---
title: "Seam Tracking for Robotic Welding: Sensors, Setup & Accuracy"
url: https://www.intouchray.com/seam-tracking-for-robotic-welding-sensors-setup-accuracy/
date: 2026-11-08
modified: 2026-11-03
lang: en
author: "Allan Hill"
description: "Seam Tracking in Robotic Laser Welding: When It Pays Off, and When It Doesn't A seam is the joint two pieces of metal form when they meet, and in robotic..."
categories:
  - "Laser Welding Machine"
image: https://www.intouchray.com/wp-content/uploads/2026/11/intouchray-bd96220e.jpg
word_count: 1605
---

# Seam Tracking for Robotic Welding: Sensors, Setup & Accuracy

# Seam Tracking in Robotic Laser Welding: When It Pays Off, and When It Doesn't

A seam is the joint two pieces of metal form when they meet, and in robotic laser welding, the weld head has to follow it exactly. Seam tracking is the sensor-and-software loop that finds that joint line, measures how far the real part has drifted from the programmed path, and corrects the robot's motion while the beam is on. It matters most when parts arrive with variation a fixture cannot absorb — and it matters least when your tooling and upstream processes are already tight.

We write this from the factory floor at Intouch, where we have built fiber laser systems since 2001 — cutting, cleaning, cladding, and robotic welding cells shipped to more than 80 countries. Buyers ask us about tracking on almost every automation project, so here is how we explain it in the sampling room.

## Why the Joint Line Moves in the First Place

![Seam Tracking for Robotic Welding: Sensors, Setup & Accuracy](https://www.intouchray.com/wp-content/uploads/2026/11/intouchray-ac23cc5e.jpg)

Before talking sensors, look at the parts. A robot follows a taught path to within a few hundredths of a millimetre; positioning accuracy on well-built machines stabilises around ±0.03 to ±0.05 mm per industry data. The problem is almost never the robot. It is the joint.

Stamped panels spring back differently from batch to batch. Extrusions vary in cross-section along their length. A part that went through a hot process may have distorted by a fraction of a millimetre — enough to push the joint off the programmed line. Laser beams are narrow. A welding spot measured in fractions of a millimetre leaves almost no tolerance for a joint that wanders. TIG or MIG arc welding forgives drift because the arc is broad; a focused 1070 nm fiber beam does not [1].

That is the whole argument for seam tracking in one paragraph: the beam is precise, the world is not.

## How Seam Tracking Actually Works

Every tracking system does the same three things in sequence: detect the joint, measure the offset, correct the path. The differences between vendors come down to how fast each step happens and what sensors do the measuring.

### Finding the Seam Before the Arc-On

Most cells run a pre-scan. Before welding, a laser triangulation sensor — or in some setups a vision camera — sweeps along the programmed path. It projects a line of light onto the surface, reads the profile through the camera, and locates the geometric signature of the joint: the V-groove of a butt weld, the step of a lap weld, the corner of a fillet. The system then compares the found joint position to the taught path and shifts the program to match. On parts with repeatable distortion, one scan per part is enough.

### Watching While the Beam Is On

Through-the-lens tracking goes further: the sensor looks through the welding optics themselves, sampling the joint profile a few millimetres ahead of the melt pool in real time. Because it measures during welding, it catches variation the pre-scan could never see — thermal distortion that builds as heat goes in, for instance. Update rates on modern triangulation sensors run into the kilohertz range (manufacturer data), fast enough to steer the beam within one spot diameter.

### Turning an Offset into a Correction

Raw sensor data is a cloud of profile points. Software classifies the joint type, extracts the joint centre, computes the vector from that centre to where the robot currently thinks it is, and feeds a correction command into the motion controller. Done well, this loop runs continuously with no operator involvement. Done badly — wrong joint type selected, reflectivity confusing the sensor — it chases edges instead of the joint and makes things worse. That second case is real; we have seen cells misconfigured to track a sharp edge rather than the groove, and the weld line followed it faithfully in the wrong direction.

## Where Tracking Helps — and Where Fixturing Does the Job

Here is the point we make to every buyer, and it occasionally disappoints them: seam tracking does not replace good fixturing. It supplements it.

If your parts arrive inconsistent because the stamping die is worn or the saw cut wanders, tracking will follow the joint — but you are now asking an expensive sensor loop to compensate for a cheap upstream failure. Weld quality may recover, but fit-up gaps may still exceed what laser welding can bridge. Fiber laser welding handles small gaps well; anything beyond roughly a tenth of the material thickness starts to need wire feed or process changes. Tracking fixes position error. It does not fix a gap.

Our honest guidance, the same we give across every robotic laser welding project:

- **Tight, repeatable parts** (machined components, laser-cut blanks): skip tracking, invest in solid fixtures. Laser-cut blanks are a good example — if you cut your own parts on a fiber laser cutting machine at ±0.05 mm, the joint lands where the fixture puts it.

- **Stampings, castings, long welds on formed parts**: tracking earns its cost. The pre-scan alone often suffices; add real-time tracking when thermal distortion is in play.

- **High-mix production**: tracking shortens changeover because you re-teach less when a new part variant lands a little differently.

The trade-off said plainly: tracking adds cost, another sensor to calibrate, and another failure mode. Buy it when part variation is real and measured — hold a few parts from production, scan them, and look at the spread before you specify the option. That ten-part study answers the question better than any brochure.

## What Real-Time Monitoring Data Tells You

The same sensor that steers the beam also collects a record. Height profiles, joint position over the weld length, measured offsets — all of it logs per part. On a monitored cell, that data reveals patterns no operator would catch:

- a fixture that has gradually shifted, showing up as a slow offset trend across a shift;

- a batch of parts arriving consistently 0.2 mm off, pointing at the upstream supplier, not your cell;

- thermal distortion on a long weld, visible as the joint drifting as heat accumulates — which tells you where to add clamping or change weld sequencing.

Per ISO 9001-style quality management thinking, this traceability is half the value. When a customer questions a weld, you can show the measured joint position for that serial number, not just a pass/fail flag.

## The System View: Tracking Is One Link in the Chain

Weld accuracy in a robotic laser welding cell is a chain, and tracking is only one link. The robot's calibration, the rigidity of the fixture, the stability of the laser source's output power, the optics' condition, even the shielding gas flow — each contributes. A perfect tracking loop on a cell with an uncalibrated robot produces perfectly tracked, slightly wrong welds.

This is why we at Intouch build welding heads, motion, and control as one engineered system rather than assembling third-party boxes. Our production lines cover over 100 machine models across cutting, welding, cleaning, and cladding, certified to ISO 9001:2015 with CE on multiple product series — and when we integrate a tracking option on a robotic laser welding cell, we commission the whole loop together: sensor, robot, welding head, and the laser welding equipment parameters that go with it. The same logic applies whether the cell pairs with a handheld laser welder for touch-up work or runs fully unattended.

The question buyers ask us most at trade fairs is not whether tracking exists — everyone's brochure says it does — but whether their specific parts need it. Bring sample parts. That is what our sampling room is for, and a trial weld on your actual joint geometry settles it in an afternoon.

## FAQ

**Does every robotic laser welding cell need seam tracking?**
No. If parts are consistent and fixtures are rigid, a taught path holds tolerance comfortably. Tracking pays when measured part variation exceeds roughly half a spot diameter, or when thermal distortion shifts the joint during welding.

**What sensor types are used for laser seam welding tracking?**
Laser triangulation is the standard: a projected light line read by a camera, either on a separate pre-scan scanner or integrated through the welding optics. Vision-based systems handle some visible joints; tactile probing is obsolete in laser work.

**Can seam tracking close a fit-up gap between parts?**
No, and this is the most common misconception we correct. Tracking corrects the path's position, not the gap width. Wide gaps need tighter upstream tolerance, wire feed, or a change in joint design.

**How much does adding tracking cost on a laser welding cell?**
It varies with sensor grade and whether real-time through-the-lens tracking or pre-scan only is specified. As a seam price reference, pre-scan tracking is a modest option; integrated real-time systems command a clear premium. Quote it against your measured part variation, not against fear.

**How do I evaluate a seam tracking supplier before buying?**
Ask them to weld your parts, not theirs. Check what joint types their classification supports, what the update rate is, and how offsets are logged. A laser welding manufacturer who cannot show a live correction on your sample geometry is selling you a feature list.

Tracking is a tool for a specific problem — part variation. Diagnose first, then buy. If you want that diagnosis done on your own parts, write to us at info@intouchray.com or visit www.intouchray.com; we will run the trial and show you the numbers.

---

*Sources: [1] fiber and disk laser wavelength data, approximately 1070 nm (manufacturer data); positioning accuracy ranges per published fiber-laser machine specifications (industry data).*

**Related reading:** [laser applications by industry](https://www.intouchray.com/applications/) · [robotic laser welding machines](https://www.intouchray.com/robotic-laser-welding-machine/) · [laser welding machines](https://www.intouchray.com/category/laser-welding-machine/)