Diagnostics and Damage Prediction Model for Heavy Duty Gas Turbine Combustor Hardware Failure

摘要

The focus of this paper is on the degradation of the combustion liner and developing the risk prediction model to predict the damage based on hours, starts and key operating parameters. The multivariate K-means clustering technique is used for classifying the data into different sets of clusters i.e. hours, starts or hours to start ratio vs. deformation. The effect of hour to start ratio on the liner deformation was studied, with the significant clusters obtained from K-means clustering. It is concluded that the hours-to-start ratio (N-ratio) can be a good indicator of component life and provides useful information while modeling the metallurgical damage for component life prediction. The damage growth model is developed using Liner Bulging data and it is shown that N-Ratio is a critical factor in damage prediction as well. The analysis is illustrated with the help of a limited set of combustor liner inspection data for actual heavy-duty gas turbine operation. Future guidelines provided in the paper are expected to spawn additional work in the area of advanced gas turbine diagnostics.