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

229

2

1. Introduction The fretting fatigue (FF) phenomenon has attracted significant attention due to its damagingeffects on the fatigue life of mechanicalassemblies. Fretting is defined as the micro slip between contactingparts. If one of those parts are also subjected to fatigue loads, one has the so-called fretting fatigue phenomenon. In this case, the high local contact stresses might lead to considerable decreases in the fatigue life of components subjected to fretting (Nowell et al., 2006). Even nowadays, numerous approaches are continually being proposed to address the fretting fatigue phenomenon due to all the complexities involved in this problem, for instance, the presence of mutiaxial stress states and strong stress gradients (Almeida et al, 2023; Dieu et al, 2023; Glodek et al, 2023; Moreno-Rubio et al., 2023). Due to the complexities involved in experimental setups, fretting fatigue problems have mostly been tested in laboratories using constant amplitude loading. However, this condition does not reproduce what truly occurs in practical applications, where variable amplitude loading predominates. Recently, some studies have focused on addressing this issue by attempting to replicate experiments with variable amplitude loading (Kouanga et al., 2023; Matos et al., 2023; Pinto et al., 2023; Rousseau et al., 2019). Additionally, a recent trend in the field of FF is the utilization of Machine Learning techniques (Brito Oliveira, 2022, 2023; Han et al., 2023; Liu and Yuan, 2023), which has gained popularity due to its excellent results and generalization capabilities. In this setting, this work aims to test some damage model , inspired in Miner’ s cumulative damage mode, combined with a consolidated ANN-based model to predict the fretting fatigue life under varying shear loading amplitude. 2. Fretting fatigue under variable amplitude data In this study, one considers the experimental FF data from Pinto et al. (2023). The material is the Al 7075-T651, which has an elastic modulus of 68 GPa, a Poisson's ratio of 0.33, a yield stress of 506 MPa, and an ultimate tensile stress of 570 MPa. The specimens have a cross-section of 13 mm × 13 mm, their geometries and applied loads are depicted in Fig. 1. The tests were performed under load control. Tables 1 and 2 present the key experimentalvalues that will be employed in this work. The friction coefficient ( ) is also presented in Table 2. Table 1. Fretting fatigue tests under H-L and L-H loading sequence. H-L (High-Low) L-H (Low- High) d 1 n 1 (cycles) n 2 (cycles) N f (cycles) d 1 n 1 (cycles) n 2 (cycles) N f (cycles) 0.75 1.07E+05 2.26E+04 1.30E+05 0.75 2.61E+05 5.79E+04 3.19E+05 0.75 1.07E+05 4.48E+04 1.52E+05 0.75 2.61E+05 6.29E+04 3.24E+05 0.5 7.16E+04 9.64E+04 1.68E+05 0.50 1.74E+05 8.70E+04 2.61E+05 0.5 7.16E+04 1.25E+05 1.97E+05 0.5 1.74E+05 7.90E+04 2.53E+05 0.5 7.16E+04 1.34E+05 2.06E+05 0.5 1.74E+05 6.71E+04 2.41E+05 0.4 5.73E+04 2.01E+05 2.59E+05 0.35 1.22E+05 9.22E+04 2.14E+05 0.3 4.30E+04 2.55E+05 2.98E+05 0.25 8.70E+04 9.80E+04 1.85E+05 0.25 3.58E+04 3.33E+05 3.68E+05 0.25 8.70E+04 1.76E+05 2.63E+05 0.25 3.58E+04 3.04E+05 3.40E+05 0.25 8.70E+04 1.69E+05 2.56E+05 0.25 3.58E+04 3.09E+05 3.45E+05 0.25 8.70E+04 1.50E+05 2.37E+05 - - - - 0.15 5.22E+04 1.45E+05 1.97E+05 In Table 2, the main loads are reported, named High and Low due to the shear load ( Q ) amplitude. In the conducted tests, only the tangential load amplitude, Q , varies, while the amplitude of the bulk fatigue load, , remains constant, and the normal load, P , remains unchanged. Fig. 1a illustrates a High-Low (H-L) loading sequence, where the tangential load amplitudes analyzed are Q 1 and Q 2 , corresponding to a given number of cycles, n 1 and n 2 , respectively, with Q 1 > Q 2 . Conversely, Fig. 1b represents a Low-High (L-H) loading sequence, where the first tangentialload of amplitude Q 1 is applied, followed by the one of amplitude Q 2 ( Q 1 < Q 2 ). Table 1 shows the variable shear loading FF data, under H-L and L-H loading configurations, such that n 1 represents the predefined number of cycles for the first load block (High or Low), which produces a damage d 1 according to Miner’s rule. In this setting, n 2 is the remaining number cycles experienced by the second load block until failure is observed. Therefore, N f is the totalFF life ( n 1 + n 2 ) in the tests. Table 2 presents the experimental FF lives of tests conducted considering one single loading block for each case, High or Low.