Topology Optimization and Design for Manufacturing: Lighter Parts You Can Actually Make

You define a design space (the volume the part may occupy), the loads and supports, an objective (usually maximum stiffness) and a constraint (usually how much material to keep). The most common method, SIMP, gives every element a density between 0 and 1 and repeatedly moves material to where it does the most work, while penalising in-between densities so the result becomes solid or empty.

The cover image is a real topology optimization we ran for this article: a 240 × 120 mm cantilever design space, fixed on the left, loaded at the middle of the right edge, keeping 40 % of the material. After 120 iterations the material formed the familiar truss of ties and struts that engineers have long used for cantilevers, found here by the algorithm alone.

• Keep the interfaces: mark bolt holes, bearing seats and mating faces as non-design regions so they are not removed.
• Include every load case: optimizing for one load produces a part that fails under the others.
• Use realistic supports: over-stiff supports give shapes that will not work in the real assembly.
• Add the right constraints: stress, displacement or natural frequency limits, not just a volume target.
• Add manufacturing constraints: minimum member size, a pull direction for casting or moulding, extrusion, symmetry, or overhang limits for 3D printing. Ansys Mechanical and other tools include these.

The raw result is a guide, not a finished design. Engineers redraw it in CAD with smooth surfaces, standard features and sensible thicknesses for the chosen process:

• CNC machining: pockets the tool can reach, inner radii that match standard cutters, no very deep narrow pockets, and flat datum faces for fixturing.
• Casting: even sections without isolated thick spots (which cause porosity), draft and generous radii.
• Injection molding: uniform walls, ribs and draft, as described in our guide to plastic part design for injection molding.
• Sheet metal: bends, flanges and embossed ribs instead of organic shapes.
• Metal 3D printing: organic shapes are welcome, but overhangs beyond roughly 45° need supports, internal powder must be able to escape, and critical faces are usually machined afterwards.

• Fewer parts: combine parts where the process allows.
• Standard features, hole sizes and fasteners.
• Tight tolerances only where the function needs them.
• Materials that suit the process and the volume.
• Easy assembly: one assembly direction, self-locating parts and access for tools (design for manufacture and assembly, DFMA).

The redrawn part is checked with a full FEA, because smoothing and DFM changes alter the stresses. Slender members created by the optimization also need a buckling check, and parts under repeated loads need a fatigue check. Our guide to static structural analysis in Ansys Workbench covers how to run and check these analyses.

Weight savings depend on how conservative the original design was, and can reach tens of per cent. FEA results usually point to them clearly:

• In our FEA of an industrial fixture, the peak stress of 146.8 MPa against a yield strength above 250 MPa showed clear room to remove material.
• In our FEA of a battery module frame, only one small region dropped to a safety factor of about 0.41, so local reinforcement there beat thickening the whole frame.
• In our FEA of a beam bracket, the deformation plot showed which areas work hard and which barely move, exactly the information an optimization uses.

• 3D CAD of the current part and its surroundings, or drawings for our 3D CAD team to model.
• All load cases, supports and interfaces that must stay.
• The manufacturing process and material you plan to use.
• Your targets: weight, stiffness, stress or frequency limits.