By Zhao Liang, Senior Applications Engineer · EVST Cobot Applications Team · · Reviewed by EVST robotics integration engineering
Welders are hard to hire, harder to train — and three months after the welder lands, they’re gone. The pain is sharpest on multi-variant, small-batch work. Drag-to-teach is the other path: the operator holds the torch, traces the seam by hand, and a collaborative robot records the path. No G-code, no teach pendant, no programmer call-out. The welding parameters — arc current, gas, weave width, crater fill — sit in a pre-built process library; the operator picks one from a menu. That library encapsulates the senior welder’s craft and scales it from one person to a full line.
Key takeaways
- 3-minute teach time per new seam by hand-guided drag-to-teach.
- Process library = pre-built menu of welding parameters (arc, gas, weave, crater fill).
- Fenceless cobot with force limiting and collision detection per ISO/TS 15066.
- Minute-scale changeover: drag path + pick process + quick-change fixture.
- Fits multi-variant small-batch welding (3C, appliance, motorcycle, custom metal furniture).
- Industrial arm guarded under ISO 10218; operator qualified under ISO 14732.
This article is for welding line designers, production managers and engineers on multi-variant small-batch welding work who face the welder-shortage / training-cost problem. It covers drag-to-teach cobot welding with a process-library architecture; it does not cover high-cycle production welding, which needs industrial six-axis behind a safety light curtain.
The welder shortage problem
Across most industrial economies, certified welders are aging out faster than they are being trained. A new hire takes 6–18 months to reach production-grade quality; turnover within the first 12 months runs 30–50% in the worst regions. For multi-variant small-batch shops — where a senior welder also needs to reprogram for each batch — the problem compounds: the certified welder is the bottleneck twice (welding speed and programming speed).
Drag-to-teach addresses both bottlenecks by separating the path from the process. The path is taught by the operator in minutes by hand. The process — arc current, shielding gas, weave width, crater fill — is captured once from the senior welder and stored as a reusable parameter set the operator picks from a menu.
When drag-to-teach earns its premium — and when it doesn’t
The decisive factors are batch size, variant count and cycle time. Use this frame:
| Choose drag-to-teach when… | Reconsider when… |
|---|---|
| Batches are small and variants are many | Single high-volume SKU, run for months |
| Operator labor cost > welder-programmer labor cost | Programmer cost is already amortized |
| Cycle time per seam > 30 seconds | Sub-3-second hard-cycle production rate |
| Parts share a process library (e.g., 3C, appliance) | Each part needs a unique process |
| Compliance benefits from fenceless layout | Safety case requires hard fencing |
EVST scopes a drag-to-teach line with what our engineers call the Path-Process Split method: design the path as a per-part hand-taught artifact, and the process as a senior-welder-built menu — because mixing the two locks you back into needing the welder for every batch.
Manual vs hard-programmed robot vs drag-to-teach
| Factor | Manual welding | Hard-programmed industrial | Drag-to-teach cobot |
|---|---|---|---|
| Per-seam programming time | None (welder welds) | Hours (teach pendant) | Minutes (hand-drag) |
| Parameter consistency | Welder skill | Stored program | Stored process library |
| Fencing capex | None | Required | None (ISO/TS 15066) |
| Operator skill needed | Certified welder | Programmer + operator | Operator only |
| Best fit | Single artisan part | High-volume single SKU | Multi-variant small batch |
How an EVST drag-to-teach line is built
| Layer | What EVST does | Why it matters |
|---|---|---|
| Hardware — cobot | Collaborative robot rated to ISO/TS 15066 | Fenceless operation, operator shares space |
| Hardware — torch | Standard MIG/MAG or TIG torch, weight-matched to cobot payload | Path drag-teach works smoothly |
| Hardware — fixture | Quick-change part fixture (10-second swap) | Changeover stays minute-scale |
| Software — drag-teach | Hand-guided trajectory recording, gravity-compensated | Operator drags torch by hand, cobot records |
| Software — process library | Senior-welder-built parameter menu (per material × thickness × joint) | Reusable craft, not per-batch reprogramming |
| Software — safety | Force limit + collision detection per ISO/TS 15066 risk assessment | Operator stands next to robot, no fence |
In practice the failure we see most is buying drag-to-teach hardware without investing in the process library. The result: the operator drags the path in 3 minutes, then spends 30 minutes fiddling with parameters. EVST scopes the process library as a first-class engineering artifact, built once from senior-welder consultation, then reused across thousands of batches.
The welding process library
A welding process library is a structured menu of welding parameter sets, each tagged by:
- Material — mild steel, stainless, aluminum, others
- Thickness range — e.g., 1.5–3 mm thin sheet
- Joint type — butt, lap, fillet, T-joint
- Position — flat, horizontal, vertical-up, overhead
Each entry holds arc current, voltage, wire-feed speed, shielding gas flow, weave amplitude/frequency, crater fill time, and any other process-specific parameters. The operator picks the closest match from the menu; the cobot loads the entry and welds.
For a typical multi-variant shop, the library reaches 30–80 entries within the first quarter and stabilizes there. Adding new entries takes the senior welder a few hours per entry; reusing them takes the operator seconds.
Standards the cell runs under
- ISO/TS 15066 — Collaborative robot safety; power and force limits for fenceless operation.
- ISO 10218 — Robot safety baseline; ISO/TS 15066 is the collaborative-specific extension.
- ISO 3834 — Welding quality requirements; the WPS sits under one of the three classes.
- ISO 14732 — Welding personnel qualification for fully mechanized and automatic welding; the cobot operator is qualified under this standard, not the senior welder cert.
- AWS D1.1 — Structural Welding Code: Steel; applies to structural cobot welding work.
Where it applies across industries
- 3C enclosures — small metal chassis, brackets, frames.
- Small home-appliance weldments — kitchen electrics housings, water heater inner shells.
- Motorcycle and electric-bike thin-sheet parts — fenders, brackets, exhaust shrouds.
- Custom metal furniture — chairs, tables, racks at low volume per SKU.
- Architectural metalwork — railings, decorative panels, custom brackets.
The common thread: small batches, many variants, manual welding currently dominant.
FAQ
How long does a real teach session take? For a typical seam under 200 mm, the hand-drag teach session takes 1–2 minutes; for a longer or multi-pass seam, 3–5 minutes. The process library pick is a second. Total changeover with quick-change fixture: usually under 5 minutes.
Does drag-to-teach hold accuracy across runs? Yes — the cobot records the taught path as a digital trajectory and welds it identically every time. Path repeatability is at the cobot’s stated spec (typically ±0.1 mm). The hand-taught path captures geometric intent; the cobot executes it with mechanical precision.
What if the part fit-up varies between batches? Real-time seam tracking handles fit-up variation within a part family. For step-change fit-up (very different part), a new path is taught — the same 3-minute process. Process library entries usually carry over without change.
Can a non-welder operator run the cell? Yes — that’s the entire point. The operator is qualified under ISO 14732 (robotic welding operator), not under welder-procedure qualification. The senior welder’s expertise lives in the process library, not in the daily operator’s hands.
What’s the payback vs hard-programmed industrial? On a multi-variant small-batch line with 8+ changeovers per shift, drag-to-teach typically pays back in 9–12 months — driven by recovered programming downtime and the elimination of the dedicated programmer’s labor cost. On single-SKU high-volume work, industrial wins on cycle time.
Is it safe to stand next to a welding robot? With proper ISO/TS 15066 risk assessment, yes — the cobot’s force limit, collision detection and speed limits keep the operator safe even in contact. Welding arc and fume are managed with conventional PPE and fume extraction — the cobot doesn’t change that.
Bringing it into your plant
Drag-to-teach robot welding doesn’t replace the senior welder — it scales the senior welder’s craft from one person to a full line. The path is hand-taught in minutes; the process is library-loaded in seconds. The decision starts with batch size, variant count and cycle time; the engineering that makes it work is in the process library investment, not just the cobot hardware. EVST designs drag-to-teach welding lines with the Path-Process Split method. See our guides to cobot ROI payback, collaborative palletizing, and welding positioner selection, or talk to EVST about scoping a drag-to-teach line.
About the author — Zhao Liang is a Senior Applications Engineer on the EVST Cobot Applications Team, with 10+ years of experience deploying collaborative robots in welding, fastening and packaging lines. He builds process-library architectures with customer senior welders during cell scoping — the senior welder’s craft is captured once, then runs unattended. Reviewed by EVST robotics integration engineering for technical accuracy; figures are typical achievable ranges, not guarantees, and are sized per project. Corrections and updates: see the Last Updated date.