PSI - Issue 34

N. Lammens et al. / Procedia Structural Integrity 34 (2021) 247–252 N. Lammens/ Structural Integrity Procedia 00 (2019) 000 – 000

251

5

The results in Figure 5 show the conservative predictions of the material model, illustrate a good correspondence between experimental results and predicted performance from the material model. Note that the machine learning model is capable of predicting fatigue performance for parameter settings that have not been tested experimentally, which is an essential requirement in order to analyze a complex geometry such as the present use-case, where variable surface roughness and orientation (between build direction and principal stress direction) will be found. 4.2. AM-aware fatigue prediction In order to properly predict local fatigue performance throughout the use-case component, surface roughness is estimated at each element in the FE model and used as an input variable to the Durability calculation. Several simple test-coupons were printed, and surface roughness was measured for different overhang angles. A least-squares fitting approach was adopted to fit a continuous curve through the test data while ensuring correct periodicity of the fitted curve is respected. Figure 6 shows the fitted roughness curves (left) and the mapped surface roughness (right) for the use-case discussed in this work. The element-by-element definition of surface roughness is directly integrated into the Durability calculations to update the material performance locally.

Figure 6. Calibration curve for roughness (left) and predicted roughness on the use-case component (assuming a Z-axis build direction) Performance prediction

Using the machine learning model, the predicted surface and an FE model recreating the cantilever loading condition representative of in-service loading, a Durability analysis is performed using an extension to the Siemens Simcenter 3D Specialist Durability solver. The FE analysis (Figure 7 – left) indicated two potentially critical areas requiring more detailed analysis. The first zone is very close to a supported area during printing (thus possibly having been machined when the support structures have been removed). Principal stress directions were calculated in order to define the correct material performance at each location within the model. Figure 7 (right) shows the predicted component fatigue behavior at the two identified zones. As can be seen, the Durability of the component is dependent on the precise surface condition present in Zone 2. In an as-built condition, Zone 2 will lead to early failure at less than 20.000 cycles, while a machined condition will both increase the lifetime of the component to 500.000 cycles and shift the failure location to Zone 3. This highlights the importance of properly accounting for localized material conditions and proper assignment of material performance at each location.