Abstract:

Machine availability and reliability are two of the most essential concerns for a gas turbine power plant system. A good maintenance program that increases power plant availability while reducing the losses due to unplanned shutdowns should be instituted. A Risk Based Maintenance (RBM) methodology is developed in this paper. It calculates the future risk of failure of a gas turbine power plant system so that the maintenance can be planned just before occurrence of failure. To calculate the risk, first a General Log Linear Lognormal (GLL- Lognormal) model, which tells about damage growth of the machine, is developed. Bayesian approach is then used to update the model parameters (i.e., GLL- Lognormal parameters) on the basis of new inspection data (i.e., crack length) and calculate the updated risk. It is recommended that risk should be continuously updated with the age of the unit to increase the effectiveness of RBM policy. The novelty in this work is that the failure probability is directly dependent on observed crack length instead of time to failures. The whole analysis is illustrated with cap effusion plate inspection data of actual gas turbine system. It is found that the proposed risk based approach gives more accurate results than a normal fleet level model.