Design of fault detection for a hydraulic looper using dynamic neural networks

The contribution investigates the design of a fault detection system for a hydraulic looper from a hot rolling mill plant. A genetically evolved dynamic functional-link neural network is used to identify different relationships between process variables. One step ahead prediction errors, i.e. residuals, are then evaluated by a decision logic using threshold values and different patterns of residual's change in cases of known faulty behaviours. Experimental results are included referring to the separate processing of simulation and real data. The obtained results characterise the efficiency of the presented approach as depending on the availability of signals from the automation system of the considered process.