Development of Predictive Maintenance System for Nuclear Power Turbine Unit

The safe operation of nuclear power plants is the most concerned issue in the nuclear power industry. Intelligent early warning and predictive maintenance of mechanical equipment such as steam turbines can effectively prevent nuclear power plant safety accidents and reduce maintenance costs. This paper will introduce the architecture and process of predictive maintenance system and apply it to nuclear power steam turbine units. The developed predictive maintenance system of nuclear power steam turbine units is equipped with LSTM-Bayes signal anomaly recognition method, benchmarking other manufacturers' systems, and improving many indicators. For example, new early warning parameters such as abnormal probability, remaining useful life and characteristic signals are added to help maintenance, making it more intelligent. It has been verified that the input of the system reduces the failure rate of mechanical equipment, saves maintenance costs, can effectively ensure safe operation, improve productivity and reduce costs.