Page 370 - Handbook of Materials Failure Analysis
P. 370

368    CHAPTER 14 Fatigue failure analysis of welded structures




                         less than the required value, cracking might occur, especially in the bottom sheet, due
                         to the severe plastic deformation.
                            Magnesium has hexagonal crystal structure which results in limited slip systems
                         and poor ductility and formability at room temperature; therefore, SPR generally
                         causes cracking in magnesium alloys [49]. The formability of magnesium improves
                         at higher temperature because of the additional slip systems activated [50,51]. Dur-
                         andet et al. [49] used a laser beam at the joint location and showed that it can provide
                         enough heat to prevent initiation of cracks in AZ31 magnesium alloy. This method is
                         called “laser-assisted self-pierce riveting” or LSPR, and it was employed for joining
                         the components of the Demo-structure.


                         3.3 MATERIAL PROPERTIES AND MODELING
                         3.3.1 Material properties
                         As stated in Section 3.1, the components of the Demo-structure are made of three
                         different magnesium alloys: AZ31B sheet, AM30 extrusion, and AM60B cast. In
                         order to simulate the Demo-structure, the mechanical properties for these alloys were
                         obtained from the literature [10,52,53]. The microstructure of the three alloys differ
                         as a result of the different manufacturing process. The rolling and extrusion pro-
                         cesses produce highly textured microstructures for AZ31B and AM30; therefore,
                         material properties for these alloys are different under tension and compression.
                         On the other hand, a randomly oriented crystal structure is observed for AM60B cast,
                         which results in symmetric tension-compression behavior.
                            Figure 14.8 shows the monotonic behavior of the three alloys, in which RD, TD,
                         and ED represent the rolling, transverse, and extrusion directions, respectively.
                         Table 14.1 summarizes the monotonic properties of the three magnesium alloys in
                         different directions.
                            Similar to monotonic loading, cyclic loading experimental results for the three
                         Mg alloys were obtained from the literature [10,54,55]. The stabilized hysteresis
                         loops for the three alloys at different strain amplitudes are shown in Figure 14.9.
                            The strain-life curves were obtained for the three alloys under fully reversed load-
                         ing [10,54,55] and are shown in Figure 14.10. In this figure, the fatigue properties
                         correspond to RD for AZ31B, and ED for AM30 magnesium alloys.
                            The monotonic and cyclic properties of the rivets are unknown, and therefore the
                         elastic steel properties, that is, E ¼ 210GPa and ν ¼ 0:3 were used in the simulation.

                         3.3.2 Material modeling
                         After material properties are obtained from experiments or from the literature, the
                         next step is to generate a FE model for the structure of interest and assign the cyclic
                         stress-strain curves to the corresponding components. However, for the case of the
                         Demo-structure the procedure is more complex than general problems. The wrought
                         Mg alloys involved in the Demo-structure behave asymmetrically under tension and
                         compression and this behavior could not be modeled by available material models in
                         the Abaqus FE package [56]. Therefore, an asymmetric constitutive model was
   365   366   367   368   369   370   371   372   373   374   375