CNC Glass Cutting Explained

• 1
  How CNC glass cutting removes material
• 2
  Scribe and break versus abrasive and laser routes
• 3
  What tolerances and edge quality you can expect
• 4
  When glass is the right material for CNC work
• 5
  Fixturing, coolant and inspection on the shop floor
• 6
  Glass cutting method comparison

Glass is not cut the way aluminium is cut. A milling cutter shears metal and throws a chip. Glass is brittle. Push a hard tool into it and the crack runs wherever it wants, not where the toolpath says. So CNC glass cutting is really a controlled crack problem: the machine decides where the crack starts, in which direction it travels, and how deep it is allowed to go.

Two families of process exist. In the first, a hard tool scores the surface and a second operation bends the sheet so the score propagates through the thickness. In the second, the material is removed by erosion or by thermal energy, and there is no score line at all. Scribe-and-break is faster and cheaper. Erosion and thermal routes handle shapes that no score line can follow.

The CNC part is the same in both cases. A CAD model defines the outline, CAM software converts it to G-code, and the controller drives the axes to that path. On a flat sheet the third axis only sets depth. On a curved cover glass or a glass block, the rotary and tilt axes keep the tool normal to the surface, which is where 5-axis work starts to matter.

One property drives every parameter choice: glass has no plastic zone. Metals yield a little before they break, so a slightly wrong feed just rubs. Glass goes from elastic to fractured in one step. Feeds, depths and coolant are chosen to keep the stress field small, not to maximize removal rate.

Scribe-and-break is the workhorse for flat soda-lime and borosilicate sheet. A carbide or diamond wheel scores a shallow line, usually 10–20 μm deep, and a mechanical or thermal step propagates the crack downward. Straight lines and gentle curves are fast and repeatable. The limit is geometry: the score must stay continuous and the crack must stay near the score, so inside corners and holes need a different operation.

Abrasive waterjet removes material with a jet of water and garnet grit at 3,000–4,000 bar. There is no heat-affected zone and no score line, so any 2D outline is possible, including narrow slots and thick stacks. The trade-off is the edge. The jet exits with a taper and a matte, chipped surface that typically needs grinding or polishing if the edge is visible or load-bearing.

Laser cutting works by local heating and controlled thermal stress. A focused beam or a shaped beam follows the contour and the glass separates along it. The kerf is narrow, the edge is smooth, and no grit or coolant touches the part. Thickness is the constraint: the approach is practical on thin sheet, and thicker sections get slow and prone to micro-cracks near the edge.

Pick by geometry first. If the outline is a simple profile with generous radii, scribe-and-break gives the lowest cost per part. If the outline has tight inside corners, holes or mixed thickness, abrasive jet wins. If the edge must stay pristine and the sheet is thin, laser is the cleanest option.

Cut glass does not hold the same tolerance as a milled metal part. On thin sheet, a scribe-and-break line typically lands within ±0.1 mm to ±0.2 mm on a good day, and the number drifts as the sheet gets thicker. Abrasive waterjet is usually quoted around ±0.2 mm to ±0.3 mm on straight runs and looser on thick sections. Laser on thin sheet can hold tighter than either, closer to ±0.05 mm on short contours.

Edge quality matters more than most drawings admit. A scribed edge carries a shallow micro-crack zone of roughly 50–100 μm. Under tension that zone is a crack starter. If the glass is structural, tempered, or bonded into a frame, the edge needs grinding and often a polish, which adds an operation and a fixture.

Holes and internal features are where the process boundary shows up. A scribed hole needs a separate drilling or jet operation. Waterjet cuts the hole in the same pass but leaves a taper on the wall. A rule of thumb: hole diameter should stay at least equal to sheet thickness, and the hole center should sit at least one thickness away from the outer edge.

Flatness and thickness variation of the incoming sheet set the floor for everything else. If the sheet bows 0.3 mm across a 300 mm span, no toolpath compensates for it. Measure the stock, then set the tolerance, not the other way around.

Glass earns its place when you need optical clarity, chemical inertness, thermal stability or electrical insulation in one part. Sight windows, sensor covers, fluidic channels, insulator plates and display substrates all use it for reasons no polymer matches. In those cases the cutting step is a means to an end, and the edge quality is what decides whether the part survives assembly.

It is the wrong material when the part sees impact, bending or point loads and nobody wants to pay for tempering and edge polishing. Laminated or chemically strengthened glass solves some of that, but the cut edge of strengthened glass cannot be machined after strengthening without losing the surface compression near the edge. Plan the sequence: cut, then strengthen.

Small features are a poor fit. Slots narrower than the sheet thickness, sharp inside corners and thin webs between holes concentrate stress. Round every inside corner to at least half the sheet thickness, keep webs at least one thickness wide, and the part will survive handling, shipping and installation far better.

Mixed-material assemblies are common. A glass window bonded into a machined aluminium frame usually needs the frame tolerance tighter than the glass, because the frame is easier to rework. We machine the metal housing to ±0.005 mm and leave the glass at its natural cut tolerance, then control the fit with the bond gap.

Fixturing is where most glass jobs fail. Clamping glass like metal puts point loads at the clamp and starts cracks that show up two days later. The usual approach is a sacrificial backing board, a vacuum chuck, or a low-modulus pad that spreads the load. Support the full underside, not just the perimeter.

Coolant choice depends on the route. Scribe-and-break is usually dry, with a controlled bend or a thermal assist. Abrasive waterjet uses the water itself as the carrier. Laser routes may use an air or gas assist to clear debris and control the thermal gradient. In every case, once the crack has started, you cannot stop it with coolant, so the setup has to prevent it.

In-process monitoring is mostly acoustic and visual. The sound of a scribe line changes when the wheel dulls. Operators listen for it. On the inspection side, edge chips, micro-cracks and taper are checked with a loupe or a low-power microscope, and critical edges get a dye penetrant or a bend test on a sample.

Final inspection follows the same discipline we use on metal: raw material check, in-process monitoring, and a final check before shipment, with reports on request. Our tolerance capability reaches ±0.005 mm on machined metal features, and glass parts are quoted against their own achievable band rather than that number. For one-off prototypes through 10,000+ part runs there is no minimum order quantity.