and the spectral analysis method.