Application of improved BP algorithm to fault diagnosis of steam turbine-generator sets

A new network training rate controlled by a constant error correction rate and a selfadaptive momentum coefficient adjustment method were put forward to shorten the convergence time and overcome the local optimization. A fault diagnosis model based on the improved BP network was established for the vibration faults of steam turbinegenerator sets. The application of the model to diagnosis of a practical generator set shows that the method is effective.