Full-scale roll model

The third full-scale 3-D FE model was made with the same 3 homogenized layers, but was used for coupled analysis during HT.

FEA was performed, using ANSYS Workbench (ANSYS WB, [125]) on the actual cast roll geometry (Figure 34a) and taking into account its composite structure (i.e. HSS shell, intermix zone and core). All domains are considered to be homogeneous and isotropic. The core including the journals (or necks for the bearings) are made with ductile iron core, which assumed to be homogeneous and isotropic. The CAD model of the roll was imported into the ANSYS WB Design Modeler, using “sat” (i.e. transferred only solids) file format. After importing the model, ANSYS WB automatically detects and assigns contact elements (e.g., surface-to-surface) between individual parts. All introduced contacts were described as “bonded”, meaning no separation is allowed. More detailed information about this technique can be found in ANSYS WB manual [99].

The full-scale model development is shown in multiple steps in Figure 34, where actual schematics show the geometry (Figure 34a), corresponding to the as-cast condition, i.e. before the hardening

HT. This geometry was used to recreate a 3-D CAD solid model (Figure 34b) of the roll body, which has subsequently meshed (Figure 34c). Based on the mesh convergence analysis, i.e.

difference of the results is not exceeding 5%, the mesh was made sufficiently fine in the shell layers to adequately capture the higher thermal gradients during heating and, especially, upon cooling (Figure 34c).

(a)

(b) (c)

Figure 34. Full-scale modeling: (a) original CAD drawing of the roll geometry, (b) imported and meshed CAD model in ANSYS WB

Thermo-physical and mechanical properties for each considered homogenized, isotropic layer of the full-scale FE roll model (3-layers) were predicted with the thermodynamic software, using measured chemical compositions of the improved cast (Table 2) as an input. As an example, predicted instantaneous linear expansion coefficient and Young’s modulus for the shell,

Figure 35. Predicted instantaneous linear expansion coefficients of the layers vs. temperature

Figure 36. Predicted Young’s modulus of the layers vs. temperature

In Figure 37 an example of the user-defined material database in ANSYS WB environment is shown. Five main characteristic as a function of temperature are used in the FEA for all developed models including: density, isotropic thermal conductivity, specific heat, isotropic instantaneous coefficient of thermal expansion and isotropic elasticity.

Figure 37. Generated material database in ANSYS WB

Moreover, the present work examines thermo-mechanical stress evolution during high temperature hardening HT operations (i.e. heating and quenching). The thermal stresses occur from the thermal gradients, for example, from the alternating of heating-cooling within either similar or dissimilar materials, as a result of the differences in thermal expansion behavior. There are several reasons for the thermal gradient to occur: a) relatively low (or high) initial temperature and high heating (or cooling) rates on surface; b) non-uniform thermal properties due to compositional changes; c) possible phase transformations affecting thermal characteristics during HT, which is also taken into account. Therefore, it is important to control volumetric distortions in the composite roll body that result from the thermal gradients and phase transformations during HT, preventing hot cracking and de-bonding between the layers, or even inside the “homogeneous”

Two types of thermal boundary conditions were applied on the 3-D full-scale model simultaneously: 1) tabulated surface temperature from the pyrometer readings, 2) convection at the journals, corresponding to dry air at RT, during quenching stage of the roll body. For the structural analysis only bearing supports were assigned to the journals, considering HT stage. The coupled thermal-stress analysis was also performed in ANSYS WB, where the selected unfavourable temperature loads were transferred to the steady-state thermal module and coupled with mechanical model to predict thermal stresses. In other words, the results from thermal transient analysis at particular time steps were applied as BCs in the structural steady-state thermal-stress model.

The integrated ANSYS WB environment provides the possibility of linking the steady-state thermal and static structural analyses (to predict mechanical stresses that can de-bond the HSS layer), as shown in Figure 38 (the possibility of linking simultaneously both thermal and structural transient analyses is computationally expensive and not discussed here). As it can be seen, the material library, roll geometry and meshed FE model are transferred, including the temperature results. Fixed BCs at the journals were assigned to the full-scale model in structural analysis.

Figure 38. ANSYS WB: coupled thermal-stress analysis schematic

The temperature data from a steady-state thermal solution is used as the initial temperature distribution for the static structural analysis. An example of the project tree for the coupled analysis, where the thermal loads are transferred, is shown in Figure 39.

Figure 39. ANSYS WB: project tree of coupled thermal-stress analysis

The mesh and the nodes numbering must be identical for both: thermal and structural FE models, because the resulting temperature values are assigned for each particular node. Based on that, the displacements are calculated in the static structural analysis. Therefore, the mesh size should be initially fine in order to get convergence in the structural analysis.

The allowable thermal stress-strain levels were established as a function of thermal

An example of mesh quality control for the full-scale roll model is shown in Figure 40. It demonstrates shape quality control via interactive distribution diagram, containing statistics of elements and their configurations. Simple selection (“click”) on the diagram highlighted corresponding finite elements in the model (top left window, Figure 40), which helped to identify related volumes in the model. This technique was used to improve FE composite HSS roll model based on elements shape quality in particular domains.

Figure 40. FE shape quality control in full-scale 3-D model by interactive individual bar selection