3DM Launches SLS 3D Printing Solution: Laser Can Dynamically Adapt Material Melting Characteristics

Standard laser powder bed fusion runs a fixed parameter set: laser power, scan speed, hatch distance, and spot size are chosen per alloy and layer thickness, then held constant across the whole build. That works well on simple geometry. On a part with thin walls, thick bosses, and a large flat flange, the local cooling rate changes from feature to feature. The same energy density that fully melts a 3 mm rib can leave lack-of-fusion porosity in a heavy section, and can over-melt a thin one.

The consequences show up in different places. Lack-of-fusion defects cluster near contour boundaries where the scan strategy changes direction. Keyhole porosity forms when the melt pool overheats in thick sections. Residual stress builds where the thermal gradient is steepest, which is usually right at the joint between a thin wall and a solid base. All three are geometry-dependent, so a parameter study on a test cube does not predict them on a real bracket.

Adaptive control attacks this directly. Instead of holding power and speed fixed, the system reads the melt pool and shifts the laser on the fly. The claim behind 3DM's platform is that the laser follows the material's melting characteristics rather than a pre-set table. If the loop is fast enough, it can hold melt pool size steady across a section change.

• 1
  Fixed recipeOne power and speed per alloy and layer thickness, for the whole build
• 2
  Geometry effectSection thickness changes cooling rate, so the same energy density gives different results
• 3
  Defect patternLack-of-fusion near contours, keyhole porosity in thick sections
• 4
  Adaptive loopReads the melt pool and adjusts laser output during the scan

Three variables matter in a closed-loop powder bed system: laser power, scan speed, and spot size or focus offset. Some platforms also modulate the scan strategy itself, switching between hatch and contour passes based on what the sensor sees. The sensor is usually a photodiode or a coaxial camera reading the thermal emission of the melt pool. The control loop then compares that signal to a target and corrects.

The hard part is time. A typical scan speed runs between 600 mm/s and 1,200 mm/s. At 1,000 mm/s, the laser crosses a 10 mm feature in 10 ms. Any control loop slower than roughly a millisecond cannot correct within a single feature, so it can only react between layers. That is still useful, but it is a different capability than intra-layer correction. When you evaluate any adaptive SLS system, ask where the loop closes: within a layer, or between layers.

There is also a calibration cost. The melt pool signal depends on alloy emissivity, powder condition, and shield gas flow. A loop tuned on Ti-6Al-4V will not transfer directly to Inconel or to aluminum. Each alloy needs its own signal-to-temperature model, and that model drifts as the laser window clouds.

• 1
  PowerFastest lever, but changes penetration depth
• 2
  Scan speedChanges both energy density and build time
• 3
  Spot sizeWidens or narrows the melt pool without touching power
• 4
  Loop rateSub-millisecond for intra-layer, seconds for layer-to-layer

Start from the function, not the process. Mark which faces carry a seal, which carry a bearing, and which only carry load. Faces that need a seal or a bearing fit should be left with stock and finished on a CNC. That single decision removes most of the tolerance risk from a metal AM build, because as-built powder bed surfaces sit around Ra 8–12 μm and cannot hold a ±0.005 mm fit.

Orientation drives both support volume and residual stress. Put the largest flat face down when it does not need a fine finish, and rotate the part so that thin walls run vertical rather than horizontal. A horizontal thin wall needs support underneath, and support removal on a thin wall bends it. If the design allows, add a 0.5 mm machining allowance on critical faces and a small chamfer where a thin wall meets a thick base.

Build simulation is worth the hour it takes. Thermal simulation predicts the hot spots and the distortion direction, which tells you where to add sacrificial ribs or where to bias the model. After the build, stress relief before wire EDM or bandsaw cut-off prevents the part from moving once the base plate is released. Then machine the datums first, and measure from those datums only.

• 1
  Leave stock0.3–0.5 mm on sealing and bearing faces
• 2
  OrientVertical thin walls, minimal support contact on functional faces
• 3
  SimulatePredict hot spots and distortion before the build
• 4
  RelaxStress relief before cut-off, then machine datums

Most metal AM parts that reach a production line are hybrids. The AM step creates the internal channels, the organic ribs, or the topology that a cutter cannot reach. The CNC step creates the interfaces. On a typical aluminum or stainless housing, that split is roughly 70 % printed geometry and 30 % machined geometry by feature count, but nearly all of the critical dimensions come from the machined side.

That split changes the cost model. Printing time scales with volume, so hollowing a part saves machine hours. Machining time scales with the number and depth of features, so keeping machined faces flat and reachable saves more than shrinking the part. We see quotes swing by 30–40 % on the same geometry depending on orientation and stock allowance, before any material change.

For low-volume work, from one prototype to a few hundred parts, this route often beats tooling. For runs above a few thousand, die casting or injection molding usually wins once the tool is amortized. The crossover is not fixed; it moves with part size, feature count, and how much of the geometry has to be machined anyway.