Additive manufacturing

Fourteen capabilities across six plugins cover metal and polymer AM: the process chain, the design-for-additive-manufacturing checks, and the output writers.

Which process are you on?

Process	What souxmar does	Depth
LPBF (metal powder bed)	Build mesh → thermal history → melt pool + porosity risk → distortion → residual stress → DfAM → time/cost → report	Full chain. Every default is a 316L-on-a-400 W-machine number.
DED / WAAM	Time, energy, mass, cost with process: ded_waam; distortion and DfAM unchanged; the metal thermal chain reused with the arc read as a moving heat source	Partial — no arc physics, no bead geometry
FFF (polymer extrusion)	Interlayer thermal cycling → bond strength / Z-strength → DfAM → time/cost → G-code + CLI + report	Full chain, aimed at PEKK and PA12-CF
SLS (polymer powder bed)	Time, energy, mass, cost with process: sls	Partial — the polymer bonding chain is indicative only

Not covered: binder jetting, material jetting, vat photopolymerisation, sheet lamination, cold spray, hybrid additive-subtractive.

The capabilities

Capability id	Kind	Gives you
mesher.am.layered	mesher	Layer-aligned Hex8 build mesh; cell tag = layer index
reader.lattice	reader	Parametric strut lattice (cubic/bcc/fcc/octet/diamond) as Edge2 beams
solver.am.thermal.lpbf	solver	Nodal temperature (°C), one step per layer
postproc.am.melt_pool	postproc	Per-cell [depth_m, normalised_enthalpy, porosity_risk]
solver.am.distortion.inherent_strain	solver	Nodal distortion_displacement (m), one step per layer
postproc.am.residual_stress	postproc	Per-cell [sigma_vm_Pa, sigma_vm_over_yield, layer_index]
solver.am.polymer.fff	solver	Nodal interface_temperature (°C), one step per layer
postproc.am.bond_strength	postproc	Per-cell [degree_of_healing, z_strength_fraction, seconds_above_tg]
solver.am.overhang	solver	Per-cell [min_downskin_tilt_deg, support_needed, downskin_area_m2]
solver.am.printability	solver	Per-cell [printability_score, limiting_factor_code, wall_thickness_proxy_m]
solver.am.buildtime	solver	Per-cell [layer_time_s, cumulative_time_s, layer_area_m2]
writer.am.gcode	writer	FFF G-code from a planar slice
writer.am.cli	writer	Common Layer Interface ASCII
writer.am.report	writer	Markdown build report / traveller sheet

Component order is part of the contract — a viewer colouring by component 2 of melt_pool is looking at porosity risk, and that will not change under you.

Three of these are solver.* while doing no solving: overhang, printability and buildtime analyse a mesh and produce a field, and the postproc.* dispatch path requires an upstream field. See ADR-0044 — "solver" here is a vtable shape, not a claim.

Run the metal chain

souxmar run examples/am-lpbf-bracket/pipeline.yaml \
    --plugin-path build/dev/examples/plugins

The pipeline shape:

version: 1
stages:
  - id: build
    plugin: mesher.am.layered
    input:
      target_size: 0.002        # voxel edge AND simulation layer, metres

  - id: thermal
    plugin: solver.am.thermal.lpbf
    input:
      mesh: { from: build }
      laser_power: 195.0
      scan_speed: 0.8
      hatch_spacing: 1.1e-4
      layer_height: 0.002       # must match the mesh
      process_layer_height: 3.0e-5   # the machine's powder layer
      interlayer_time: 12.5
      baseplate_layers: 1

  - id: meltpool
    plugin: postproc.am.melt_pool
    input:
      mesh:  { from: build }
      field: { from: thermal }  # postproc.* REQUIRES a field
      laser_power: 195.0        # repeat the process block — a postproc
      scan_speed: 0.8           # cannot see the stage before it
      hatch_spacing: 1.1e-4
      layer_height: 0.002
      process_layer_height: 3.0e-5

Three things trip people up:

1. Two layer heights. layer_height is the simulation layer — one field step, and it must match the mesh. process_layer_height is the machine's powder layer and sets the energy density.
2. postproc.* needs field: {from: …}. A missing field is a dispatch error, not a NULL.
3. Process inputs are repeated per stage. No plugin reads a material file; every property is an explicit input with a default.

DfAM triage on real geometry

Point the three checks at a real surface mesh, not at a voxel box — mesher.am.layered meshes a bounding box, so its axis-aligned faces make the overhang answer degenerate.

  - id: part
    plugin: reader.stl
    input: { path: bracket.stl }

  - id: overhang
    plugin: solver.am.overhang
    input:
      mesh: { from: part }
      build_direction: [0.0, 0.0, 1.0]
      overhang_threshold_deg: 45.0

  - id: printability
    plugin: solver.am.printability
    input:
      mesh: { from: part }
      min_wall_thickness: 6.0e-4
      machine_build_volume: [0.25, 0.25, 0.30]
      corrosion_allowance: 0.001   # subsea part: extra wall for 25 years

  - id: cost
    plugin: solver.am.buildtime
    input:
      mesh: { from: part }
      process: lpbf
      material_cost_per_kg: 75.0
      machine_rate_per_hour: 45.0

A face is downward iff n·b < 0, and its tilt from the build plate is acos(|n·b|); support_needed is 1 when any downward face tilts below the threshold. The 45° default is the powder-bed rule of thumb, not a material property.

Slice to G-code

  - id: gcode
    plugin: writer.am.gcode
    input:
      mesh: { from: part }
      path: fairing.gcode
      layer_height: 3.0e-4
      road_width: 6.0e-4
      nozzle_temperature: 400.0
      bed_temperature: 140.0
      print_speed: 0.04
      infill_spacing: 0.003

Single perimeter, scanline infill, no offset compensation, no support generation, no bridging logic, no collision check. It is a geometrically faithful but mechanically naive toolpath — read it before you feed it to a machine.

Calibrate before you quote

Constant	Default	Fit it against
strain_calibration	0.30	Tip deflection of a printed twin-cantilever bridge coupon, cut free. Re-fit whenever layer_height changes.
absorptivity	0.35	One measured single-track melt-pool cross-section.
reptation_time_reference, activation_energy	2.0 s, 8.0e4 J/mol	Z-oriented tensile coupons at a small layer_time matrix

Until then the numbers are relative. The step-by-step procedures are in docs/MANUFACTURING.md.

From chat

propose_am_setup        propose a runnable pipeline from process + material + machine
check_printability      run the DfAM check and summarise blockers
set_build_orientation   score candidate orientations, stage the winner
estimate_build_cost     time / energy / mass / cost

set_build_orientation prompts once per session because it changes what every downstream AM answer computes; the other three are read-only and silent.

Next

• Marine and subsea
• docs/MANUFACTURING.md — every input key, unit, default and output component, plus the fidelity table.
• RFC-0012 — where the closed-form models break and what replaces them.