Figure 1A: Example of a distorted part, where the blades of an aircraft impeller have warped due to residual stresses. The red color indicates regions of high relative distortion.

One of the greatest challenges in designing high-precision parts is being able to quickly fabricate them in a repeatable way to meet tight specifications. For instance, designing aircraft engine fuel injectors demands precisely measured parts. To achieve this, engineers must often optimize specific parts and the manufacturing processes used to make them.

Additive manufacturing (AM) offers a way to create parts with less material waste and to build shapes that were previously impossible to fabricate through more traditional methods.

The Manufacturing Technology Centre (MTC) in Coventry, U.K., researches AM techniques and supplies designs and prototypes to aerospace part producers. One AM method, laser powder-bed fusion, uses powder layers tens of microns thick to build parts layer by layer using a laser. The system follows a predefined tool path to fabricate a part with very fine geometric details.

To verify the quality and performance of their additively manufactured parts, MTC researchers use COMSOL Multiphysics software for virtual design testing, validation, and performance prediction and began building apps from COMSOL models to share analysis capabilities among different teams exploring a variety of projects for customers.

A new approach

Borja Lazaro Toralles, team leader of MTC’s Physics Modelling department, says that laser powder bed fusion has certain advantages over other methods of fabrication. Deposition rates are slower than shaped-metal deposition processes but can achieve higher accuracy and resolution.

One downside is that as the metal cools, deformations can occur after a few layers have been built. Thermal cycling due to the high temperature gradient and quick cooling can cause residual stresses during deposition. This slowly alters the microstructure, which causes distortions in the final part. (See Figure 1A.)

Figure 1B: Final impeller design after distortion adjustments.

Some deformations are negligible, but in others, a difference of just 100µm (0.1mm) can be too far off specification. For these situations, the MTC team needed a way to get around the effects of thermal cycling. Since they couldn’t remove the thermal cycling and the evolution of the microstructure, they approached it another way.

“We created a simulation that predicts the stresses and deformation during a part build to give us a clear understanding of how it will distort during printing,” Toralles says. “Once we have this information, we can invert the distortion in the part’s design, which allows us to account for the warping ahead of time so that the final product distorts into the shape we actually want.”

Working backward from errors and building them directly into designs has helped researchers create parts within the required tolerances more efficiently, knowing that the predictive model will guide them to a shape that results in minimal error. (See Figure 1B, above.)

Adopting multiphysics simulation has also opened new lines of communication with the MTC’s design for AM team. Toralles’ team built an app around their COMSOL model for predicting distortions, which allows their colleagues to run the simulation and see where designs need to be changed without having to completely understand the original model.

Before sharing an app with their part designers, the team needed to build the high-fidelity model on which the app would be based.

Figure 2: Simulation results showing displacement in the impeller to predict the final part shape.

Modeling for complex, varied parts

To build a model that gives design engineers the information needed to make appropriate design adjustments, Toralles and his team defined a new modeling process that would predict the final shape of large parts.

“Traditional additive manufacturing models are very detailed, down to the microstructure. But these are not suitable for simulating large part builds because of the computational cost,” he explains. “They take forever. But we still need to understand how an entire part will behave during printing. To circumvent this, we lump the layers of the print build and impose an analytical temperature field based on experimental data. This reduces solving time but still gives an accurate solution.”

Since the MTC works with a variety of metals, from aerospace grade titanium to stainless steels, they parameterized inputs in the COMSOL model – material properties, lumped layer thickness, build plate fixturing on the part, mesh element size – to use the model to test parts of many shapes, sizes, and materials.

Within the software environment, they used the structural mechanics functionality to define a linear elastic material with temperature-dependent plasticity and thermal expansion, using the analytical temperature field. The app also performs several automated CAD operations to prepare the geometry for suitable lumped layering.

The simulation then generates a grid to represent an approximate toolpath, given that the lumping of layers makes it impractical to use the real one. The temperature field is imposed on the grid points. The software then computes the stresses generated during deposition and predicts the final shape of the part. (See Figure 2, page 18.)

Spreading predictive capabilities

Once the simulation predicts the errors in a part, getting the information to the design team is another matter. Many companies have separated design and simulation groups, benefiting from employees who specialize in one or the other. But this leaves a gap between part design and part analysis.

Apps bridge this gap by allowing simulation specialists to package their models into user-friendly interfaces, which designers use to run their own tests without needing to understand all the complexities of the simulation underneath. This allows design engineers to make adjustments, saving simulation experts from running an analysis every time a new part’s performance needs to be evaluated.

The design team creates lots of complex parts using a wide range of CAD platforms, so the model and corresponding app needed to be quite robust. It includes CAD import features so that any shape can be tested. This is especially important for organic shapes – those based on natural features like plants, animals, and land formations – such as those drawn in a program such as Rhino.

The app, created with the Application Builder available in COMSOL Multiphysics, displays the simulation results – final shape, deformation, and stress levels for a given part, in this case an aircraft impeller. (See Figure 3, page 20.)

“[The simulation team] often use the app ourselves,” Toralles adds. “Once we built it, it was easier to make a few input changes in the app rather than go back to the original model. But the design team doesn’t work on simulations. The app was built for them because it allows them to import part models and check the prediction of how it will warp during printing.”

The Application Builder included in the software allowed Toralles to have full control over what was available to the app user. As the app has evolved based on new needs from the company, he has been careful to build in the necessary outputs and displays, as well as locking certain inputs and conditions so that app users cannot inadvertently create errors. The underlying model setup remains hidden from the user, but the simulation capabilities have spread.

Figure 3: The MTC app enables the user to make design adjustments and test changes in the simulation, but without showing the underlying multiphysics model.

Solidifying collaboration

Toralles deploys the app through the COMSOL Server for distributing, managing, and running simulation apps. Hosting the app online makes it available for colleagues who have been given access, anywhere in the company.

Simulation has changed the way teams work together at the MTC. Now they have an established routine where designers and simulation engineers can communicate quickly, test designs easily, and make changes that result in the desired prototypes for their customers to manufacture on a wide scale.

Toralles admits that there was skepticism when they first began offering simulation apps to other departments. “We had to earn their trust. But since the model consistently provided good results, everyone has found it helpful. The app contains everything: it slices the print geometry, shows the mesh, guides engineers through the distortion analysis, and provides feedback.” Their simulation work has been an important part of helping the MTC establish a workflow that improves communication between the physics modeling team and the design team. This, ultimately, has changed the way they approach part design.


Manufacturing Technology Centre (MTC)