Advances in CNC robot welding machines: what changed on the shop floor

A CNC robot welding cell is a six-axis arm under numerical control, with the weld path, torch angle, travel speed, wire feed and current all defined in a program rather than by hand. The 'CNC' part matters less than people expect. The real shift is that the same program drives the positioner, the fixture clamps and the torch, so the part arrives at each weld in a known orientation.

That matters for repeatability. A manual welder adjusts by eye and by feel. A programmed cell repeats the same torch angle on part 1 and part 400, provided the joint gap stays inside tolerance. Most welding defects we see on incoming assemblies trace back to gap variation, not to the robot.

For our customers, robot welding usually pairs with CNC machining rather than replacing it. A welded frame still needs machined faces, bores or bosses so it can bolt to something else. Welding and machining are two operations on one part, and the sequence you pick decides whether the part is accurate or scrap.

Seam tracking with laser or arc sensing is the biggest practical change. The cell measures the joint in real time and corrects the path as it welds. On a 1,200 mm frame with a 0.5 mm gap variation, that correction keeps the weld centred instead of drifting to one side. It does not fix a bad fit-up, but it absorbs normal variation from laser-cut or plasma-cut edges.

Offline programming has also matured. We build the cell model and the fixture model in software, generate the path, then verify it before the first part is cut. For a new assembly this removes most of the teach-by-hand time, which used to be the reason short runs were uneconomical.

Torch and power supply control has moved in the same direction. Synergic programs set voltage and current from wire feed speed and material thickness, so a change from 6 mm to 10 mm plate does not require a full re-tune. Pulse and dip-transfer modes let the cell weld thinner sheet with less heat input than a standard spray arc.

None of this removes the need for a weld procedure. We still qualify the procedure, cut and etch a sample, and check penetration before running a batch.

A robot welds what it is given. If the gap is 1.5 mm on a 6 mm fillet, the weld either bridges it or burns through, and neither is consistent. For robot cells we ask for a gap of 0 to 0.5 mm on a fillet joint, with the edges square or with a small root face. That is a tighter fit-up than many manual shops work to.

Laser-cut and waterjet-cut edges usually hold that. Plasma-cut edges at 10 mm thickness often do not, so we either machine the edge or accept a wider gap and a slower, hotter weld. The decision is cost, not preference.

Tack placement matters too. Tacks should sit inside the weld path or be ground flush before the robot runs, because a proud tack changes the arc length and the penetration at that spot. On assemblies with a lot of tacks, we often tack in a fixture, then let the cell run the full seam.

Access is the other constraint. A torch needs roughly 15–20 mm of clearance to reach a fillet at the correct angle. If two plates form a closed corner, the torch cannot get in, and that joint goes back to manual or needs a design change.

• 1
  Fillet gap0 to 0.5 mm for consistent robot welds on 3–10 mm plate
• 2
  TacksGrind flush or place inside the weld path
• 3
  Torch accessAbout 15–20 mm clearance for a correct fillet angle
• 4
  Edge conditionLaser or machined edges hold fit-up; plasma edges often do not

Welding moves metal. A 1,000 mm steel frame can pull 0.5 to 2 mm out of flat after cooling, depending on heat input and how restrained it was in the fixture. If a drawing calls for a machined face with ±0.1 mm flatness across that frame, you cannot weld it last and hold the number.

The usual answer is stress relief, then machining. For steel we often specify a stress-relief cycle after welding, then a finish pass on the critical faces, bores and bolt patterns. On aluminium, stress relief is less effective, so we control heat input, sequence the welds to balance shrinkage, and leave 0.5 to 1 mm of stock on faces that will be cut after welding.

We hold ±0.005 mm on machined features after welding when the sequence is right. That number applies to the machined feature, not to the welded assembly as a whole. Weldment flatness and hole position are separate tolerances and should be called out separately on the drawing.

The practical rule: tolerances tighter than ±0.1 mm belong on a machined surface, and the drawing should say which surfaces those are.

• 1
  Steel weldmentsStress relieve, then machine critical faces
• 2
  Aluminium weldmentsBalance weld sequence; leave 0.5–1 mm stock on cut faces
• 3
  Drawing practiceSeparate weldment tolerance from machined-feature tolerance

Mild steel and 4130/4140 weld predictably and take a post-weld cut well. 4130 and 4140 need preheat and controlled cooling on thicker sections, otherwise the heat-affected zone hardens and cracks. We check hardness after welding on these grades when the part sees load.

Stainless 304 and 316L weld cleanly but move more than mild steel because of higher thermal expansion. On thin stainless enclosures, we back-purge and use a low-heat pulse mode, and we expect some distortion that has to be machined or straightened. 17-4PH can be welded but needs a solution-treat and age cycle if you want the full strength back.

Aluminium 6061 welds with 4043 or 5356 filler and loses roughly half its strength in the heat-affected zone unless it is re-aged. For structural aluminium frames, 5083 or 6082 is often the better choice. On 6 mm and thinner aluminium, a backing bar or chill block helps control distortion and burn-through.

Titanium and Inconel are weldable but the process is a different conversation: full inert shielding, qualified procedure, and usually a machined joint preparation rather than a sheared edge.

Incoming material is checked against the cert before anything is cut. After welding, we inspect the assembly before it goes to machining, because a bad weld found after a 4,000 mm frame is machined is an expensive scrap.

Weld inspection runs visual first, then dimensional on the fixture. For load-bearing joints we cut and etch a sample from the setup piece to confirm penetration and check for lack of fusion. Dye penetrant is used on stainless and aluminium joints where surface cracks matter.

Post-machining, we inspect the machined features at 100% before shipment and can supply reports on request. CMM reports, material certs and hardness data go out with the parts when the drawing calls for them.

If a customer needs a specific weld standard, we work to the drawing and the referenced procedure rather than a house default.