Additive Manufacturing cover
Interactive book · Semester 4-8 (elective)

Additive Manufacturing

Slice a part and watch the staircase appear, cure a resin layer, find the energy window for metal powder, nest an SLS build and cost a printed part against injection moulding

20 chapters, 20 interactives, few words. Every chapter is mapped to the Anna University, VTU, SPPU, Mumbai University and KTU units it covers: what additive manufacturing is and when it beats machining or moulding, the seven ISO/ASTM 52900 process families, the process chain, STL files, slicing and the staircase effect, part orientation and supports, tool paths and G-code, FDM, stereolithography, DLP and CLIP, SLS and multi-jet fusion, metal powder bed fusion, directed energy deposition and WAAM, binder jetting, material jetting and sheet lamination, design for AM and lattices, materials and post-processing, process selection and cost, printer hardware and firmware, and rapid tooling and applications.

  • AUCME339Additive Manufacturing
  • VTUBME714AAdditive Manufacturing
  • SPPU402045CAdditive Manufacturing
  • MUMEC404CAD/CAM (Module 5: Additive Manufacturing)
  • KTUMET468Additive Manufacturing

By Skill-Lync · Updated 17 Nov 2026 · 20 chapters · 20 interactives · free, no sign-up

Chapter 1

What additive manufacturing is: from rapid prototyping to production

Which syllabus covers thisAU CME339 Unit IVTU BME714A Module 1SPPU 402045C Unit 1MU MEC404 Module 5KTU MET468 Module 1

What is additive manufacturing, and how is it different from machining and moulding?

ISO/ASTM 52900 defines additive manufacturing (AM) as joining materials to make parts from 3D model data, usually layer upon layer. Machining is subtractive: it starts from a block and cuts material away. Moulding, casting and forging are formative: they push material into a shape set by a tool. AM needs no part-specific tool. A slicer cuts the CAD model into thin layers, typically 0.02 to 0.4 mm, and the machine builds each layer on the last. That is why the same machine can make a bracket in the morning and a hearing-aid shell in the afternoon.

What do rapid prototyping, rapid tooling and rapid manufacturing mean?

They name what the printed part is used for. Rapid prototyping (RP), the first use, makes concept models and form, fit and functional test parts in days instead of weeks, so a design can be checked before tooling is ordered. Rapid tooling (RT) prints the tool rather than the product: mould inserts, casting patterns, sand moulds, jigs and fixtures. Rapid manufacturing (RM), or direct digital manufacturing, prints end-use parts, such as medical implants, aircraft brackets and short runs of spares. In product development RP shortens the design loop; RT and RM shorten the path to production.

Why use AM, and where does it fall short?

Benefits: shape complexity costs almost nothing extra, so internal channels, lattices and organic shapes are possible; no tooling, so one-offs and customised parts are cheap; several parts can be consolidated into one; little material is wasted, especially compared with machining a titanium billet; and spares can be made on demand from a file. Limits: builds are slow, typically hours to days; build volumes are small; surfaces show layer steps and often need finishing; properties can differ between the layer plane and the build direction; and at high volumes the cost per part stays flat while moulding gets cheaper.

When is AM cheaper than CNC machining or injection moulding?

Each route has a fixed cost spread over the batch plus a cost per part: cost per part = F/n + c. AM has a small F (build preparation) but a high c. CNC needs programming and fixtures. Moulding needs a mould costing hundreds of thousands of rupees but makes each part for a few rupees. So AM wins small batches, CNC the middle and moulding the large runs. Each crossover sits where two routes cost the same: n = (F₂ − F₁)/(c₁ − c₂). With the illustrative numbers in the figure, AM is cheapest up to 81 parts, CNC from 82 to 2,088, and moulding beyond.

Cost per part against quantity: additive manufacturing, CNC machining and injection moulding, and the break-even batch sizes: interactive loads as you scroll

How did AM develop, and where does it fit in Industry 4.0?

Two developments made it possible: solid CAD models that describe a closed volume, and the idea of building in thin layers. Charles Hull filed his stereolithography patent in August 1984 and it was granted in March 1986 (US 4,575,330). Carl Deckard's selective sintering patent, filed in 1986 for the University of Texas, was granted in 1989. Scott Crump's patent behind FDM, filed in 1989 for Stratasys, was granted in 1992. In Industry 4.0, AM is the step where a digital file becomes a part. That supports digital spare-part inventories, local production, mass customisation and sensor-monitored builds.

Predict, then check

A mould costs ₹500,000 and moulds each part for ₹25. AM costs ₹1,500 to set up and ₹600 a part. At 2,000 parts, which is cheaper per part?

Worked examples

Break-even between AM and injection moulding · AM: set-up ₹1,500 and ₹600 a part. Moulding: mould ₹500,000 and ₹25 a part (illustrative). Find the break-even quantity and the cost per part at 500 parts
  1. At break-even the totals match: 1,500 + 600 n = 500,000 + 25 n
  2. n = (500,000 − 1,500)/(600 − 25) = 498,500/575 = 867.0
  3. Check totals: at 866 parts AM is still cheaper; at 867 the mould is cheaper
  4. At 500 parts: AM 1,500/500 + 600 = ₹603; moulding 500,000/500 + 25 = ₹1,025

Answer: Break-even at about 867 parts; at 500 parts AM costs ₹603 a part against ₹1,025 for moulding

Three routes and the cheapest at each quantity · As above, plus CNC machining: programming and fixtures ₹30,000 and ₹250 a part (illustrative). Which route is cheapest at 500 parts, and where does each take over?
  1. AM vs CNC: n = (30,000 − 1,500)/(600 − 250) = 28,500/350 = 81.43
  2. CNC vs moulding: n = (500,000 − 30,000)/(250 − 25) = 470,000/225 = 2088.9
  3. At 500: AM ₹603, CNC 30,000/500 + 250 = ₹310, moulding ₹1,025
  4. AM vs moulding (867) lies inside CNC's range, so AM is never cheapest beyond 81

Answer: CNC is cheapest at 500 parts (₹310 each). AM is cheapest up to 81 parts, CNC from 82 to 2,088, and moulding from 2,089

Variety instead of volume · 1,000 parts in total, but 50 different designs of 20 each (for example, sizes fitted to users). Each design needs its own mould at ₹100,000 (illustrative small aluminium tool); AM as above
  1. AM: one set-up for the batch, 1,500 + 600 × 1,000 = ₹601,500
  2. Moulding: 50 × 100,000 + 25 × 1,000 = ₹5,025,000
  3. Ratio: 5,025,000/601,500 = 8.4
  4. With one design of 1,000 the same mould costs 100,000 + 25,000 = ₹125,000, and moulding wins

Answer: AM is about 8.4 times cheaper here: tooling cost scales with the number of designs, not the number of parts

Chapter 2

The seven process families of ISO/ASTM 52900

Which syllabus covers thisAU CME339 Unit IVTU BME714A Module 1SPPU 402045C Unit 1MU MEC404 Module 5KTU MET468 Module 1

What are the seven process families?

ISO/ASTM 52900:2021 groups every AM process by how a layer is made. Vat photopolymerisation (VPP): light cures liquid resin in a vat. Material extrusion (MEX): material is pushed through a nozzle. Powder bed fusion (PBF): heat fuses regions of a powder bed. Binder jetting (BJT): a liquid binder joins powder. Material jetting (MJT): droplets of build material are deposited. Sheet lamination (SHL): sheets are bonded together. Directed energy deposition (DED): focused heat melts material as it is deposited. Trade names such as SLA, FDM, SLS, PolyJet and LENS each sit in one family.

The seven ISO/ASTM 52900 process families: feedstock, energy source, how a layer forms and which materials each can print: interactive loads as you scroll

How else can AM processes be classified?

By the starting material's form: liquid (resin vats, jetted droplets), solid (filament, wire, sheet, pellets) and powder (powder beds, blown powder). By material class: polymers, metals, ceramics and composites. By application: prototypes, tooling or end-use parts. By how many steps: single-step processes give the final shape and material in one go (laser powder bed fusion of metal), while multi-step ones need a second stage, such as sintering a binder-jetted green part. By scale: micro and nano AM, such as micro-stereolithography and two-photon writing, makes features of microns or less.

What is the difference between direct and indirect AM?

Direct AM prints the part itself in its final material: a titanium implant by electron beam melting, or a nylon duct by SLS. Indirect AM prints something used to make the part: a wax or resin pattern for investment casting, a binder-jetted sand mould for a cast iron housing, or a mould insert for injection moulding. Indirect routes give access to materials and sizes the printer cannot make directly, and to the established properties of cast or moulded metal. They add process steps, but are often the cheapest way to a metal part in a small batch.

Which families print metal, and what is a hybrid machine?

Metal parts come mainly from powder bed fusion (laser or electron beam), directed energy deposition (laser, arc or electron beam, with powder or wire) and binder jetting followed by sintering. Bound-metal filament extrusion, sheet lamination by ultrasonic welding of foils, and nanoparticle jetting also exist. A hybrid machine puts AM and machining in one setup, typically a DED head on a five-axis CNC machine: deposit near-net material, then mill it to tolerance without moving the part. Hybrid systems are used to repair worn parts and to add features to existing ones.

Predict, then check

Which family needs no support structures, even under a flat overhanging ceiling?

Worked examples

Place the trade names in their families · SLA, DLP, FDM, SLS, EBM, MJF, binder-jet 3DP, PolyJet, LOM, LENS, WAAM
  1. Liquid resin cured by light: SLA, DLP → VPP
  2. Material pushed through a nozzle: FDM → MEX
  3. Heat fuses a powder bed: SLS (laser), EBM (electron beam), MJF (fusing agent + infrared lamp) → PBF
  4. Binder printed onto powder: 3DP → BJT; droplets of build material: PolyJet → MJT
  5. Sheets bonded and cut: LOM → SHL; material melted as it is fed: LENS (powder + laser), WAAM (wire + arc) → DED

Answer: VPP: SLA, DLP; MEX: FDM; PBF: SLS, EBM, MJF; BJT: 3DP; MJT: PolyJet; SHL: LOM; DED: LENS, WAAM

Layer count across families · A part 30 mm tall, built at a typical layer thickness for each family: VPP 0.05 mm, MEX 0.2 mm, metal PBF 0.04 mm, MJT 0.016 mm, DED 1.0 mm
  1. VPP: 30/0.05 = 600 layers
  2. MEX: 30/0.2 = 150 layers
  3. PBF: 30/0.04 = 750 layers
  4. MJT: 30/0.016 = 1,875 layers
  5. DED: 30/1.0 = 30 layers

Answer: From 30 (DED) to 1,875 (MJT) layers, a factor of 62.5: fine layers give detail and smooth surfaces; coarse layers give deposition rate

Choose a family for three jobs · (a) An 80 mm titanium bracket with internal cooling channels; (b) rebuilding a worn 20 mm band on a 600 mm steel shaft; (c) a multi-colour concept model with soft grips
  1. (a) Fine metal detail and internal channels: laser or electron beam powder bed fusion
  2. (b) Adding metal to an existing large part: DED, laser cladding or a hybrid DED + CNC machine, then machine to size
  3. (c) Several materials and colours in one build: material jetting (PolyJet-type), with soluble or gel support

Answer: (a) PBF, (b) DED, (c) MJT

Chapter 3

The AM process chain: eight steps from CAD to part, and reverse engineering

Which syllabus covers thisAU CME339 Unit IVTU BME714A Module 1SPPU 402045C Unit 1MU not in syllabusKTU MET468 Modules 1, 2

What are the eight steps of the generic AM process chain?

1. Concept and CAD: a solid model, or a scan of an existing object. 2. Convert to STL (or AMF or 3MF): a closed triangle mesh. 3. Transfer and manipulate the file: check and repair it, then scale, orient, place and nest the parts. 4. Set up the machine: material, layer thickness, power and speed, supports and slicing. 5. Build: mostly automatic, with monitoring. 6. Remove the parts: cool down, take them off the plate, recover the powder. 7. Post-process: remove supports, clean, cure or heat-treat, machine and finish. 8. Use: inspect the part and put it into service.

What decides how long a build takes?

Two parts: the time to scan or deposit the volume, roughly volume divided by the machine's rate, and a fixed time for every layer (spreading powder, recoating resin, moving the platform). The layer count is height divided by layer thickness, so a tall part built in 40 µm layers can spend more time recoating than melting. Laying a part flat keeps the volume and scan time the same but cuts the layer count. Nesting several parts in one build shares the recoats among them. Process planning trades this time against supports, surface finish and strength.

Build time from layers: why layer count and recoating dominate, and why a part printed flat finishes sooner than tall: interactive loads as you scroll

What is reverse engineering, and how does a scan become a printable model?

Reverse engineering creates a CAD model from a physical object that has no usable drawing: a worn part, a hand-made pattern or a patient's anatomy. Digitising captures the surface as a point cloud. Methods: a touch probe on a CMM (accurate, slow), laser line or structured-light scanners (fast, non-contact), and CT scanning (inside as well as outside). Software then cleans and aligns the scans, meshes the points into triangles, fills holes and decimates the mesh. For editing, it fits NURBS surfaces or rebuilds features. The mesh can be printed directly; the surface model can be edited like any CAD part.

What goes into process planning and machine setup?

Choose the process and material for the part's job; check the design against that process's rules (minimum wall, hole size, overhangs); orient and place parts on the plate; generate supports; choose layer thickness and process parameters (power, speed, hatch spacing and temperatures for the material); slice and generate tool paths; and estimate time and cost. The result is a build file for one machine. Planning also covers post-processing steps and inspection, because a part printed in the wrong orientation can need hours of support removal.

Predict, then check

On a powder-bed machine you double the layer thickness for a tall part. What happens to the build time?

Worked examples

Upright or flat · Square prism 100 mm long, cross-section 20 × 20 mm (400 mm²); layers 0.04 mm; scan rate 3 mm³/s; recoat 8 s per layer
  1. Volume = 100 × 400 = 40,000 mm³; scan time = 40,000/3 = 13,333 s = 3 h 42 min in both orientations
  2. Upright: 100/0.04 = 2,500 layers × 8 s = 20,000 s; total 33,333 s = 9 h 16 min
  3. Flat (20 mm tall): 20/0.04 = 500 layers × 8 s = 4,000 s; total 17,333 s = 4 h 49 min
  4. Recoating is 60% of the upright build

Answer: Flat takes 4 h 49 min against 9 h 16 min upright: 48% less, from layers alone

Nesting five parts in one build · Five of the same prisms, all laid flat side by side in one build
  1. Scan: 5 × 13,333 = 66,667 s
  2. Recoat: still 500 layers × 8 s = 4,000 s (the bed height is set by the tallest part)
  3. Total 70,667 s = 19 h 38 min; per part 14,133 s = 3 h 56 min

Answer: 3 h 56 min a part against 4 h 49 min alone: nesting spreads the recoat time

Size of a scanned mesh · A 200 × 150 mm face scanned on a square grid at 0.1 mm point spacing, meshed into two triangles per grid cell and saved as binary STL; then decimated to 0.5 mm spacing
  1. Points at 0.1 mm: 2,001 × 1,501 = 3,003,501; triangles 2 × 2,000 × 1,500 = 6,000,000
  2. Binary STL = 84 + 50 × 6,000,000 = 300,000,084 bytes ≈ 300 MB
  3. At 0.5 mm: 401 × 301 = 120,701 points, 240,000 triangles, 12,000,084 bytes ≈ 12 MB

Answer: About 300 MB raw and 12 MB after decimation: point spacing ×5 cuts the mesh 25 times

Chapter 4

STL files: tessellation, errors, repair, AMF and 3MF

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Module 4SPPU not in syllabusMU not in syllabusKTU MET468 Module 4

What is inside an STL file?

STL, developed in 1987 for 3D Systems' first commercial printers, describes only the surface, as a list of triangles (facets). Each facet stores an outward unit normal and three vertices, listed anticlockwise when seen from outside: the right-hand rule. It has no units, colour, material or topology, and shared vertices are repeated in every facet that uses them. ASCII STL is readable text (solid, facet normal, outer loop, vertex, endloop, endfacet, endsolid). Binary STL has an 80-byte header, a 4-byte facet count and 50 bytes per facet, so its size is 84 + 50F bytes. The original rules also required all coordinates to be positive.

How does the chord tolerance set the number of facets?

A curved surface is replaced by flat triangles whose edges are chords. The chord tolerance limits how far a chord can sit from the true curve: for a circle of radius r divided into N segments, the gap is r(1 − cos(π/N)). Halving the tolerance raises N by about √2, so file size grows quickly for little gain. Too coarse a mesh shows flats on holes and loses volume, because the facets lie inside the curve. Too fine a mesh gives huge files and slow slicing. Exporters also offer an angle tolerance to refine small, tight radii.

STL tessellation of a cylinder: facets for a chord tolerance, file size, volume error and the checks that catch gaps and flipped normals: interactive loads as you scroll

What are the common STL errors, and what makes a mesh valid?

A valid tessellated model is watertight and manifold: every edge is shared by exactly two facets, the two facets run along that edge in opposite directions so all normals point outwards, facets meet only at shared edges and vertices (vertex-to-vertex rule), and V − E + F = 2 for a solid without holes. Common errors: gaps or cracks, where facets are missing or edges do not meet; flipped normals; non-manifold edges shared by three or more facets; overlapping or intersecting facets; degenerate facets with zero area; and stray shells. A slicer fed an open mesh cannot tell inside from outside.

How are STL files repaired and manipulated?

Repair software merges vertices that should coincide, stitches edges across small cracks, fills holes with new facets, unifies the facet winding so all normals point out, deletes degenerate and duplicate facets, and removes self-intersections by re-meshing. Then the model is manipulated for the build: scaled (including shrinkage compensation), rotated into the chosen orientation, cut into pieces that fit the machine, hollowed, labelled, given supports and nested with other parts. Each step must keep the mesh watertight. Repairing at the CAD end, by re-exporting with sensible tolerances, is often quicker than patching a broken mesh.

What other file formats are used, and why were AMF and 3MF proposed?

CAD exchange formats (STEP and IGES) carry exact surfaces but are not read directly by most printers. Other mesh formats include OBJ, PLY and VRML, which can carry colour. Slice formats such as CLI and SLC store the contours of each layer. AMF (ISO/ASTM 52915:2016) is XML-based and adds units, colours, multiple materials, curved triangles and lattices. 3MF, first released in 2015 by the 3MF Consortium, is a ZIP package of XML files. It stores the mesh with units, colours, materials and print settings in one file, so they need not be sent separately.

Predict, then check

You halve the chord tolerance when exporting a cylinder. The facet count roughly:

Worked examples

Facets and file size for a cylinder · Cylinder r = 10 mm, h = 20 mm, chord tolerance 0.01 mm
  1. N = ⌈π / cos⁻¹(1 − 0.01/10)⌉ = ⌈π/0.04473⌉ = ⌈70.24⌉ = 71
  2. Facets F = 2N (sides) + 2(N − 2) (caps) = 280; V = 2N = 142; E = 6N − 6 = 420
  3. Euler check: 142 − 420 + 280 = 2
  4. Binary size = 84 + 50 × 280 = 14,084 bytes
  5. Volume: polygon area/circle area = N sin(2π/N)/(2π), so 0.130% is lost (6,275.0 vs 6,283.2 mm³)

Answer: 71 segments, 280 facets, 14,084 bytes, volume 0.13% low

A coarse export · Same cylinder at chord tolerance 0.1 mm
  1. N = ⌈π / cos⁻¹(0.99)⌉ = ⌈22.2⌉ = 23
  2. F = 4 × 23 − 4 = 88; binary size 84 + 50 × 88 = 4,484 bytes
  3. Volume lost: 1 − 23 sin(2π/23)/(2π) = 1.24% (6,205.3 mm³)
  4. Sag of each flat: 10(1 − cos(π/23)) = 0.0931 mm

Answer: 23 flats you can see on a 20 mm bore: 88 facets (4,484 bytes) and 1.2% less volume

Checking a cube · A cube exported as STL; then the same file with one facet missing
  1. Each face is two triangles: F = 12; unique vertices V = 8; edges E = 12 cube edges + 6 face diagonals = 18
  2. V − E + F = 8 − 18 + 12 = 2: a closed solid; binary size 84 + 50 × 12 = 684 bytes
  3. The file stores 12 × 3 = 36 vertex records for 8 vertices: STL repeats shared vertices
  4. Missing facet: F = 11, its 3 edges now belong to only one facet, and V − E + F = 1

Answer: The intact cube passes (every edge in exactly 2 facets, V − E + F = 2); the damaged one fails with 3 open edges

Chapter 5

Slicing, layer thickness and the staircase effect

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Module 1SPPU 402045C Unit 5MU not in syllabusKTU MET468 Module 2

What does slicing do?

The slicer intersects the mesh with a stack of horizontal planes, one per layer. Each plane cuts some facets along line segments, and the slicer joins the segments into closed loops: outer contours and inner contours for holes. It then decides which side is material, offsets the loops for the bead or beam width, and fills the inside with the tool path. Gaps in the mesh give open loops, which is why the file is repaired first. Direct slicing, which cuts the exact CAD surfaces instead of the STL, avoids the facet error but needs the CAD kernel.

What is the staircase effect, and how big is the cusp?

Each layer has vertical edges, so a sloping or curved surface becomes a flight of steps. The cusp height, the largest gap between the steps and the true surface, is h = t·|cos θ|. Here t is the layer thickness and θ the angle between the surface normal and the build direction, the same as the surface's tilt from horizontal. Vertical walls (θ = 90°) have no steps. Near-flat slopes (θ near 0) have cusps approaching a full layer. The profile is a sawtooth, so the roughness Ra is about h/4: a 0.2 mm layer on a 60° slope gives h = 0.1 mm, Ra ≈ 25 µm.

The staircase effect: cusp height and roughness against layer thickness and surface angle, and adaptive slicing: interactive loads as you scroll

How do you choose the layer thickness?

Thinner layers mean smaller steps, finer vertical detail and smoother slopes, but more layers and a longer build. Build time is roughly proportional to the layer count, and so to 1/t. The range is set by the process (typical: FDM 0.1-0.3 mm for a 0.4 mm nozzle, SLA 0.025-0.1 mm, metal powder bed fusion 0.02-0.06 mm) and by the feedstock: a powder layer must be thicker than most of the powder particles. A common compromise: a thick layer for internal and vertical regions, and thin layers only where shallow slopes or fine features need them.

What is adaptive slicing?

Uniform slicing uses one thickness for the whole part, so it must be thin enough for the shallowest slope, everywhere. Adaptive slicing varies the thickness: t = h_max/|cos θ|, limited to the machine's thinnest and thickest layer. On steep walls layers are thick; near horizontal faces they thin down. For a 40 mm sphere with a 0.05 mm cusp limit, uniform layers need 800 slices, while adaptive layers between 0.05 and 0.3 mm need about 410. Cura offers this as adaptive layers. Region-based slicing goes further: thick layers inside, thin ones at the skin.

Predict, then check

Same layer thickness everywhere. Which surface shows the biggest steps?

Worked examples

Cusp and roughness · Layer 0.2 mm; surfaces tilted 60° and 30° from horizontal, and a vertical wall
  1. 60°: h = 0.2 × cos 60° = 0.100 mm; Ra ≈ h/4 = 0.025 mm = 25 µm
  2. 30°: h = 0.2 × cos 30° = 0.1732 mm; Ra ≈ 43 µm
  3. 90°: cos 90° = 0, no steps (only bead texture)

Answer: Cusps of 0.10 mm and 0.173 mm; the 30° surface is 1.73 times rougher

Layer thickness for a cusp limit · Allowed cusp 0.05 mm; machine layers 0.05 to 0.3 mm (typical FDM); surfaces at 30°, 75° and 85°
  1. 30°: t = 0.05/cos 30° = 0.0577 mm
  2. 75°: t = 0.05/cos 75° = 0.1932 mm
  3. 85°: 0.05/cos 85° = 0.574 mm, capped at the machine's 0.3 mm

Answer: 0.058, 0.193 and 0.3 mm: the steep wall takes a layer 5.2 times thicker than the 30° slope

Uniform against adaptive slicing of a sphere · Sphere 40 mm in diameter; layers 0.05 to 0.3 mm; cusp limit 0.05 mm
  1. Uniform at 0.2 mm: 40/0.2 = 200 layers, but cusps reach 0.2 mm near the poles
  2. Uniform meeting 0.05 mm everywhere: t = 0.05 mm (needed at the poles, cos θ = 1) → 800 layers
  3. Adaptive: on a sphere cos θ = |z|/R, so t = 0.05 × 20/|z|, between 0.05 and 0.3 mm
  4. Stepping layer by layer from the bottom pole gives 413 layers (integral of dz/t: 411)

Answer: 413 adaptive layers against 800 uniform ones for the same 0.05 mm cusp: 48% fewer

Chapter 6

Part orientation and support structures

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Module 4SPPU 402045C Unit 4MU not in syllabusKTU MET468 Module 2

Why does part orientation matter so much?

Orientation fixes the build height (and so the layer count and time), which faces overhang and need support, where the staircase appears, and which direction is weakest. Turning the same L-bracket changes all of these together. Lying flat it needs no support and is 50 mm tall. Upside down it needs 1,600 mm² of support under one leg. Stood on its outer corner (rotated 135°) it needs none, but every face is stair-stepped and the build is 78 mm tall. There is rarely one best answer. Planners weigh time, support volume, surface finish on the faces that matter, accuracy of critical features and strength.

Part orientation: rotate an L-bracket and watch build height, overhangs past the self-supporting angle and support area change: interactive loads as you scroll

Why are support structures needed?

Supports hold up overhangs and bridges that would otherwise print onto air, anchor the part to the build plate, and stop thin regions from moving. In metal powder bed fusion they also conduct heat away from the melt and resist the residual stresses that curl the part; even down-facing surfaces surrounded by powder need them. In SLA they keep fresh layers in place against peel forces. Polymer SLS and binder jetting need none, because the surrounding powder carries every layer. Supports cost material, build time and post-processing, and they leave marks on the surfaces they touch.

What is the self-supporting angle?

A down-facing surface can be built without support if it is close enough to vertical. Each new layer then steps out only a little from the one below. The usual rule of thumb is 45° from vertical for FDM and metal powder bed fusion (typical; it depends on material, layer and cooling). In FDM, each layer steps out by t·tan α; at 45° and 0.2 mm that is 0.2 mm, less than half of a 0.45 mm bead, so the bead is mostly supported. Short horizontal bridges between two supports also print, by stretching the bead. Chamfers and teardrop holes replace flat ceilings and round tops.

What kinds of supports are there, and how are they removed?

Block or lattice supports fill the volume under an overhang with a thin grid. Line, point and cone supports touch the part only at small spots. Tree supports branch up from a few bases, reaching overhangs with less material. Soluble supports, printed in a second material, dissolve in water or a chemical bath and suit internal cavities. Contact is weakened deliberately with small gaps or perforated teeth. Removal is manual (pliers, knives), by dissolving, or for metal by wire EDM, sawing and machining, followed by finishing the scars.

How does orientation affect strength and accuracy?

Layers bond less strongly than the material within a layer, especially in FDM, where parts are typically weakest across the layers (in the build direction). So orient a part so the main load runs along the layers, not across them: a hook printed lying down is stronger than one printed upright. Holes come out roundest with their axis vertical. Accuracy is usually best in the plane of the layers, and flat faces parallel to the plate are the smoothest. Critical features may have to be machined after the build whatever the orientation.

Predict, then check

The L-bracket printed with its long leg on the plate needs no support. What happens if you turn it upside down (180°)?

Worked examples

Three orientations of an L-bracket · Section: long leg 60 × 10 mm, upright leg 10 × 50 mm, 10 mm 45° inner chamfer; self-supporting angle 45° from vertical; layers 0.2 mm
  1. 0° (long leg on the plate): height 50 mm, 250 layers, no down-facing edges off the plate: support 0
  2. 90° (short leg on the plate): height 60 mm, 300 layers, support 0
  3. 180°: height 50 mm, 250 layers; the long leg's 40 mm underside is a ceiling 40 mm up: 40 × 40 = 1,600 mm²
  4. 135° (standing on a corner): height 77.8 mm, 389 layers; all faces at exactly 45°, so no support, but every face is stair-stepped

Answer: 0° is best here: lowest (50 mm), no support, and only the chamfer shows steps

A tighter self-supporting angle · Same bracket at 180°, but the material only self-supports to 40° from vertical
  1. The chamfer is 45° from vertical, now beyond the limit, so it needs support
  2. It spans 10 mm across, from 40 mm down to 30 mm above the plate (just beside the upright leg)
  3. Support under it, down to the plate: 10 × (40 + 30)/2 = 350 mm²
  4. Total: 1,600 + 350 = 1,950 mm²

Answer: 1,950 mm² of support in the section, 22% more

Why 45° is the usual FDM rule · FDM, layer 0.2 mm, line width 0.45 mm. Compare overhangs of 45° and 60° from vertical
  1. Each layer steps out by t·tan α: 45° → 0.2 × 1 = 0.20 mm; 60° → 0.2 × 1.732 = 0.346 mm
  2. Half the bead width is 0.225 mm: at 45° the bead is more than half supported; at 60° most of it is over air
  3. The limit where the step equals half a bead: α = tan⁻¹(0.225/0.2) = 48.4°

Answer: About 48° from vertical for these settings: the 45° rule of thumb, with a little margin

Chapter 7

Tool paths, infill and G-code

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Module 2SPPU 402045C Unit 5MU not in syllabusKTU MET468 Module 2

How is each layer filled?

First the contours: one or more perimeters (shells) traced along the slice outline, offset inwards by half a line width, which give the surface finish and most of the stiffness. Then solid top and bottom skins, usually several layers of 100% raster. Then the infill inside, at a chosen density (typically 10-30% for prototypes). Patterns: lines or zigzag raster (fast, with alternating directions layer to layer), grid, triangles, honeycomb, gyroid (similar stiffness in all directions) and concentric. Laser machines use the same ideas: contour scans plus hatch vectors, often rotated every layer or split into stripes and islands.

How much filament does a line of bead use?

A bead of width w and height t is close to a stadium: a rectangle with rounded ends, area A = (w − t)·t + πt²/4. Moving at speed v, the nozzle needs a volumetric flow Q = A·v, and each millimetre of path uses E = A/(πD²/4) of filament of diameter D. That E is what the slicer writes into G-code. The hot end can melt only so much plastic a second (typically 10-15 mm³/s for a standard hot end), so it caps the speed at Q_max/A. Under-extrusion, gaps and weak layers follow if you go faster.

Extrusion maths: bead area, volumetric flow, filament feed per mm (the E value in G-code) and the hot-end speed limit: interactive loads as you scroll

What does G-code look like?

Each line is a command. G28 homes the axes; G90 and G91 select absolute or relative positions; M82 and M83 the same for the extruder. G0 is a travel move and G1 a printing move: G1 X50 Y20 E1.69 F3600 moves to (50, 20) at 3,600 mm/min (60 mm/s) while feeding 1.69 mm of filament. M104 and M109 set the nozzle temperature (M109 waits), M140 and M190 the bed, M106 the part fan. Firmware such as Marlin turns these lines into step pulses for each motor. The slicer, for example Cura, writes the file.

What else does path planning control?

Where each layer starts and stops (the seam, hidden in a corner or spread at random), retraction (pulling filament back before a travel move so it does not ooze strings), travel paths that avoid crossing the part, the order of perimeters and infill, speeds and cooling for small layers, and acceleration at corners. Plotting and path control in extrusion machines means keeping the flow matched to the speed: the nozzle slows at corners and starts, so the slicer must cut flow there or blobs and gaps appear. Bio-extrusion follows the same rules with pastes and gels.

What design rules apply to extrusion-based AM?

Typical guidelines for a 0.4 mm nozzle: walls at least two line widths (about 0.8-1.2 mm) thick; vertical pins at least about 2 mm across; overhangs within about 45° of vertical, or chamfered; bridges kept short; horizontal holes given a teardrop or flat top; holes printed slightly undersize, then drilled where fit matters; clearances of about 0.2-0.5 mm between mating parts printed together; text embossed rather than finely engraved. Lay the part so loads run along the layers, add fillets at the bed to resist warping, and allow for elephant's foot at the first layer.

Predict, then check

You double the print speed and keep the same bead. What happens to the flow Q and to E per mm?

Worked examples

The E value of a G1 move · Nozzle 0.4 mm, line width 0.45 mm, layer 0.2 mm, 1.75 mm filament, 60 mm/s; one straight line 50 mm long, relative extrusion (M83)
  1. Bead A = (0.45 − 0.2) × 0.2 + π × 0.2²/4 = 0.05 + 0.03142 = 0.08142 mm²
  2. Filament area = π × 1.75²/4 = 2.4053 mm²; E per mm = 0.08142/2.4053 = 0.03385
  3. E for 50 mm = 1.6924 mm; F = 60 × 60 = 3,600 mm/min
  4. Flow Q = 0.08142 × 60 = 4.885 mm³/s

Answer: G1 X50 Y0 E1.6924 F3600, at 4.88 mm³/s

Speed limit of the hot end · The same 12 mm³/s hot end (typical) with a 0.6 mm nozzle, 0.7 mm line, 0.3 mm layer and 2.85 mm filament
  1. Small bead (example 1): v_max = 12/0.08142 = 147 mm/s
  2. Large bead: A = (0.7 − 0.3) × 0.3 + π × 0.09/4 = 0.19069 mm²; v_max = 12/0.19069 = 62.9 mm/s
  3. E per mm with 2.85 mm filament: 0.19069/6.379 = 0.02989

Answer: The 2.3 times larger bead caps the speed at 63 mm/s instead of 147 mm/s, but deposits the same 12 mm³/s

Filament for a part · A part needs 12 cm³ of PLA (density 1.24 g/cm³, typical), printed with the bead and speed of example 1
  1. Filament length = 12,000/2.4053 = 4,989 mm ≈ 4.99 m
  2. Mass = 12 × 1.24 = 14.9 g
  3. Time spent extruding at 4.885 mm³/s = 12,000/4.885 = 2,457 s ≈ 41 min (travel and slow-downs extra)

Answer: 4.99 m of filament, 14.9 g, at least 41 minutes of extrusion

Chapter 8

Fused deposition modelling (material extrusion)

Which syllabus covers thisAU CME339 Unit IVVTU BME714A Module 2SPPU 402045C Unit 3MU MEC404 Module 5KTU MET468 Module 3

How does FDM build a part?

A drive gear pushes thermoplastic filament (1.75 or 2.85 mm) into a heated liquefier, and the molten plastic leaves a fine nozzle, 0.4 mm being typical, as a bead. The head traces each layer's outline and fill like a plotter while the bed or head steps by one layer, typically 0.1 to 0.3 mm. Each hot bead fuses to the one beside and below it, then cools and sets. Overhangs need supports, printed in the same material and broken off, or in a soluble material from a second nozzle. The process is also called fused filament fabrication (FFF).

Which materials does FDM use, and how hot does it run?

Amorphous thermoplastics, which soften over a range instead of melting sharply, print most easily. One printer maker's guide gives typical settings: PLA, the easiest, at a 210 to 215 °C nozzle and 60 °C bed, softening above about 60 °C in use; PETG at 230 to 240 °C and 85 to 90 °C, good to about 80 °C; ABS at 255 °C and an 80 to 110 °C bed, best in an enclosure because it warps as it cools. Nylon, polycarbonate, flexible TPU and fibre-filled grades need hotter, harder nozzles and dry filament.

Why are FDM parts weaker in Z, and what limits their quality?

Beads within a layer are continuous plastic, but each layer bonds to the cooler layer below only where the two briefly melted together, and small gaps remain between round beads. So a part pulled along the build direction is weakest, and loads should run along the layers. Cooling shrinkage curls corners off the bed (warping), the layers show as a staircase on slopes, and fine holes print undersize. Hotter nozzles, slower printing, enclosures and heated beds improve bonding and warping. Gains: cheap, office-safe and many materials. Costs: slow, visible layers, anisotropic strength.

What are plotting and path control in extrusion?

The head moves like a pen plotter, but the ink is a viscous melt under pressure. Bead width is set by the flow rate divided by the head speed, so flow and speed must change together. At the start of a path the pressure takes time to build, and at the end the nozzle keeps oozing: slicers retract the filament, coast to the end, and wipe the nozzle. At corners the head decelerates and a constant flow leaves a blob, so firmware links extrusion to the actual speed. Path order matters too: outer walls give accuracy, the seam is hidden at a corner, and travel moves are kept short.

Where do the mass, time and cost of an FDM part go?

A slicer does not fill a part solid. It prints a few perimeters, solid top and bottom layers, and a sparse infill inside. For a 60 × 40 × 30 mm PLA box with 3 walls, 4 solid layers top and bottom and 20 % infill, the plastic is 33 % of the box: 29.1 g and 9.7 m of filament. Doubling the infill to 40 % adds only half of the infill share, because the shells do not change. Time follows the volume divided by the average flow the printer sustains. The interactive prices filament at ₹1,200/kg for PLA, ₹1,100 for ABS and ₹1,300 for PETG, and machine time at ₹60/h (illustrative).

An FDM part: walls, top and bottom layers and infill to mass, filament length, print time and cost: interactive loads as you scroll

What are bio-extrusion, direct ink writing and robocasting?

They are extrusion without melting. A syringe, piston or screw pushes a paste or gel through a nozzle at about room temperature. Direct ink writing uses shear-thinning inks that flow under pressure and stand as soon as they leave the nozzle. Robocasting extrudes a concentrated ceramic slurry, which is dried and sintered into a dense ceramic. Bio-extrusion, or bioprinting, extrudes hydrogels carrying living cells at low pressure and temperature so the cells survive, then crosslinks the gel with UV light, ions or heat. Melt-extruded biodegradable polymers make porous tissue scaffolds.

Predict, then check

A small FDM box (60 × 40 × 30 mm, 3 walls, 4 solid layers) is printed at 40 % infill. You drop the infill to 20 %. The mass falls by about:

Worked examples

Mass, filament and cost of a printed box · PLA (1.24 g/cm³, typical) box 60 × 40 × 30 mm; 3 walls of 0.45 mm lines; 4 solid layers of 0.2 mm at top and at bottom; 20 % infill; filament ₹1,200/kg and machine ₹60/h (illustrative); average flow 5 mm³/s
  1. Wall thickness e = 3 × 0.45 = 1.35 mm; top + bottom slabs 8 × 0.2 = 1.6 mm thick: V = 60 × 40 × 1.6 = 3,840 mm³
  2. Between them, height 28.4 mm: walls = (2,400 − 57.3 × 37.3) × 28.4 = 7,461 mm³; core = 57.3 × 37.3 × 28.4 = 60,699 mm³
  3. Infill 0.20 × 60,699 = 12,140 mm³; total V = 23,441 mm³
  4. Mass = 1.24 × 23.44 cm³ = 29.1 g; filament = 23,441 / (π × 1.75²/4) = 9.75 m
  5. Time = 23,441/5 = 4,688 s = 1.30 h; cost = ₹34.9 + ₹78.1

Answer: 29.1 g, 9.7 m of filament, about 78 min, ₹113

What lighter infill saves · The same box at 40 % and at 10 % infill
  1. 40 %: V = 3,840 + 7,461 + 0.40 × 60,699 = 35,581 mm³ → 44.1 g
  2. 10 %: V = 17,371 mm³ → 21.5 g, time 58 min
  3. Going from 20 % to 10 % saves 20 min and ₹29
  4. From 40 % to 20 % the mass falls 34 %: the shells (11,301 mm³) do not change

Answer: 44.1 g at 40 %, 29.1 g at 20 %, 21.5 g at 10 %

How much filament is on a 1 kg spool · 1 kg spools of PLA (1.24 g/cm³) and ABS (1.04 g/cm³), 1.75 mm; PLA also in 2.85 mm
  1. Volume of PLA = 1,000 g / 1.24 = 806.5 cm³ = 806,452 mm³
  2. Area of 1.75 mm filament = π × 1.75²/4 = 2.4053 mm² → length = 335.3 m
  3. ABS is less dense: 961,538 mm³ → 399.8 m
  4. 2.85 mm PLA: area 6.379 mm² → 126.4 m

Answer: About 335 m of PLA, 400 m of ABS, and 126 m of 2.85 mm PLA

Chapter 9

Stereolithography: photopolymers and the working curve

Which syllabus covers thisAU CME339 Unit IIIVTU BME714A Module 2SPPU 402045C Unit 2MU MEC404 Module 5KTU MET468 Module 3

How does stereolithography work, top-down and bottom-up?

A vat holds liquid photopolymer. An ultraviolet laser, steered by two galvanometer mirrors, traces each layer's section, and the resin it touches turns solid. In the top-down layout the platform sits just under the resin surface and sinks one layer after each exposure while a recoater blade sweeps a fresh, even film over the part. In the bottom-up layout the laser shines up through a clear window in the vat floor; the part hangs from a platform that lifts one layer at a time, peeling the new layer off the window. Bottom-up needs less resin; top-down avoids peel forces.

What happens chemically when resin cures?

The resin is a mix of monomers and oligomers with a photoinitiator. UV light splits the photoinitiator into radicals (for acrylates) or acids (for epoxies) that start chain reactions, linking the molecules into a crosslinked network. The material reaches the gel point and stops flowing once it has received a minimum dose, the critical exposure Ec. The result is a thermoset: it cannot be remelted. Acrylates cure fast but shrink more, so parts curl; epoxies shrink less but cure slower, and many resins blend both. Light absorbers set how deep the light reaches.

What is the working curve?

Light fades exponentially into the resin, falling by 1/e every penetration depth Dp. Resin cures down to where the dose drops to Ec, so cure depth Cd = Dp ln(E/Ec): a straight line on a semilog plot. A Gaussian laser of power P and beam radius W0 moving at speed Vs gives the peak dose Emax = √(2/π)·P/(W0·Vs). With 30 mW, W0 = 0.125 mm, 500 mm/s and a resin with Dp = 0.12 mm and Ec = 10 mJ/cm², Emax = 38.3 mJ/cm² and Cd = 0.161 mm.

Stereolithography working curve: laser exposure, cure depth and overcure, and the width of a cured line: interactive loads as you scroll

Why must the cure depth exceed the layer thickness?

Each layer must glue itself to the one below, so the laser cures a little deeper than one layer; that extra is the overcure. In the example above a 0.1 mm layer gets 0.061 mm of overcure. Too little and layers delaminate; too much and downward-facing surfaces grow extra cured skin (print-through) and lose accuracy. The cured line is a parabolic cylinder 0.205 mm wide here, so the laser traces borders and then fills the inside with hatch lines spaced closer than one line width, so neighbours overlap. Faster scanning cuts the dose and the depth together.

What happens after the build, and where is SLA used?

The part leaves the machine wet and only partly cured (the green state). It is washed in a solvent such as isopropyl alcohol, its supports are clipped off and the marks sanded, and it is post-cured under UV light, often with heat, to finish the reaction and reach full strength. SLA gives smooth surfaces, fine detail and clear parts, so it is used for appearance models, master patterns for silicone moulds and investment casting, dental models and microfluidic chips. Drawbacks: resins are brittle compared with thermoplastics, degrade in sunlight, need careful handling, and only photopolymers can be used.

Predict, then check

On an SLA machine you double the scan speed and change nothing else. The cure depth:

Worked examples

Dose, cure depth and line width · Laser 30 mW, beam radius W0 = 0.125 mm, scan speed 500 mm/s; resin Dp = 0.12 mm, Ec = 10 mJ/cm² (illustrative); layer 0.1 mm
  1. Work in cm: W0 = 0.0125 cm, Vs = 50 cm/s
  2. Emax = √(2/π) × 30 / (0.0125 × 50) = 0.7979 × 48 = 38.30 mJ/cm²
  3. Cd = 0.12 × ln(38.30/10) = 0.12 × 1.343 = 0.161 mm; overcure = 0.061 mm
  4. Lw = 2 × 0.125 × √(0.161/0.24) = 0.205 mm

Answer: Emax = 38.3 mJ/cm², Cd = 0.161 mm (0.061 mm overcure), line width 0.205 mm

Finding Dp and Ec from two test exposures · Windowpane test on a new resin: 20 mJ/cm² cures 0.083 mm, 80 mJ/cm² cures 0.250 mm
  1. Two points on Cd = Dp ln E − Dp ln Ec: subtract to remove Ec
  2. Dp = (0.250 − 0.083) / ln(80/20) = 0.167 / 1.3863 = 0.1205 mm
  3. Ec = 20 × e^(−0.083/0.1205) = 20 × 0.5021 = 10.04 mJ/cm²
  4. In practice fit a straight line through five or more points on a semilog plot

Answer: Dp ≈ 0.120 mm, Ec ≈ 10.0 mJ/cm²

Scan speed for a chosen overcure · Same laser and resin; 0.1 mm layers with 0.05 mm overcure, so Cd = 0.15 mm
  1. Dose needed: E = Ec e^(Cd/Dp) = 10 × e^(0.15/0.12) = 34.90 mJ/cm²
  2. Vs = √(2/π) × P / (W0 × E) = 0.7979 × 30 / (0.0125 × 34.90) = 54.86 cm/s
  3. = 549 mm/s, about 10 % faster than the 500 mm/s run

Answer: Scan at about 549 mm/s

Chapter 10

DLP, CLIP and direct laser writing

Which syllabus covers thisAU CME339 Unit IIIVTU BME714A Module 2SPPU 402045C Unit 2MU not in syllabusKTU not in syllabus

How does DLP differ from laser SLA?

Digital light processing cures a whole layer in one flash. A projector forms the layer's image with a digital micromirror device, a chip of tiny tilting mirrors, one per pixel, each sending light into the resin or away from it. Masked SLA does the same with an LCD screen over a UV backlight. The layer is cured as a mosaic of voxels, so resolution in the plane is the pixel size: projected width divided by the number of pixels. Greyscale at the edges smooths the steps. Most machines work bottom-up, so every layer ends with a lift and peel.

When is a projector faster than a laser?

A laser's layer time grows with the area it must hatch; a projector's does not. With illustrative settings (0.1 mm hatch at 1,000 mm/s plus 12 s recoat for SLA; 2.5 s exposure plus 6 s peel for DLP), four parts of 400 mm² section, 30 mm tall in 0.05 mm layers, take 4.67 h by SLA and 1.42 h by DLP. Twelve parts take 10 h by SLA but still 1.42 h by DLP, as long as they fit in the projected field. The laser keeps its edge on large platforms and fine, uniform spots.

SLA, DLP and CLIP build times: scanning a laser against projecting a whole layer, as the platform fills with parts: interactive loads as you scroll

What makes CLIP continuous?

Continuous liquid interface production is a bottom-up projection process whose window lets oxygen through. Oxygen stops the radical curing reaction, so a thin layer of resin just above the window never solidifies: the dead zone. The part therefore never sticks to the window, there is no peel step, and the platform can rise smoothly while the projector plays the slices as a film. The 2015 paper that introduced it reported parts tens of centimetres in size with features below 100 µm, drawn out at rates of hundreds of millimetres per hour. Resin must keep flowing into the gap, so wide solid sections print slower.

What is micro-stereolithography?

Stereolithography scaled down to make parts with features of tens of micrometres or less. A laser can be focused to a very small spot through a high-quality lens, or a micromirror projector can image a whole layer through reduction optics (projection micro-stereolithography). Layers are a few to tens of micrometres: one study on a desktop DLP machine with 912 × 1,140 micromirrors used 10, 25 and 50 µm layers. Thinner layers there gave stronger parts but take longer. Uses: microfluidic channels, micro-lenses, small medical devices and tissue scaffolds.

How does direct laser writing (two-photon polymerisation) work?

A pulsed near-infrared laser is focused deep inside a resin that ignores single infrared photons. Curing needs two photons absorbed almost at once, which happens only where the intensity peaks, at the focal spot. So the resin cures in a tiny voxel anywhere in the volume, not just at a surface, and moving the focus writes 3D structures directly. Features well below a micrometre are typical. It is slow and builds parts of millimetres at most, so it is used for photonic structures, micro-needles, micro-optics and cell scaffolds.

Predict, then check

A bottom-up DLP printer takes 1.42 h for 4 small parts. You add 4 more identical parts that still fit inside the projected image. The build now takes about:

Worked examples

SLA and DLP build times for a batch · 4 parts, 400 mm² section each, 30 mm tall, 0.05 mm layers. SLA: hatch 0.1 mm, 1,000 mm/s, recoat 12 s. DLP: 2.5 s exposure, 6 s peel (all illustrative)
  1. Layers = 30 / 0.05 = 600
  2. SLA layer: scan 4 × 400 / (0.1 × 1,000) = 16 s + 12 s = 28 s → 600 × 28 = 16,800 s = 4.67 h
  3. DLP layer: 2.5 + 6 = 8.5 s → 600 × 8.5 = 5,100 s = 1.42 h
  4. 12 parts: SLA layer 48 + 12 = 60 s → 10 h; DLP unchanged at 1.42 h

Answer: 4 parts: SLA 4.67 h, DLP 1.42 h; 12 parts: SLA 10 h, DLP 1.42 h

Pixel size of a projector · A 1920-pixel-wide projector imaging 96 mm; then the same projector stretched to 192 mm; then a 3840-pixel projector over 192 mm
  1. 96 mm / 1,920 px = 50 µm per pixel; the field is 96 × 54 = 5,184 mm²
  2. Stretched to 192 mm: 192 / 1,920 = 100 µm, twice as coarse, for four times the area
  3. 3,840 px over 192 mm: 50 µm again

Answer: 50 µm; a bigger platform needs more pixels to keep the same detail

CLIP against layer-by-layer DLP · The 30 mm parts above; CLIP at 100 mm/h (illustrative)
  1. CLIP time = 30 / 100 h = 18 min
  2. Equivalent to 1.8 s per 0.05 mm slice, against 8.5 s for DLP with peel
  3. DLP / CLIP = 85 / 18 = 4.7

Answer: 18 min by CLIP, about 4.7 times faster than 1.42 h by DLP

Chapter 11

Selective laser sintering and multi-jet fusion of polymers

Which syllabus covers thisAU CME339 Unit IVVTU BME714A Module 2SPPU 402045C Units 2, 3MU MEC404 Module 5KTU MET468 Module 3

How does selective laser sintering work?

A roller or blade spreads a thin layer of polymer powder, about 0.1 mm typical, over a bed held hot, just below the powder's melting point (one PA12 powder melts at 187 °C). A CO₂ laser scans the layer's section, adding just enough heat to fuse the particles, and the bed drops one layer. The loose powder around the parts holds them up, so no supports are needed and parts can be stacked in three dimensions. The whole block of powder, the cake, then cools slowly before the parts are dug out and bead-blasted clean.

What are the powder fusion mechanisms?

Solid-state sintering: below the melting point, atoms diffuse at the contacts between particles and necks grow; it is slow and leaves pores. Chemically induced binding: a reaction joins the particles. Liquid-phase sintering or partial melting: one component, often a polymer binder or low-melting alloy, melts and wets the solid particles, which is how metal or ceramic powders coated in polymer were laser sintered into green parts, then debound and furnace-sintered or infiltrated. Full melting: the particles melt completely, as PA12 does in SLS and metals do in laser melting, giving nearly dense parts.

Which materials suit SLS, and what are the parts like?

Nylon PA12 and PA11 dominate. They melt over a narrow range and recrystallise slowly, so the hot bed stays stable. Glass- or carbon-filled nylons are stiffer; TPU gives flexible parts; polypropylene is also used. Parts are tough and nearly the same in every direction but slightly porous with a grainy surface. In one PA12 datasheet the loose powder has a bulk density of 0.425 g/cm³ and the parts 1.01 g/cm³, so a powder bed more than halves its volume where it fuses. Uses: functional prototypes, ducts, snap-fits, living hinges and small batches of end-use parts.

Why pack an SLS build tightly, and what is the refresh ratio?

The machine spreads and heats every layer of the bed whether it holds one part or many, so cost per part falls as the build fills. With illustrative rates (machine ₹1,200/h, powder ₹4,000/kg, labour ₹1,500 per build), one 60 × 40 × 20 mm bracket costs ₹2,932, but a full build of 66 brackets costs ₹260 each, at only 11 % packing density. Powder that has sat hot ages, so each fill mixes in fresh powder: the refresh ratio. One maker recommends 30 % for its PA12 and 20 % for TPU; one PA12 for multi-jet fusion is rated at 20 %.

Nesting parts in an SLS build: packing density, build time, powder used and cost per part: interactive loads as you scroll

How does multi-jet fusion differ from SLS?

Multi-jet fusion replaces the laser with an inkjet head and a lamp. After each layer of powder is spread, the head prints a fusing agent, which absorbs infrared, wherever the part should be, and a detailing agent around the edges that keeps neighbouring powder cool and edges sharp. An infrared lamp then sweeps the whole bed: powder with fusing agent heats up and fuses, the rest does not. Like DLP against SLA, the layer time hardly depends on how much is printed. Parts are grey (the agent is dark) and the same nylon and TPU families are used.

What are the limits of polymer powder bed fusion?

Long cycles: the cake must cool slowly, often for many hours, or parts warp. Shrinkage on cooling must be scaled out in the build file, and large flat parts curl. Surfaces are grainy and need blasting, dyeing or vapour smoothing for appearance. Closed hollows trap powder, so drain holes are needed. Powder handling needs care: fine powder in the air is a breathing and dust-explosion hazard. And the unfused powder ages, so the refresh ratio sets how much of the bed has to be bought fresh.

Predict, then check

An SLS build runs at a 30 % refresh ratio with parts filling about 10 % of the powder bed. The fresh powder bought per build is set by:

Worked examples

How many parts fit, and the packing density · Build volume 160 × 160 × 300 mm; bracket bounding box 60 × 40 × 20 mm, 12,000 mm³ of solid; 5 mm gaps between parts and at the walls
  1. Across: (160 − 5)/(60 + 5) = 2.38 → 2; deep: (160 − 5)/(40 + 5) = 3.44 → 3; 6 per level (turning the part also gives 3 × 2)
  2. Up: (300 − 5)/(20 + 5) = 11.8 → 11 levels; 66 parts at most
  3. Height used = 11 × 25 + 5 = 280 mm; bed volume = 160 × 160 × 280 = 7,168,000 mm³
  4. Packing = 66 × 12,000 / 7,168,000 = 11.0 %

Answer: 66 brackets at 11.0 % packing density

Cost per part, part-full and full · Layers 0.1 mm, 10 s recoat; scan 0.2 mm hatch × 5,000 mm/s; ₹1,200/h machine, ₹4,000/kg powder, ₹1,500 labour per build; 30 % refresh; bulk powder 0.425 g/cm³ (all rates illustrative)
  1. 20 parts: 4 levels, bed 105 mm → 1,050 layers × 10 s = 10,500 s; scan 20 × 12,000 / 0.1 / 1,000 = 2,400 s; total 3.58 h
  2. Bed 160 × 160 × 105 mm = 2,688 cm³; fresh powder 0.3 × 2,688 × 0.425 = 343 g; cost ₹4,300 + ₹1,371 + ₹1,500 = ₹7,171 → ₹359 per part
  3. 66 parts: 2,800 layers × 10 s + 7,920 s scan = 9.98 h; fresh 914 g; ₹17,129 → ₹260 per part
  4. A single part would cost ₹2,932

Answer: ₹359 per part for 20, ₹260 for a full build of 66

Fresh powder, and when the parts set it · The full build: bed 3.05 kg of powder, parts 0.80 kg (part density 1.01 g/cm³)
  1. At 30 % refresh: fresh = 0.3 × 3046 = 914 g, more than the 800 g of parts
  2. The surplus 114 g of used powder cannot all go back in and is set aside
  3. At 20 %: 0.2 × 3046 = 609 g is less than the parts, so fresh = 800 g: you only replace what left as parts

Answer: 914 g fresh at 30 % refresh; 800 g at 20 %

Chapter 12

Metal powder bed fusion: SLM, DMLS and EBM

Which syllabus covers thisAU CME339 Unit IVVTU BME714A Module 2SPPU 402045C Unit 2MU not in syllabusKTU MET468 Modules 3, 4

How do SLM and DMLS make metal parts?

Both are laser powder bed fusion (PBF-LB/M in ISO/ASTM terms), and the names are used for the same process: a fibre laser of a few hundred watts fully melts a thin layer of metal powder, typically 20 to 60 µm, along the section, and the melt solidifies into dense metal. The chamber is filled with argon or nitrogen to keep oxygen out, and a gas stream carries away spatter and fumes. Parts grow welded to a thick steel or titanium build plate. Common alloys: 316L and 17-4PH stainless steels, maraging steel, Ti-6Al-4V, AlSi10Mg, Inconel 718 and 625, and cobalt-chrome.

What is volumetric energy density, and why is there a window?

VED = P/(v·h·t) is the laser energy per unit volume of layer, with power P, scan speed v, hatch spacing h and layer thickness t. Too little leaves unmelted powder between tracks and layers: lack-of-fusion pores. Too much drives the melt pool into a deep vapour keyhole that collapses and traps round pores. Published windows differ with machine and powder. On one Ti-6Al-4V machine 41 to 64 J/mm³ gave over 99.6 % density, with lack of fusion at 34 and lower density by keyholing at 86 to 128. 316L at 200 W, 800 mm/s, 0.1 mm hatch and 30 µm layers gets 83.3 J/mm³.

Metal powder bed fusion: volumetric energy density, the process window between lack of fusion and keyholing, and build rate: interactive loads as you scroll

What goes wrong in metal powder bed fusion?

Pores: lack of fusion (irregular, at too little energy), keyhole (round, at too much) and gas pores from the powder. Balling: the melt beads up instead of wetting the layer below. Spatter: ejected droplets land on the bed and spoil later layers. The worst problem is residual stress: each tiny melt pool shrinks as it cools against cold metal below, so parts curl, crack or tear off their supports. Remedies are supports that anchor and conduct heat away, short scan vectors in islands or stripes rotated each layer (67° is a common choice), a heated plate, and stress relief before cutting the parts off.

How is electron beam melting different?

EBM melts the powder with an electron beam in a vacuum. Magnetic coils steer the beam with no moving mirrors, so it can jump very fast. Before melting each layer, a defocused beam preheats the whole bed to several hundred °C, lightly sintering the powder so it holds still and conducts the charge away. The hot bed means much lower residual stress and fewer supports, and the vacuum suits reactive metals such as titanium. The costs: only conductive metals, coarser powder and layers, rougher surfaces, and a sintered cake that must be blasted off.

What post-processing do metal PBF parts need, and where are they used?

A typical sequence: stress-relieve the parts while still on the plate, cut them off with wire EDM or a band saw, remove supports, then heat-treat. Laser-melted Ti-6Al-4V cools so fast it forms brittle α′ martensite, which heat treatment converts. Hot isostatic pressing, holding parts at high temperature under high-pressure argon, closes internal pores and raises fatigue life. Critical faces and holes are machined, and surfaces blasted or polished. Uses: lightweight aerospace brackets, fuel and hydraulic manifolds, heat exchangers, tooling inserts with conformal cooling, and orthopaedic implants with porous lattice surfaces.

Predict, then check

You double the layer thickness in laser powder bed fusion, keeping power, speed and hatch the same. The VED:

Worked examples

VED and build rate for 316L · P = 200 W, v = 800 mm/s, h = 0.10 mm, t = 0.03 mm; illustrative 316L window 60-150 J/mm³
  1. VED = 200 / (800 × 0.10 × 0.03) = 200 / 2.4 = 83.3 J/mm³: inside the window
  2. Melt rate = v h t = 2.4 mm³/s = 8.64 cm³/h (scanning only)

Answer: 83.3 J/mm³ and 8.64 cm³/h

Thicker layers for speed · Same 316L parameters, but 60 µm layers
  1. VED = 200 / (800 × 0.10 × 0.06) = 41.7 J/mm³: lack-of-fusion risk
  2. To restore 83.3 J/mm³ at 800 mm/s: P = 83.33 × 800 × 0.10 × 0.06 = 400 W
  3. Melt rate doubles to 17.28 cm³/h, and half as many layers need recoating
  4. Slowing to 400 mm/s at 200 W would restore the VED but give back the speed

Answer: 41.7 J/mm³ at 200 W; about 400 W restores the window and doubles the rate

Why recoating time matters · A Ti-6Al-4V part of 50 cm³, 60 mm tall; P = 280 W, v = 1,200 mm/s, h = 0.12 mm, t = 0.03 mm; 8 s recoat per layer (illustrative)
  1. VED = 280 / (1,200 × 0.12 × 0.03) = 64.8 J/mm³
  2. Melt rate = 1,200 × 0.12 × 0.03 = 4.32 mm³/s = 15.55 cm³/h → scan time 3.22 h
  3. Layers = 60 / 0.03 = 2,000 → recoat time 2,000 × 8 s = 4.44 h
  4. Total about 7.7 h; recoating is 58 % of it, and does not shrink if more parts share the plate

Answer: About 7.7 h, of which 4.4 h is recoating

Chapter 13

Directed energy deposition: LENS, laser cladding and WAAM

Which syllabus covers thisAU CME339 Unit IIIVTU BME714A Module 3SPPU 402045C Units 2, 3MU not in syllabusKTU MET468 Module 3

What is directed energy deposition?

A focused heat source, a laser, an electron beam or an electric arc, melts a small pool on a substrate, and metal powder or wire is fed straight into the pool. The head, carried by a CNC gantry or a robot, lays beads side by side and layer on layer, like automated welding. Because material goes only where the head goes, DED can build on an existing part, on curved surfaces, and to sizes far beyond a powder bed. Beads are larger than in powder bed fusion, so parts come out near-net shape and are finish-machined, often in hybrid machines that also mill.

How do LENS, laser blown powder and laser cladding work?

In laser blown powder deposition (laser engineered net shaping, LENS, and direct metal deposition, DMD, are names for it), a nozzle round the laser blows powder in argon into the melt pool, inside a shielding gas. Not every grain lands in the pool: the fraction that does is the catchment efficiency, which one 316L study measured from 25 to 96 % as spot size and gas flow changed, at about 10 g/min of powder. Two powder feeders can grade the composition along a part. Laser cladding uses the same head to coat a surface with a thin wear- or corrosion-resistant layer, under 1 mm, and to rebuild worn turbine blades, dies and shafts.

How do wire-fed DED, WAAM and electron beam DED work?

Wire puts nearly all of its metal into the part, is cheaper than powder and is clean to handle; some laser-wire systems use 0.8 to 1.2 mm welding wire. Wire arc additive manufacturing (WAAM) uses welding arcs: gas metal arc (MIG), where the wire itself is the electrode and the rate is highest; cold metal transfer, a lower-heat MIG variant; and tungsten or plasma arcs with a separate wire feed. Typical WAAM rates are about 2 to 12 kg/h with layers 1 to 3 mm high. Electron beam DED melts wire in a vacuum, which suits large titanium parts.

How fast does DED deposit metal, and how big is a bead?

With wire, the deposition rate is wire area × feed speed × density: 1.2 mm steel wire at 6 m/min lays 3.2 kg/h. Spread along the travel speed, that gives the bead area, and a bead shaped as a parabola of width w and height h has area (2/3)·w·h. With powder, the rate is the feed times the catchment efficiency. Typical bands: laser powder bed fusion 0.1 to 0.5 kg/h, laser metal deposition 0.1 to 2 kg/h, WAAM 2 to 12 kg/h. Speed costs resolution: WAAM beads are millimetres high.

Wire arc and laser directed energy deposition: deposition rate, bead size and the time to build a wall: interactive loads as you scroll

Which parameters control DED, and what microstructure results?

Power or arc current, travel speed, feed rate, spot size or stand-off, shielding gas, path pattern and the interpass temperature. Heat input per unit length (power / travel speed) sets bead size, dilution into the layer below, distortion and cooling rate. Steels, stainless steels, Ti-6Al-4V, nickel superalloys, aluminium and copper alloys are all deposited. Each new layer re-melts the top of the last and reheats several below, so grains grow upward through layers as long columns along the build direction, with banded structure and direction-dependent properties. High-heat WAAM gives the coarsest grains, so parts are often heat-treated and always machined.

Predict, then check

A WAAM wall is built at a fixed wire feed speed. You double the travel speed. The deposition rate (kg/h):

Worked examples

WAAM deposition rate and bead size · Steel wire 1.2 mm (7.85 g/cm³), wire feed 6 m/min, travel 8 mm/s, bead width 7 mm
  1. Wire area = π × 1.2²/4 = 1.1310 mm²; feed 6 m/min = 100 mm/s → 113.1 mm³/s
  2. Rate = 113.1 × 7.85 × 10⁻³ g/mm³ × 3,600 s/h = 3,196 g/h = 3.20 kg/h
  3. Bead area = 113.1 / 8 = 14.14 mm²
  4. Parabolic bead: h = 1.5 A / w = 1.5 × 14.14 / 7 = 3.03 mm

Answer: 3.20 kg/h, bead 14.1 mm² and 3.03 mm high

Time to build a wall · A single-bead wall 200 mm long and 60 mm high with the beads above
  1. Layers = ceil(60 / 3.029) = 20
  2. Arc-on time = 20 × 200 mm / 8 mm/s = 500 s = 8.3 min
  3. Metal = 14.14 × 200 × 20 mm³ × 7.85 × 10⁻³ = 444 g
  4. Check: 3.196 kg/h × 500/3,600 h = 444 g
  5. Real builds add cooling pauses between layers to control the interpass temperature

Answer: 20 layers, 8.3 min of arc time, about 444 g of steel

Laser powder DED of 316L · Powder fed at 10 g/min, catchment efficiency 60 %, travel 10 mm/s, bead width 2.5 mm, 316L 7.99 g/cm³
  1. Deposited = 10 × 0.60 = 6 g/min = 0.36 kg/h; 4 g/min is lost as overspray
  2. Volume = 6 / (7.99 × 10⁻³) / 60 = 12.52 mm³/s; bead area = 12.52/10 = 1.252 mm²
  3. Bead height = 1.5 × 1.252 / 2.5 = 0.75 mm
  4. The WAAM example deposits 8.9 times faster, with beads four times taller

Answer: 0.36 kg/h, bead 1.25 mm² and 0.75 mm high

Chapter 14

Binder jetting and sintering shrinkage

Which syllabus covers thisAU CME339 Unit VVTU BME714A Module 3SPPU 402045C Unit 3MU MEC404 Module 5KTU MET468 Module 4

How does binder jetting build a part?

A roller or blade spreads a thin layer of powder across the build box. An inkjet printhead then passes over it and drops a liquid binder only where the slice has material, gluing those grains together. The box drops by one layer and the cycle repeats. Nothing is melted while printing, so there is no thermal stress, and the loose powder holds the part up, so no supports are needed and parts can be stacked in 3D through the whole box. The result is a fragile green part that is cured, lifted out of the powder (depowdering) and then densified.

What can binder jetting print, and what are the variants?

Metals (316L and 17-4PH stainless, tool steels) for parts that are later sintered; silica sand for foundry moulds and cores, poured straight after printing; ceramics; and gypsum-like powders for full-colour models, where coloured binders are jetted like the inks of a paper printer (often called colour-jet printing). Universities teach the original powder-and-binder process as 3DP or three-dimensional printing. Sand moulds are an indirect route: the print is the tool, and the casting is the part.

Why does a sintered binder-jet part shrink, and by how much?

The green part is only loosely packed powder: for 316L, published studies report green relative densities of about 50-60%. Sintering pulls the grains together, to well above 90% dense and above 99% in the best cases, and since the mass is unchanged the volume falls by ρg/ρf, a 40-50% volume loss. Shrinkage is not quite uniform: one study of fine 316L measured 18.4%, 17.7% and 21.5% linear shrinkage in X, Y and Z, Z being largest. Designers scale the CAD model up by 1/(1 − s) per axis, and support long spans so they do not sag in the furnace.

Binder jetting shrinkage: from green density to sintered density, linear shrinkage and the CAD scale-up factor: interactive loads as you scroll

How much binder goes into each layer?

Saturation is the share of the pore space between grains that the binder fills. Binder volume per layer = saturation × (1 − packing density) × section area × layer thickness. With too little binder the green part crumbles when depowdered; with too much, binder bleeds sideways into loose powder, edges grow and fine features blur. The 316L studies above printed at 80% saturation. Binder for a 40 × 40 mm section at 55% packing, 0.1 mm layers and 80% saturation is 57.6 mm³ per layer.

Where is binder jetting strong, and where is it weak?

Strong: it is fast because the head prints a whole swath at once, it needs no supports, the whole box can be filled with parts, it prints at room temperature, and it is the main way to print colour and sand moulds. Weak: the green parts are fragile; metal parts must be debound and sintered, with large shrinkage that has to be predicted; some porosity can remain unless the part is sintered fully or infiltrated (a lower-melting metal is wicked into the pores); and accuracy depends on how evenly the part shrinks.

Predict, then check

A binder-jet steel part is printed at 55% relative density and sintered to 97%. Roughly how much does each dimension shrink?

Worked examples

CAD scale factor for an isotropic part · Green relative density 55%, sintered relative density 97%, a 40 mm feature on the finished part; assume equal shrinkage in all directions
  1. Volume ratio V_f/V_g = ρg/ρf = 0.55/0.97 = 0.5670
  2. Length ratio = (0.5670)^(1/3) = 0.8277, so linear shrinkage s = 17.23%
  3. Scale factor k = 1/(1 − s) = 1.2082
  4. Print the feature at 40 × 1.2082 = 48.33 mm
  5. Check: 48.33³ × 0.55 = 40³ × 0.97 = 62,080 mm³ of solid metal

Answer: s = 17.2%, scale the CAD model by 1.208: 40 mm is printed as 48.3 mm

Extra shrinkage along Z · Same densities, but the part shrinks 3% more along Z: (1 − s_z) = (1 − s_xy)(1 − 0.03). Finished part a 50 mm cube
  1. Mass balance: (1 − s_xy)² (1 − s_z) = ρg/ρf, so (1 − s_xy)³ × 0.97 = 0.5670
  2. 1 − s_xy = (0.5670/0.97)^(1/3) = 0.8361, s_xy = 16.39%
  3. 1 − s_z = 0.8361 × 0.97 = 0.8110, s_z = 18.90%
  4. Green size: 50/0.8361 = 59.80 mm in X and Y, 50/0.8110 = 61.65 mm in Z

Answer: Print a 59.8 × 59.8 × 61.6 mm block (s_xy = 16.4%, s_z = 18.9%)

Binder needed for a build · Part section 40 × 40 mm, 30 mm tall, 0.1 mm layers, powder packed at 55%, saturation 80%
  1. Pore volume per layer = (1 − 0.55) × 1,600 × 0.1 = 72 mm³
  2. Binder per layer = 0.8 × 72 = 57.6 mm³
  3. Layers = 30/0.1 = 300
  4. Binder for the part = 57.6 × 300 = 17,280 mm³ = 17.28 mL

Answer: 57.6 mm³ per layer, about 17.3 mL for the part

Chapter 15

Material jetting, sheet lamination and direct write

Which syllabus covers thisAU CME339 Unit VVTU BME714A Module 3SPPU 402045C Unit 3MU MEC404 Module 5KTU MET468 Modules 3, 4

How does material jetting (PolyJet, multi-jet modelling) work?

Piezoelectric printheads, like those in an office inkjet, fire droplets of liquid photopolymer straight onto the build tray; UV lamps on the carriage cure each pass as it lands. A second, softer support material is jetted into overhangs and later washed or melted away. Because every droplet can be a different material, one print can mix rigid, rubbery, clear and coloured resins. Surfaces are smooth and detail is fine: Stratasys' Objet30 V5 sheet lists 600 × 600 dpi in X and Y and layers of 16-36 µm. The resins are costly, and parts are weaker and age under light more than moulded ones.

How do droplet volume and resolution set the layer thickness and speed?

At R dots per inch the drop pitch is 25.4/R mm. If one droplet of volume V lands in each pitch cell and levels out, it forms a layer t = V/(px·py) thick. At 600 × 600 dpi the pitch is 42.3 µm, so a 30 pL droplet gives 16.7 µm. The peak volume rate is nozzles × heads × firing frequency × V; real machines run slower, because nozzles fire only where the slice has material and the head makes several passes. Halving the pitch in both directions with the same droplet makes the layer four times thicker.

Material jetting: droplet volume, print resolution, layer thickness and how fast a printhead lays down material: interactive loads as you scroll

What other jetting processes appear in the syllabus?

Multi-jet modelling (MJM) jets either photopolymer or wax; wax prints serve as patterns for investment casting. Nanoparticle jetting suspends metal or ceramic nanoparticles in a liquid, jets them, drives off the liquid with heat and then sinters the part. Colour-jet printing is really binder jetting with coloured binders on a white powder. Drop-on-demand jetting of molten material is used for wax patterns. Jetting is accurate and multi-material, but it is limited to fluids thin enough to pass through a nozzle tens of micrometres wide.

How do laminated object manufacturing and ultrasonic consolidation work?

Sheet lamination stacks and bonds sheets, then cuts each to the slice outline. In LOM, a heated roller presses adhesive-coated paper onto the stack (thermal bonding); other machines spread glue first (gluing). A CO₂ laser or knife cuts the contour and crosshatches the waste into cubes, which are broken out afterwards. Parts are cheap and large but woody, and they absorb moisture unless sealed. Ultrasonic consolidation welds thin metal foils together with ultrasonic vibration under pressure, well below the melting point, and a CNC cutter trims the outline. Because it runs cool, sensors or fibres can be embedded between foils.

What is direct write, and what is it used for?

Direct write places small amounts of functional material precisely on a surface, often a curved one, rather than building bulk parts. Ink-based methods push or jet conductive, dielectric or ceramic inks: direct ink writing (DIW) extrudes a paste through a fine nozzle, and robocasting is DIW with stiff ceramic slurries that hold their shape until sintered. Laser transfer (laser-induced forward transfer) uses a laser pulse to fire a small spot of material from a coated film onto the substrate. Thermal spray sprays molten particles through a mask. Uses include printed antennas, strain gauges, conductive tracks on 3D surfaces and ceramic scaffolds.

Predict, then check

A material jetting head is switched from 600 × 600 dpi to 1200 × 1200 dpi with the same 30 pL droplet. What happens to the layer thickness?

Worked examples

Layer thickness from droplet volume · 600 × 600 dpi, 30 pL droplets, one droplet per cell; part 50 mm tall
  1. Pitch = 25.4/600 = 0.04233 mm = 42.33 µm
  2. Droplet volume 30 pL = 30 × 10⁻¹² L = 30 × 10⁻⁶ mm³
  3. t = V/p² = 30 × 10⁻⁶ / 0.04233² = 0.01674 mm = 16.7 µm
  4. Layers for 50 mm = ⌈50/0.01674⌉ = 2,987

Answer: Pitch 42.3 µm, layer 16.7 µm, 2,987 layers

Upper bound on jetting speed · 4 heads × 1,024 nozzles, 10 kHz firing, 30 pL droplets, 600 × 600 dpi, a full 100 × 100 mm layer
  1. Drops per second = 4 × 1,024 × 10,000 = 40,960,000
  2. Drops needed = 10,000 mm² / 0.04233² = 5,580,011
  3. Time per layer ≥ 5,580,011 / 40,960,000 = 0.136 s
  4. Peak volume rate = 40,960,000 × 30 × 10⁻⁶ = 1,228.8 mm³/s = 4,424 cm³/h
  5. Real builds are far slower: several passes per layer, curing, and nozzles that fire only inside the slice

Answer: At least 0.14 s per full layer; peak 1,229 mm³/s

LOM build time (illustrative values) · Part 60 mm tall, paper 0.1 mm thick; per layer: 450 mm of contour and 900 mm of crosshatch cut at 250 mm/s, plus 25 s to feed and laminate the sheet
  1. Layers = 60/0.1 = 600
  2. Cutting per layer = (450 + 900)/250 = 5.4 s
  3. Time per layer = 5.4 + 25 = 30.4 s
  4. Build time = 600 × 30.4 = 18,240 s = 5.07 h
  5. Laminating, not cutting, dominates, which is why sheet lamination is quick for bulky parts

Answer: About 5.1 h for 600 sheets

Chapter 16

Design for AM: part consolidation, topology optimisation and lattices

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Module 1SPPU 402045C Unit 4MU not in syllabusKTU not in syllabus

What does additive manufacturing change for a designer?

In machining or moulding, every pocket, undercut or internal channel adds cost or is impossible. In AM, the cost of a part depends mostly on its volume, its height in the build and its post-processing, not on how complex it is. That frees internal channels, organic shapes, graded or lattice infills and custom one-offs. New limits replace the old ones: overhangs that need support, trapped powder or resin that must drain, minimum wall and hole sizes, anisotropic strength between layers, residual stress and distortion in metals, and surfaces that need finishing where they mate.

What is part consolidation, and when is it worth it?

Part consolidation redesigns an assembly of many parts (brackets, tubes, fasteners, seals) as one printed part. The gains: fewer joints that can leak or loosen, fewer part numbers to stock, no assembly labour, and often lower mass because flanges and bolts disappear. It is worth it when the parts are made of the same material and are never serviced separately. Check that the consolidated part fits the build volume, that a worn region does not force scrapping the whole part, that internal passages can be cleaned and inspected, and that the print still costs less than the assembly it replaces.

How do topology optimisation and generative design differ?

Topology optimisation starts from a design space, loads, supports and a material budget, and lets a finite-element solver decide where material is needed. The common SIMP method gives each element a density between 0 and 1 and minimises compliance (maximises stiffness) for a target volume fraction. The output is a rough, organic shape that is then smoothed into CAD and checked by FEA. Generative design runs many such studies with different goals, materials and manufacturing methods and offers a range of candidate designs, so the engineer chooses among trade-offs instead of refining a single answer.

How dense is a lattice, and how stiff and strong does it stay?

Relative density ρ* is strut volume over cell volume. To first order ρ* ≈ c₂(d/a)²: c₂ = 3π/4 for simple cubic, √3π for BCC and 3√2π for octet. This counts node material twice, so it overestimates. At d/a = 0.1 the error is about 6% for simple cubic, 10% for BCC and 17% for octet; a −c₃(d/a)³ node term removes most of it. Ashby's scaling laws: bending-dominated cells such as BCC keep E*/Es ≈ ρ*² and σ*/σy ≈ 0.3ρ*^1.5. The stretch-dominated octet keeps E*/Es ≈ ρ*/3 and σ*/σy ≈ ρ*/3 (an upper bound), several times stiffer at the same weight.

Lattice structures: relative density of cubic, BCC and octet cells and the stiffness and strength they keep (Gibson-Ashby): interactive loads as you scroll

What design rules apply to extrusion (FDM) parts?

These are typical guidelines, not fixed limits. Keep overhangs within about 45° of vertical, or add supports or chamfers. Bridges print cleanly only over short spans. Make walls a whole number of extrusion widths, at least two lines thick. Horizontal holes print rough at the top, so use a teardrop or diamond shape, or drill them afterwards. Vertical holes print undersize, so add clearance. Strength across layers (Z) is lower than along the roads, so orient the part so loads run along the layers. Use fillets or chamfers on the first layer, avoid tall thin pins, and leave 0.2-0.5 mm clearance between parts printed in place.

Predict, then check

Two titanium lattices both have relative density 0.1: one is BCC, the other octet. Roughly how much stiffer is the octet?

Worked examples

A BCC titanium lattice block · BCC cell a = 5 mm, strut d = 0.8 mm, Ti-6Al-4V (typical E = 114 GPa, σy = 880 MPa, 4.43 g/cm³); a 50 mm cube
  1. d/a = 0.16; first order ρ* = √3π × 0.16² = 0.1393
  2. Node correction: − 4.897 × 0.16³ = − 0.0201, so ρ* = 0.1192
  3. Bending-dominated: E* ≈ ρ*² Es = 0.1192² × 114 = 1.62 GPa
  4. σ* ≈ 0.3 ρ*^1.5 σy = 0.3 × 0.1192^1.5 × 880 = 10.9 MPa
  5. Mass: solid cube 4.43 × 125 = 553.8 g; lattice 0.1192 × 553.8 = 66.0 g

Answer: ρ* = 0.119, E* ≈ 1.6 GPa, σ* ≈ 11 MPa, 66 g instead of 554 g

Sizing octet struts for a target density · Octet cells a = 4 mm in 316L (typical E = 193 GPa, proof stress 170 MPa), target ρ* = 0.10
  1. Solve 3√2π x² − 19.32 x³ = 0.10 for x = d/a (Newton's method): x = 0.0931
  2. Strut diameter d = 4 × 0.0931 = 0.373 mm (first order alone would give 0.346 mm, too thin)
  3. Stretch-dominated: E* ≈ ρ*/3 × Es = 6.43 GPa
  4. σ* ≈ ρ*/3 × σy = 5.67 MPa (upper bound)
  5. A BCC lattice at the same density: E* ≈ 0.01 × 193 = 1.93 GPa

Answer: d ≈ 0.37 mm; E* ≈ 6.4 GPa, 3.3 times a BCC lattice of the same mass

How wrong is the first-order formula? · Simple cubic cell a = 4 mm, strut d = 0.6 mm
  1. d/a = 0.15; first order ρ* = (3π/4) × 0.15² = 0.05301
  2. At each node three struts cross at right angles; the strut volumes overlap by two Steinmetz solids' worth, 8√2 r³ per node
  3. Exact: ρ* = (3π/4)x² − √2 x³ = 0.05301 − 0.00477 = 0.04824
  4. First order overestimates by 9.9%; the error grows in proportion to d/a

Answer: ρ* = 0.0482, not 0.0530 (10% too high)

Chapter 17

Materials, multi-material parts, post-processing and density checks

Which syllabus covers thisAU CME339 Unit IIVTU BME714A Modules 4, 5SPPU 402045C Unit 4MU not in syllabusKTU MET468 Module 1

Which materials can be printed, and in what form?

Polymers: PLA, ABS, PETG and nylon filaments for extrusion, photopolymer resins for vat and jetting, and nylon PA12 powder for SLS (EOS lists its PA 2200 at 48 MPa tensile strength in X and Y). Metals, as powder or wire: stainless 316L and 17-4PH, Ti-6Al-4V, AlSi10Mg, nickel alloys such as Inconel 718, tool steels, cobalt-chrome. Ceramics such as alumina and zirconia, from slurries, binder jetting or resin filled with ceramic, then sintered. Composites: chopped carbon or glass fibre in filament, or continuous fibre laid in with the plastic. The feedstock form (filament, resin, powder, wire, sheet) decides which process family can use a material.

What are shape-memory alloys and 4D printing?

A shape-memory alloy such as NiTi (nitinol) can be deformed and then recovers its original shape when heated, thanks to a reversible change in crystal structure. Printing NiTi by powder bed fusion makes porous implants and actuators that machining cannot. 4D printing adds time as the fourth dimension: the part is designed to change shape after printing when exposed to a stimulus such as heat, water or light. Examples are shape-memory polymers, hydrogels that swell, and laminated strips whose layers expand by different amounts, so a flat print folds itself into a box or a gripper.

What are discrete, porous and blended (graded) multi-material parts?

Discrete: different materials in separate regions with a sharp boundary, such as a rigid PolyJet housing with a rubbery seal, or an FDM part with dissolvable support. Porous: one material with designed porosity, such as a hip-implant stem with a lattice surface that bone can grow into, or a filter. Blended or functionally graded: the composition changes gradually through the part. DED can do this by changing the ratio of two powders layer by layer, for example from stainless steel to a nickel alloy. Grading spreads out the mismatch in thermal expansion and stiffness that would crack a sharp joint.

What post-processing do printed parts need?

Support removal: snapped or cut off in polymers, washed out where the support is soluble, and sawn or wire-EDM cut from the plate for metals. Surface: sanding, bead blasting, tumbling and polishing; acetone vapour smooths ABS by melting its skin; critical faces and holes are machined. Thermal property enhancement: UV post-cure for resin, stress relief before metal parts leave the plate, solution and ageing heat treatment, and hot isostatic pressing (HIP), which closes internal pores with high-pressure gas at high temperature. Non-thermal: infiltration, coatings, shot peening. Parts can also serve as patterns for casting or moulding.

Where do dimensional and surface errors in AM come from?

From the data: STL chord error and the staircase effect of slicing. From the machine: positioning resolution, backlash, beam spot or nozzle size, and a badly levelled bed. From the material and process: shrinkage on cooling or curing, warping and residual stress, over- or under-extrusion, overcure in resin, powder spreading defects, and porosity from too little or too much energy. From post-processing: support scars, distortion when parts are cut from the plate, and sintering shrinkage. Each source is cut by its own fix: finer tessellation, adaptive layers, calibration, a smart choice of orientation, and scale factors.

How is the density of a printed metal part measured?

By Archimedes' principle: weigh the part in air (mₐ), then hanging fully under water (apparent mass m_w). The displaced water weighs mₐ − m_w, so the volume is (mₐ − m_w)/ρw and the density is mₐρw/(mₐ − m_w). Use the water density at the bath temperature: 0.99821 g/cm³ at 20 °C and 0.99705 at 25 °C (NIST). Relative density is the result over the fully dense value (typical: 316L 8.0, Ti-6Al-4V 4.43, AlSi10Mg 2.67 g/cm³). Dense powder-bed parts usually aim at 99% or more. Pores open to the surface fill with water and hide, so seal the part first or section it.

Archimedes density check: part density, relative density and porosity of a printed metal part: interactive loads as you scroll
Predict, then check

A 316L part weighs 20.000 g in air and 17.480 g in water at 22 °C (ρw = 0.99777 g/cm³). About how porous is it, taking 8.0 g/cm³ as fully dense?

Worked examples

Archimedes density of a titanium part · Ti-6Al-4V part: 12.500 g in air, 9.665 g in water at 24 °C (ρw = 0.99730 g/cm³); fully dense 4.43 g/cm³ (typical)
  1. Displaced water: 12.500 − 9.665 = 2.835 g, so V = 2.835/0.99730 = 2.8427 cm³
  2. ρ = 12.500/2.8427 = 4.3973 g/cm³
  3. Relative density = 4.3973/4.43 = 99.26%, porosity 0.74%
  4. Taking ρw = 1 would give 4.4092 g/cm³, 0.27% too high, as big as the porosity being measured

Answer: ρ = 4.397 g/cm³, 99.26% dense: passes a 99% bar

A porous aluminium cube · AlSi10Mg cube: 2.612 g in air, 1.610 g in water at 20 °C (ρw = 0.99821 g/cm³); fully dense 2.67 g/cm³ (typical)
  1. V = (2.612 − 1.610)/0.99821 = 1.0038 cm³
  2. ρ = 2.612/1.0038 = 2.6021 g/cm³
  3. Relative density = 2.6021/2.67 = 97.46%, porosity 2.54%
  4. Below 99%: check the energy density (lack of fusion) and section the cube to see the pore shape

Answer: 97.5% dense (2.5% porosity): reject or adjust parameters

What HIP does to the volume · A 50 g 316L part at 1.5% porosity is hot isostatically pressed to full density (8.0 g/cm³); assume all pores are internal
  1. Before: V = 50/(0.985 × 8.0) = 6.3452 cm³
  2. After: V = 50/8.0 = 6.2500 cm³, a 1.5% volume drop
  3. Linear change ≈ 1 − 0.985^(1/3) = 0.50% in each direction
  4. So HIP shifts dimensions by about half a percent: finish-machine critical faces after HIP, not before

Answer: Volume 6.345 → 6.250 cm³; dimensions shrink about 0.5%

Chapter 18

Choosing a process and costing a part

Which syllabus covers thisAU CME339 Unit IVTU BME714A Modules 4, 5SPPU 402045C Units 1, 4MU not in syllabusKTU not in syllabus

How do you choose an AM process for a part?

First screen out processes that cannot meet hard requirements: the material (a nylon snap-fit rules out metal PBF, a titanium bracket rules out FDM), the size against the build volume, the smallest feature and tolerance, and any certification needed. Then rank the survivors on the softer criteria with a weighted decision matrix: give each criterion a weight that sums to 1, score each process from 1 to 5, and add weight × score. Typical criteria are material properties, accuracy and surface, cost per part at the quantity needed, lead time, and post-processing effort. Finally, compare the winner with machining or moulding at the same quantity.

How is the cost of a printed part built up?

Machine hourly rate = (purchase price/life + yearly maintenance) / productive hours a year, where productive hours = available hours × utilisation. A build of n parts costs rate × build time + material (part mass × (1 + waste for supports or unrecoverable powder) × price per kg) + labour (set-up and unpacking per build, finishing per part). Divide by n for the cost per part. Machine time dominates for metal; material and finishing matter more for polymers. Utilisation is the lever most often forgotten: a machine idle half the year doubles its hourly rate.

Cost of a printed part: machine hourly rate, material, labour and how parts per build drive the cost per part: interactive loads as you scroll

Why does filling the build lower the cost, and where does it stop?

Some of each build's cost is fixed whatever the number of parts: warm-up and cool-down (hours on a polymer SLS machine), recoating layers that are mostly empty, plate preparation, and unpacking. Packing more parts into the build spreads this fixed share. The cost falls towards a floor set by what each part needs on its own: its scan or print time, its material and its finishing. That is why service bureaus nest jobs from many customers into one SLS build, and why a single small part on a big metal machine is expensive.

What does production planning for AM involve?

Planning turns orders into builds. Group parts by material and layer thickness. Orient and nest them for the best packing and the least support. Estimate each build's time and slot long builds to run nights and weekends. Book post-processing capacity (depowdering, heat treatment, support removal, machining, inspection) so finished builds do not queue. Plan powder stock and its refresh, and keep a record of each build (machine, powder lot, parameters) for traceability. Software helps at each step: build preparation, scheduling and machine monitoring.

What is lifecycle costing, and why does it favour AM in some cases?

Lifecycle cost adds everything a part costs over its life: design and tooling, production, inventory, operation (fuel or energy it causes to be used), maintenance and spares, and disposal or recycling. A printed part can cost more to make and still win over its life. A lighter bracket saves fuel on every flight, a consolidated part needs fewer spares and inspections, and digital inventory (printing spares on demand instead of stocking them) removes warehousing and the cost of tooling kept for old models. Compare options on the total over the life, discounted to today, not just the purchase price.

Predict, then check

An SLS build of 30 nylon parts costs ₹1,159 per part (illustrative preset). If 60 parts are packed into the build instead, what is the cost per part?

Worked examples

Machine hourly rate · SLS system ₹15,000,000, written off over 7 years, maintenance 10% of price per year, 6,000 h/year available, 60% utilisation (illustrative values)
  1. Depreciation = 15,000,000/7 = ₹2,142,857 per year
  2. Maintenance = 0.10 × 15,000,000 = ₹1,500,000 per year
  3. Productive hours = 6,000 × 0.6 = 3,600 h
  4. Rate = (2,142,857 + 1,500,000)/3,600 = ₹1,011.90 per hour
  5. At 30% utilisation the rate doubles to ₹2,024/h

Answer: About ₹1,012 per machine hour

Cost per part for 30 and 60 parts per build · Rate as above; build time 10 h + 0.3 h per part; 30 g parts, 50% powder waste, ₹7,000/kg; labour ₹2,500 per build + ₹120 per part
  1. n = 30: build 19 h; machine 1,011.9 × 19/30 = ₹640.9
  2. Material 0.03 × 1.5 × 7,000 = ₹315; labour 2,500/30 + 120 = ₹203.3
  3. Cost per part = ₹1,159.2
  4. n = 60: build 28 h; machine ₹472.2 + material ₹315 + labour ₹161.7 = ₹948.9
  5. Check by a year of builds: 3,600/19 = 189.5 builds give the same ₹1,159.2

Answer: ₹1,159 per part at 30, ₹949 at 60 (18% less)

Weighted decision matrix · Nylon housing, 50 off. Weights: material 0.3, accuracy 0.2, cost 0.3, lead time 0.2. Scores 1-5 (illustrative): FDM 2, 2, 5, 5; SLS 5, 4, 4, 3; SLA 3, 5, 3, 4
  1. FDM = 0.3×2 + 0.2×2 + 0.3×5 + 0.2×5 = 3.5
  2. SLS = 0.3×5 + 0.2×4 + 0.3×4 + 0.2×3 = 4.1
  3. SLA = 0.3×3 + 0.2×5 + 0.3×3 + 0.2×4 = 3.6
  4. SLS wins; check how sensitive the result is by changing the weights, because the scores are judgements

Answer: SLS (4.1) ahead of SLA (3.6) and FDM (3.5)

Chapter 19

Inside a 3D printer: frame, motion, extruder, firmware and calibration

Which syllabus covers thisAU not in syllabusVTU not in syllabusSPPU 402045C Unit 5MU not in syllabusKTU not in syllabus

What holds a printer together and moves the head?

The frame (aluminium extrusion, sheet steel or cast parts) must be stiff, because any flex shows up as ringing and layer shift. Motion layouts: Cartesian, where one motor per axis often moves the bed in Y; CoreXY, where two motors share crossed belts to move a light head in X and Y; and delta, where three arms carry the head. Linear rods or rails guide the moving parts. The build plate, often a heated bed with a removable spring-steel sheet, keeps the first layers stuck and stops warping. An enclosure holds the air warm for ABS and nylon. The bed must be levelled or probed before printing.

How do the extruder, hot end and nozzle work?

The extruder is a stepper-driven gear that grips the filament and pushes it. Mounted on the head, it is a direct drive: precise, and good for flexible filaments. Mounted on the frame, it pushes through a PTFE tube (Bowden), which makes the head lighter. The hot end melts the filament: a heater cartridge and a thermistor in an aluminium block, a heat break that stops heat creeping up, and a fan-cooled heatsink. The brass or hardened-steel nozzle (0.4 mm is a typical size) sets the line width. A part-cooling fan freezes each line as it lands. A sensor can detect when the filament runs out.

How does the firmware turn millimetres into motor steps?

A typical hybrid stepper makes 200 full steps per turn (1.8°). The driver splits each step into microsteps (1/16 is common) for smoother motion. Steps per mm = (steps per rev × microsteps)/(travel per motor turn). For a GT2 belt (2 mm pitch) on a 20-tooth pulley that is 200 × 16/(2 × 20) = 80. For an 8 mm-lead screw it is 400. For an extruder, one turn feeds π × the drive gear's effective diameter, divided by any gear reduction. Calibrate by commanding a move, measuring it, and scaling: new = old × commanded/measured. Marlin takes the value with M92.

3D printer motion: steps per mm for belts, lead screws and extruders, resolution, and calibrating with M92: interactive loads as you scroll

What extra hardware does a laser powder-bed machine have?

Optics: a fibre laser, a beam expander, two galvanometer mirrors that steer the spot across the bed, and an f-theta lens that keeps the spot focused flat on the bed. Gas: the chamber is flooded with argon or nitrogen to keep oxygen low, and a steady gas flow across the bed blows away spatter and fumes into a filter. Powder handling: a dosing piston or hopper, a recoater blade or roller, overflow bins, and a sieve to recycle unused powder. Sensors: oxygen sensors, bed and chamber temperatures, cameras on each layer and melt-pool monitors. A control computer coordinates the scanner, recoater and gas.

What do the slicer, G-code and firmware each do?

A slicer such as Cura cuts the model into layers. It chooses the slicing (uniform or adaptive layer heights), the walls and the infill pattern (lines, grid, triangles, honeycomb, gyroid) and writes G-code. G-codes move the machine: G28 homes, G1 makes a straight move with an extrusion amount E, G92 resets a position. M-codes set the rest: M104 and M109 the hot-end temperature (M109 waits), M140 and M190 the bed, M106 the fan, M92 steps per mm, and M500 saves settings. Firmware such as Marlin plans accelerations, runs PID heater control and stops if a heater runs away. Calibration covers steps, flow, bed levelling and PID tuning (M303).

Predict, then check

A belt axis is set to 80 steps/mm at 1/16 microstepping. The driver is switched to 1/32 without changing the firmware. What happens?

Worked examples

Belt axis steps and resolution · 200-step (1.8°) motor, 1/16 microstepping, GT2 belt (2 mm pitch), 20-tooth pulley
  1. Microsteps per turn = 200 × 16 = 3,200
  2. Travel per turn = 20 teeth × 2 mm = 40 mm
  3. Steps per mm = 3,200/40 = 80
  4. Resolution = 1/80 mm = 12.5 µm per microstep (real accuracy is coarser: microsteps are not perfectly even)
  5. With a 0.9° (400-step) motor: 160 steps/mm

Answer: 80 steps/mm, 12.5 µm per microstep: M92 X80 Y80

Z lead screw and good layer heights · 200-step motor, 1/16 microstepping, lead screw with 8 mm lead
  1. Steps per mm = 3,200/8 = 400
  2. One full step moves 8/200 = 0.04 mm
  3. Layer 0.20 mm = 5 full steps exactly; layer 0.15 mm = 3.75 full steps, which ends on a microstep
  4. Layers that are whole multiples of 0.04 mm (0.12, 0.16, 0.20, 0.24) end on full steps, where the motor holds position most precisely
  5. A 2 mm-lead screw would give 1,600 steps/mm

Answer: 400 steps/mm; prefer layer heights in steps of 0.04 mm

Calibrating the extruder · E steps/mm set to 93. Mark the filament 120 mm above the extruder, command G1 E100 (slowly, hot end at temperature); the mark is then 23.5 mm from the extruder
  1. Filament actually fed = 120 − 23.5 = 96.5 mm for 100 mm commanded
  2. New value = 93 × 100/96.5 = 96.37 steps/mm
  3. Send M92 E96.37, then M500 to store it
  4. Repeat the test: the mark should now end 20 mm from the extruder

Answer: E = 96.37 steps/mm (M92 E96.37)

Chapter 20

Rapid tooling, direct digital manufacturing and applications

Which syllabus covers thisAU CME339 Unit IVTU BME714A Module 5SPPU 402045C Unit 1MU MEC404 Module 5KTU MET468 Module 5

What are rapid prototyping, rapid tooling and rapid manufacturing, direct and indirect?

Rapid prototyping makes models to look at and test: concept models, form-and-fit checks and functional prototypes. Rapid tooling makes tools: mould inserts, dies, jigs, fixtures and gauges. Rapid manufacturing, or direct digital manufacturing, makes end-use parts. Direct means the printer makes the tool or part itself, such as a printed mould insert or a printed bracket. Indirect means the print is a master or pattern from which the real item is made: an SLA master for a silicone mould, a wax or resin pattern for investment casting, or a binder-jetted sand mould for a casting. Indirect routes reach materials no printer can.

Why do printed conformal-cooling inserts shorten moulding cycles?

Cooling is usually the longest part of an injection-moulding cycle. Drilled channels must be straight, so curved or deep parts have hot spots that set the cycle. A printed insert can carry channels that follow the part at a constant depth, keeping the wall cooler and more even. Cooling time t = s²/(π²α) · ln((4/π)(Tm − Tw)/(Te − Tw)) is the time for the centre line to reach the ejection temperature, so every degree off the hottest wall helps. One open-access study cut cooling time by up to 8%; earlier work it cites reached 28%, with cycle savings of 15-60% depending on complexity.

Conformal cooling: injection-moulding cooling time and cycle time with a 3D-printed mould insert: interactive loads as you scroll

How do Align Technology and the hearing-aid makers use direct digital manufacturing?

Both mass-produce parts that are all different. Align's annual report for 2007 describes using stereolithography to build a series of moulds of each patient's teeth, then pressure-forming plastic sheet over each mould to make Invisalign aligners. In 2018, 3D Systems said Align uses its SLA printers in a workflow producing more than 320,000 unique aligners a day; the printed moulds are sacrificial tooling. For hearing aids, Sonova (Phonak) began digital manufacturing of custom in-the-ear shells in 2001/2002: the ear impression is laser-scanned and the shell is built layer by layer in light-cured acrylic. VTU also names Siemens as a hearing-aid example; a trade report says the rest of the industry followed Phonak to 3D printing in under two years.

How do aerospace, automotive and medical sectors use AM?

Aerospace and defence: lightweight, topology-optimised brackets, consolidated ducts and fuel parts in titanium and nickel alloys, and spares printed on demand for old or remote equipment. Automotive: fast prototypes, jigs, fixtures and assembly aids on the shop floor, conformal-cooled tooling, and low-volume or custom parts for motorsport and premium models. Medical and dental: patient-specific implants with porous surfaces for bone ingrowth, surgical guides drilled to each patient's scan, anatomical models for planning operations, crowns, and moulds for aligners. In every case the value comes from customisation, complexity or speed, not from cheap volume.

What are building, bio, food and electronics printing?

Construction printing extrudes concrete or mortar from a gantry or robot arm to build walls layer by layer on site, saving formwork. Bio-printing deposits bio-inks (hydrogels carrying living cells) or prints porous scaffolds that cells colonise, aimed at tissue models and, in research, implants. Food printing extrudes pastes such as chocolate, dough or purées into shapes that cannot be moulded. Electronics printing uses direct write and inkjet to lay down conductive tracks, antennas and sensors, even onto curved parts. Together with digital inventory and distributed production, these fit the Industry 4.0 idea of manufacturing driven straight from data.

Predict, then check

A moulded PP lid's wall is thickened from 2 mm to 3 mm. Roughly how does the cooling time change?

Worked examples

Conformal against drilled cooling · PP plate, wall 2.5 mm, α = 0.096 mm²/s, melt 230 °C, mould set at 50 °C, ejection 90 °C (illustrative values); drilled channels leave a hot spot at 65 °C; fill, pack and open/eject take 8 s
  1. Conformal: t = 2.5²/(π² × 0.096) × ln((4/π) × 180/40) = 6.596 × 1.7456 = 11.52 s
  2. Drilled (hot spot at 65 °C): ln((4/π) × 165/25) = 2.1286, t = 14.04 s
  3. Cycles: 22.04 s drilled, 19.52 s conformal (11.5% shorter)
  4. Parts per 24 h, one cavity: 3,920 → 4,427
  5. Check: summing 200 terms of the slab Fourier series gives the same cooling time

Answer: Cooling 14.0 s → 11.5 s; 507 more parts a day per cavity

A thicker PC part in a two-cavity tool · PC, wall 3 mm, α = 0.144 mm²/s, melt 300 °C, mould 90 °C, ejection 125 °C (illustrative); drilled hot spot 10 °C hotter; 10 s fill, pack and open/eject; 2 cavities
  1. Conformal: t = 9/(π² × 0.144) × ln((4/π) × 210/35) = 12.88 s
  2. Drilled: t = 9/(π² × 0.144) × ln((4/π) × 200/25) = 14.70 s
  3. Parts per day = 2 × 86,400/cycle: 6,997 drilled, 7,554 conformal

Answer: About 557 more parts per day from the same tool

Printers needed for custom shells · A lab needs 600 custom hearing-aid shells a day; a printer builds about two dozen shells an hour (the rate reported for Sonova's EnvisionTEC Perfactory printers); it runs 16 h a day
  1. Shells per printer per day = 24 × 16 = 384
  2. Printers = ⌈600/384⌉ = 2
  3. Each shell is different, but the printer does not care: this is mass customisation, where AM beats moulding outright

Answer: 2 printers (768 shells a day of capacity)

Keep going

Recommended reading

  • Additive Manufacturing Technologies, I. Gibson, D. Rosen, B. Stucker, M. Khorasani
  • 3D Printing and Additive Manufacturing: Principles and Applications, C. K. Chua, K. F. Leong
  • Understanding Additive Manufacturing, A. Gebhardt
  • Rapid Prototyping and Manufacturing, P. F. Jacobs
  • Additive Manufacturing, A. Bandyopadhyay, S. Bose (eds.)
  • ISO/ASTM 52900, Additive manufacturing: general principles and terminology

How this book was made

Chapters follow the official syllabi listed above. Every interactive's formulas and every worked-example number were checked by script. Explanations, animations and examples are original to Skill-Lync. Topics informed by MIT OpenCourseWare 2.008 (Design and Manufacturing II (lectures on layered and additive manufacturing)), CC BY-NC-SA. All explanations, animations and examples are original to Skill-Lync; MIT does not endorse this page.

From the exam to a core job

Talk to a counsellor.

Ask how design and CAE engineers use additive manufacturing on real projects, and what the Additive Manufacturing and 3D Printing using Ansys SpaceClaim covers.

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.

Talk to a career counsellor

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.