PSI - Issue 57

Giorgio A. B. Oliveira et al. / Procedia Structural Integrity 57 (2024) 228–235 Giorgio A. B. Oliveira et al./ Structural Integrity Procedia 00 (2023) 000 – 000

234

7

configurations. These results are presented in Fig. 5 (orange triangles), which illustrates the comparison between experimental and estimated fatigue life. It is evident that the majority of the data falls within the 1.5 band limit, exhibiting similar behavior as reported in Pinto et al. (2023).

4.4. SHEAR model + Miner rule approach results

This last approach combinesthe ANN model (SHEAR) results with the Miner cumulative damage model(Eq. 1). Hence, this modelis regarded as the simplest model and as expected,it exhibits lower accuracy compared to Miner's model (Table 4). Miner's model considers the experimental results of N H and N L lives, whereas the discussed model solely relies on estimates derived from SHEAR. Nevertheless, these estimates fallwithin the specified two band limit , as shown in Fig 5.

10 6

SHEAR + ANN-damage SHEAR + Pinto-damage SHEAR + Miner

Predicted fatigue life - N f ( cycles ) 1.5x band limit

2x band limit

10 5

10 5

10 6

Experimental fatigue life - N f ( cycles )

Fig. 5. Estimated fatigue life considering all the three models vs. experimental one for varying amplitude FF data.

Table 4. Mean percentage error of all models considered. Model

Miner SHEAR LSM (%) SHEAR ANN-d (%) SHEAR Pinto_d (%) SHEAR Miner (%)

ME (%) 17.37

19.63

5.24

24.21

25.10

5. Conclusions

This study presents four hybrid damage models, utilizing both damage rules and ANN-based approaches, to predict the fretting fatigue totallife under varying shear loading amplitudes. The Miner cumulative damage model has been used as a benchmark for comparison. The results demonstrate that, overall, all models provided good estimates, with predictions falling within the two-band limit for fretting fatigue. However, among the three models under consideration (SHEAR LSM , SHEAR Miner , and SHEAR Pinto,d ), none showed significant improvements compared to Miner's model. Although SHEAR LSM and SHEAR Miner models have the advantage of not requiring the values of total FF life under purely High or Low loading conditions as in Miner's model, their error values remain comparable. Furthermore, the hybrid model that combines two ANN models yielded more accurate estimates of fretting fatigue life compared to models combining the SHEAR model with another analytical approach, as well as the Miner's damage model. Therefore, this hybrid model proved to be the best-performing model in this study.