STUDY ON CRACK PROPAGATION TENDENCIES OF NON-REPAIRED AND REPAIRED NOZZLES

摘要

A system designed to control and predict the length of cracks that generate in the first-stage nozzles of E and F class gas turbines was developed. This system consists of three programs for (1) inputting cracks, (2) displaying cracks, and (3) predicting cracks, and a database consisting of approximately 350,000 cracks generated in first-stage nozzles taken from past repair records of five power plants operating in Japan. The database also contains data on operating time and number of starts of gas turbines. The distinctive features of this system are described below. 1) The crack data can be entered on the nozzle drawing as a picture by using the mouse. 2) The accumulated data allows the sections of nozzles in which cracks have generated most frequently to be identified. 3) The correlation formula of cracks and operating time or number of starts can be obtained simply. 4) By entering the scheduled operating time or number of start-ups to the time of the next scheduled inspection in the correlation formula, the length of cracks in optional sections and propagating in optional directions can be predicted. Using this system, the statuses of cracks generated in non-repaired and repaired nozzles of E class gas turbines were compared. The comparison focused on 11 patterns with comparatively long cracks selected from the cracks propagating together with the increase in operating time or number of starts. The propagation of cracks covering a period of approximately two years, which corresponds to the inspection interval of power plants in Japan, was also compared. The results showed that the extent of crack propagation tends to increase with the increase in the number of repairs. Furthermore, the propagation of cracks in repair nozzles is about two times greater than that in non-repair nozzles. It was also found that the system could identify the sections in which the longest cracks are generated.