PSI - Issue 42

2

Author name / Structural Integrity Procedia 00 (2019) 000 – 000

Lukáš Suchý et al. / Procedia Structural Integrity 42 (2022) 1128–1136

1129

1. Introduction Interference fits are commonly used for shaft – hub connections in the drive trains of many common machines, such as gear boxes. In rotating applications, they are normally subject to dynamic torque and bending moments, which requires a fail-safe design in the high-cycle domain. Fatigue failures occur at the contact edges of the joints themselves, where multiaxial stresses are concentrated in all three spatial directions, even under a single load. Also, tribological stresses occur due to the different stiffnesses of the joined shaft and hub. In the field of fretting fatigue, it is known that this combined stress has a critical effect on the strengths of components. Although the factors influencing fatigue for steels can be relatively well estimated, fretting fatigue also involves the additional effects of contacts, such as contact slip, surface damage, and prestressing. Different approaches are summarized in [1], for example. In most cases, a uniaxially loaded specimen with friction pads is used to examine fretting fatigue under laboratory conditions [2 – 5]. Usually, the resulting calculation algorithms lead to larger deviations for multiaxially loaded shaft – hub connections. Also, only [6] and use fretting-fatigue pads to investigate the influence of the material within the steel grade. The uniaxial case already shows a strongly increased notch sensitivity of specimens with contact. The method that many authors e.g. [7 – 11] consider promising for calculating contact fatigue is the critical-distance method. In this, the length of the non-growing crack determines the detection location. Using the material-specific crack-structure length of a crack mode and the associated maximum tolerable stress amplitude, the point, line, area or volume method can be determined. For contacts, a crack-length tolerance is a plausible approach due to the more critical crack-initiation mechanism. However, the applicability to a multiaxial condition leaves some questions unanswered: How should the non-growing crack length be evaluated in a mixed mode? How do users calculate properly the equivalent stress and the equivalent threshold? Furthermore, the equivalent threshold depends on other parameters, such as mean stress, phase shift of the time-dependent stress curves and — especially for contacts — compression (crack-orthogonal mean stress). The threshold value is therefore very difficult to determine. Consequently, although authors like [12] determine a threshold for laboratory models in crack mode I, the load superposition leads to stress states that are again not transferable to mixed-mode fatigue. In practice, such recalculation is still very cumbersome and insufficiently accurate. In fact, in product development, load superposition in compression joints is still accounted for by using single-load notch factors in the nominal stress design [13]. Previous studies by authors of this paper [14] have already shown that this approach is too optimistic and that the products can only reliably if high safety margins are used. The reason for this is the mutual influence of concentrated loads and notch factors. A bending load, for example, leads to changed local pressure conditions, which change the notch-effect number of the torsion. Likewise, torsion further alters the slip state, which can be more critical in bending, so the notch-effect number for bending is increased. These processes are strongly material and load dependent. A more conservative calculation should be made for this reason. In the process, example cases that have been extensively tested can be prepared as support points for an application-free procedure. The aim is to reflect in particular the high material sensitivity of steels mentioned in [15] in the fatigue calculation of shrink fits. For this reason, the potential of new method for multiaxial fatigue combined with critical-distance methods is examined in this paper. 2. Experimental setup In the present study, the authors investigate the material sensitivity of the normalized C45+N steel and quenched and tempered 42CrMo4+QT steel, respectively. The authors have already published the fatigue results for shrink fits of C45+N in the preliminary study [16]. Analogously, for 42CrMo4+QT steel, interference fits with the same geometry and stepped shafts with free surface were tested. The symmetry of the entire test setup was ensured on the one hand by the shape and position tolerances of the specimens, and on the other hand by measuring the alignment when the specimen was installed in the test rig.