Basic factors to forecast maintenance cost and failure processes for nuclear power plants

Two types of maintenance interventions are usually administered at nuclear power plants: planned and corrective. The cost incurred includes the labor (manpower) cost, cost for new parts, or emergency order of expensive items. At the plant management level there is a budgeted amount of money to be spent every year for such operations. It is very important to have a good forecast for this cost since unexpected events can trigger it to a very high level. In this research we present a statistical factor model to forecast the maintenance cost for the incoming month. One of the factors is the expected number of unplanned (due to failure) maintenance interventions. We introduce a Bayesian model for the failure rate of the equipment, which is input to the cost forecasting model. The importance of equipment reliability and prediction in the commercial nuclear power plant is presented along with applicable governmental and industry organization requirements. A detailed statistical analysis is performed on a set of maintenance cost and failure data gathered at the South Texas Project Nuclear Operating Company (STPNOC) in Bay City, Texas, USA.