Monitoring of industrial equipments requires the processing of a certain number of signals sensors. The aim is to detect any deviation by generating alarms. The diagnosis function has to locate then the fail and to identify the cause of the failure. Radial Basis Function (RBF) neural network seems to be a powerful tool for this kind of processing. In order to consider the dynamic of a monitoring process, we propose a new architecture of dynamic radial basis function (RRBF - Recurrent Radial Basis Function) neural network. We demonstrate also the advantages of the RRBF for dynamic monitoring problems like earlier detection of degradation and false alarms.