5. RS predictions by finite element simulations
5.4. Summary of chapter 5
For cold-water quenched plates, finite element simulations of quenching showed that: • In the as-quenched 20 mm thick AA7449 plate, the RS at mid-thickness predicted
with or without precipitation taken into account compare well with the measurements. Indeed, the cooling of the 20 mm plate is so fast that precipitation hardening is limited. However, it is not negligible as revealed by the 50 MPa difference between simulations with and without precipitation at the surface. • In plates thicker than 20 mm, precipitation has to be taken into account since the
RS underestimation by the TM model increases with increasing thickness. Thus, as plate thickness increases, a model with precipitation becomes necessary to predict reasonably well the RS profile.
Two models were used to account for precipitation in heat treatable aluminium plates: • A Thermo-Metallurgical-Mechanical model taking into account precipitation using
a precipitation model in a one-way coupling temperature→precipitation→stresses (TMM model).
• A Thermo-Mechanical model taken into account precipitation in a simple-way using a few Gleeble interrupted quench-tests (TMG model),
In the TMM model, the yield strength is position dependent through the precipitation and yield strength models. In the TM and TMG models, the yield strength parameter is considered constant through the thickness and interpolated linearly as a function of temperature.
The TMG and TMM models predict identical RS profiles in the as-quenched 20 mm thick AA7449 plate. In the as-quenched 75 mm thick AA7449 plate, very little differences were found between the RS profiles predicted by these two models. For these two thicknesses, the agreement between measurements and simulations is excellent, thus validating both models.
For the TMG model, this shows the relevance of a model based on tensile tests after surface coolings similar to the industrial ones. Indeed, the TMG model provided also an excellent agreement between measurements and simulations in the as-quenched 75 mm and 140 mm thick AA7040 plates.
For the TMM model, this shows the quality of the calibration of the precipitation and yield strength models. Indeed, the calibration was done on SAXS coolings slightly different from surface coolings.
Compared to the TMM model, the TMG model:
• does not require a precipitation model. Therefore, the characterisation of precipitation kinetics to calibrate the precipitation model is not needed,
• decreases considerably the number of Gleeble tests,
• is easier to implement in a FE code especially for complex geometries, • is less expensive in terms of computational time and memory.
158 However, the major drawback of the TMG model is that it is not fully predictive. Indeed, it requires a few tensile tests for each alloy and plate thickness. Therefore, the physically- based TMM model remains necessary for modelling complex parts with different thicknesses or plates with a full range of thicknesses without achieving specific Gleeble tests for each thickness.
The TMM model was applied to a thicker (140 mm) AA7449 plate for which no RS measurements are done since such thick AA7449 plates are not industrially produced yet. The predicted surface and mid-thickness as-quenched residual stresses are slightly higher in absolute value than those in the 75 mm AA7449 plate. This is consistent with the trend observed for AA7040 plates for which measurements in 75 mm and 140 mm plates were performed.
For boiling-water quenched AA2618 forgings, finite element simulations of quenching showed that:
• In the small AA2618 forging, the residual strains predicted without precipitation compare well with the measurements. Almost identical residual strains were predicted by a TMG model based on Gleeble interrupted quench-tests at 20 K/s, suggesting that the effect of precipitation is small for this forging.
• In the large AA2618 forging, the residual strains predicted without precipitation compare well with the measurements except in the region where tensile stresses are maximal. There, the simulation without precipitation underestimates residual strains. The agreement was improved using a TMG model based on Gleeble interrupted quench-tests at a cooling rate of 20 K/s, meaning that precipitation hardening during cooling has to be taken into account to predict reasonably well residual strains and stresses.
As mentioned for the plates, a TMG model is not fully predictive. For the forgings, a TMG model based on Gleeble interrupted quench-tests at 20 K/s was used. This means that the effect on RS of large precipitates forming during quenching was ignored since it was found in chapter 3 that the high critical cooling rate of S phase in AA2618 is ~17 K/s. For the investigated forgings, the relatively good agreement between the measurements and results of the TMG model based on Gleeble interrupted quench-tests at 20 K/s indicates that high temperature precipitation can indeed be neglected. However, this might not be the case for larger forgings with slower coolings. Further work is required to determine the forging size above which high temperature precipitation of S phase will significantly affect RS. This could be done by means of:
• a TMG model based on Gleeble interrupted quench-tests at low cooling rate (e.g. 2 K/s).
• a TMM model based on a precipitation model involving at least two phases S and
S’. For this, the calibration at high temperature (≥ 250°C) performed in section