PSI - Issue 72

Nenad Vidanović et al. / Procedia Structural Integrity 72 (2025) 499– 506 505 values for exponent are similar to their initial values (difference for Model 1 |Δ 1 | = 3.98%, and difference for Model 2 |Δ 2 | = 0.27%) which directly means that aged material still has similar exponent as virgin material ( is not significantly changed after 4 580 ground-air-ground cycles). Based on previously mentioned, it can be concluded that parameter is not dependent on the crack shape (the existence or inexistence of the notch in the FE model), it is not dependent on the FE mesh density, and it is not dependant on extreme working conditions in this section of the engine. On the contrary, there is a large difference between obtained values for coefficient (|Δ| = 81%), and the differences between obtained and initial values for both mod els are: |Δ 1 | = 22.94% and |Δ 2 | = 6.15%. This means that is highly dependent on the crack shape, it is highly dependent on the FE mesh density, and as a general conclusion, aging of the Inconel 718 due to high pressures and temperatures is followed by the change of the coefficient Ǥ Based on calculated Paris’ coefficients residual life of the HPT case could be presented in the form of − diagram (Fig. 7.) where presented results show a significant difference in the number of cycles to complete failure for these two models.

Fig. 7. Comparison of residual life for Model 1 and Model 2.

Like the diagram presents, the number of cycles N = 2 807 to grow initial cracks to the targeted length of a = 0.41 mm is the same for both models, as demanded, but with difference in necessary number of steps (nos 1 = 30 & nos 2 = 17), and where Model 1 (N 1 = 11 479, nos1 = 413) predicts a much shorter life to failure than Model 2 (N 2 = 26 638, nos 2 = 354). The cut in Model 1 caused fatigue crack appearance and significantly accelerated crack growth through the structure of the HPT casing. Since this component had been used for 2 807 cycles until the crack of length 0.41 mm was discovered, the obtained residual life is approximately 4 times longer than the life until the problem was recognized which means that the damaged component will be safe for use at least for next few years. Also, it is obvious that crack appearance and growth in Model 2 is highly unlikely and can be caused only by some kind of accident, which implies that the crack would never appear in this area if the notch had not been made there. Simulation for Model 1 lasted 33h with allocated 95.8 GB RAM, where simulation for Model 2 lasted 17h and with allocated 56.2 GB RAM. These differences are explained by the fact that a finer mesh for Model 1 was needed than for Model 2, which is caused by a 20 times smaller radius of the initial crack for Model 1. Workstation engaged for simulations is equipped with Threadripper PRO 5955WX with 16-Cores (4.5GHz), and with 128 GB RAM. 4. Conclusions Based on all the above-elaborated in the main text, optimisation approach presented in this paper, as well as the highly reliable framework based on numerical simulations, can be successfully used to obtain Paris’ coefficients for aged materials based on real operating data: