Share:

ANSYS 19.2 Release Highlights

Probabilistic Determination of Crack for Components Subjected to TMF

In recent times, there have been major recalls for transport sector components like exhaust manifolds, turbocharger casings, etc., and unanticipated crack growth has cost OEMs thousands of dollars in warranty costs. Minor cracks in these components are prone to carbon monoxide leakage and can be hazardous to health. And, once cracks appear, the components may no longer be serviceable. The steady trend toward high-efficiency, slick, lightweight components, subjected to high thermomechanical loads is pushing designers to find more confidence on the final crack-safe designs before product deployment.

Their major challenge is identifying the most vulnerable locations for cracking in the design development stage. Because of the uncertainties in the characterization of material properties and thermo-structural loading, locations of the hotspots may not show up in their true locations in deterministic simulation, or correctly predict a conservative fatigue life estimate. Additionally, these high temperature components are typically mold-casted and variations in thickness from the design blueprint are very common. Such manufacturing variations are highly random in nature and cannot be controlled unless hotspots are identified with certain probabilistic confidence. Together, the aforementioned uncertainties are making deterministic virtual prototyping less useful.

ANSYS optiSLang, along with ANSYS partner Dynardo’s Statistics on Structure® (SoS®) software, provides a high-fidelity solution to quantify the design uncertainties and probabilistic assessment of crack locations through correct thermo-mechanical fatigue (TMF) life estimate. ANSYS’ state-of-the-art material models for TMF are coupled with random variations in loading and manufacturing tolerance. Probability assessments of critical life locations are performed using optiSLang and SoS.

Manufacturing variability can be addressed through model-order reduction (ROM) techniques coupled with mesh morphing. Typically, a large number of 3D-surface scan data for production models can be processed in SoS to find the mean and standard deviation from the nominal mesh grid point. A random field model can be generated that can explain any nominal random variation in geometry by morphing the mesh. Even if there is no scan data available, a manual definition of mean and standard deviation for the nominal mesh can generate a synthetic random field model from the mesh node distance based on an autocorrelation matrix. Also, a blend of a limited number of scan data and manual definition can also be used to generate the random field model of geometry variation.

Synthetic random field model generated for an exhaust manifold outer-surface variationFigure 1 shows a synthetic random field model generated for an exhaust manifold outer-surface variation. The generated shape represents a normal distribution of directional mesh movement values over the surface. In this case, the synthetic random field model generates 14 shapes that, in combination, can explain any arbitrary random variation on the outer surface with up to 95 percent accuracy. Shapes are deterministic in nature but each is scaled with a random amplitude (parameters for design of experiment) and combined algebraically to arrive at a final, arbitrary random variation. The limit of the maximum variation is ±0.5 mm on top of the nominal 5 mm thickness of the manifold wall. Compared to a traditional mesh morpher, in which each morphing node will be a separate parameter, the number of shapes changing parameters here is reduced to few. Each of the random geometry variations is generated in a design of experiment (DOE) study.

Quantile plot of 95 percent non-exceedance of manifoldAny TMF material model parameter can be used as an input parameter that is subjected to any random variation in the material casting process. Cyclic temperature and temperature-dependent film coefficients of convection can be parameters as well for a DOE study. Additionally, structural loads like bolt pretensions can be used as random inputs. Finally, the life of the component can be estimated through a crack tip opening displacement (CTOD) approach.

For this case study, sensitivity analysis is performed using the Latin hypercube sampling scheme. Critical locations, where life values are low and prone to crack appearance, are estimated through a quantile plot of 95 percent non-exceedance as shown in Figure 2. Legend values represent the life values which will not be exceeded 95 percent of the time at those locations within the given range of parameter variation. Figure 3 shows the important parameters that influence the minimum life. It also shows how the four geometry changing shapes influence the life of the manifold.

>Figure shows the important parameters that influence the minimum life. It also shows how the four geometry changing shapes influence the life of the manifold.

With the above approach, ANSYS software can comprehensively predict the critical life of the component and its location, and further reduce expensive physical prototype casting and testing. It will also accelerate the design development time from six months or one year to one month.

Learn More

Read more about thermo-mechanical fatigue in this white paper.

Share: