The paper examines statistical data on failures of gas-dynamic equipment, exemplified by a fleet of stationary gas turbine units (GTUs) over a 10-year operational period from the perspective of applying a regression model for predicting gas-dynamic equipment failures. The authors conduct an analysis of data on the number of failures of various types (electrical supply and equipment malfunctions – r1, automation and safety system malfunctions – r2, mechanical equipment malfunctions without node and component destruction – r3, mechanical equipment malfunctions with node and component destruction – r6) depending on operating time after repair. The selected generalized linear model with a Poisson distribution allows estimating the average number of failures and demonstrates a significant contribution of both operating time and the failure indicator. The results are compared with the Weibull curve (bathtub curve) concept to confirm the identified patterns. The work establishes that the most significant failure indicator is r2, which also correlates more with other indicators with the wear-in period stage of the Weibull-Gnedenko curve.
Simonov et al. (Tue,) studied this question.