“On the robot must be selected without teaching” ‘fully automated welding = the future of competitiveness’ - the anxiety of the manufacturing industry is being infinitely amplified by the marketing rhetoric. As a deep-rooted welding field for more than 20 years practitioners, I was saddened to see: 60% of the customers in the selection of the early stage of the “technology path dependence”, while ignoring the depth of their own process analysis. This article from the essence of the process, three steps to end the “pseudo-needs”, to find the optimal solution.
Welding scene “three-dimensional positioning method”: first know yourself, and then choose the technology
Dimension 1: process complexity - the starting point for determining “intelligence”.
Simple scene (suitable for traditional teaching robots):
✅ Single type of weld (straight line/ring)
✅ Consistency > 95% (e.g. mass production of automotive exhaust pipes)
✅ ≤ 3 material types (carbon steel/stainless steel/aluminum alloy)
✅ Cost Warning: The payback period for such scenarios can be extended by 2-3 times with strong no-tutorials.
Complex scenarios (no teaching value highlights):
✅ Multi-species and small batch (e.g. customized parts for construction machinery)
✅ Workpiece tolerance > ± 1.5mm (real-time correction)
✅ Dissimilar material welding (steel + copper, aluminum + titanium, etc.)
✅ Typical case: after the introduction of a no-demonstration program in an agricultural machinery enterprise, the commissioning time for production changeover was shortened from 8 hours to 15 minutes
Dimension 2: production volume - to calculate the “automation” of the economic accounts
Formula: Break-even point = equipment cost / (single piece of labor savings × annual output)
When the production volume <5000 pieces/year, give priority to collaborative robot + simple teaching
When the output is >20,000 pieces/year and the product life cycle is >3 years, the teaching-free solution is more cost-effective.
Dimension 3: Environmental constraints - the “invisible threshold” of technology implementation
Four major constraints that must be evaluated:
① Workshop dust/oil level (affecting vision system accuracy)
① Workshop dust/oil level (affects vision system accuracy)
② Grid fluctuation range (whether the equipment can work stably under ±15% voltage variation)
③ Spatial accessibility (pipelines/tight spaces require customized robotic arms)
③ Space accessibility (customized robotic arms for pipelines/narrow spaces)
④ Process certification requirements (automotive industry needs to comply with IATF 16949 process specifications)
Process selection of the five “fatal misunderstanding”: to avoid 90% of the customer procurement pit
Myth 1: “Fully automated = completely unmanned”.
Reality: no teaching still need process experts to set quality rules, the blind pursuit of unmanned may lead to a spike in scrap rate
Avoid the pit strategy: require suppliers to provide process parameters debugging interface, retain the key nodes of manual review rights
Myth 2: “The more functions the software has, the smarter it is.”
Truth: Functional redundancy will increase the complexity of operation, a customer purchased “all-in-one” equipment because the operator mistakenly touched the AI button, resulting in batch rework.
Core principle: choose a system that supports modular subscription (e.g., purchase basic positioning functions first, then upgrade as needed).
Myth 3: “Hardware parameters equal actual performance”.
Key indicators disassembled:
Repeat positioning accuracy ± 0.05mm ≠ welding trajectory accuracy (affected by torch deformation, heat input deformation)
Maximum speed 2m/s ≠ effective welding speed (need to consider the acceleration and deceleration process energy stability)
Suggestion: Use the actual workpiece to carry out zigzag trajectory welding, and test the consistency of the depth of fusion at the inflection point.
Myth 4: “One-time investment to end the battle”
Long-term cost list:
Annual fee for software licenses (some vendors charge by number of robots)
Process database update fee (new material adaptation requires the purchase of data packages)
Four Steps to Scientific Decision Making: A Complete Map from Requirements to Landing
Step 1: Digital modeling of the process
Toolkit:
✅ 3D scans of welded seams (to assess trajectory complexity)
✅ Material heat input sensitivity analysis (to determine control accuracy requirements)
✅ Welding process evaluation report (to define certification criteria)
Output: “Digital Portrait of Welding Process” (with 9 dimensions of scoring)
Step 2: Technology Path AB Test
Comparison of program design:
Program A: high-precision demonstration teaching robot + expert process package
Scheme B: Teach-free robot + adaptive algorithm
Test metrics:
✅ First-piece pass rate ✅ Changeover time ✅ Consumables cost/meter welded seam
Step 3: Supplier Capability Penetration Assessment
Soul six-question checklist:
① Can you provide test weldments of the same material? (Rejected generic demo parts)
② Is the algorithm open to process weight adjustment? (Prevent “black box” decision-making)
① Can you provide test weldments of the same material (reject generic demo parts)?
④ Is the after-sales service response time less than 4 hours?
⑤ Does it support acceptance by third-party testing organizations?
⑤ Does it support acceptance by third-party testing organizations?
⑥ Is the sovereignty of data clearly attributed? (Prevent process data from being locked)
Step 4: Small Scale Validation → Rapid Iteration
30-day validation plan template:
Week 1: Basic function acceptance (positioning accuracy, arc stability)
Week 2: Extreme working condition test (large angle climbing welding, strong electromagnetic interference)
Week 3: Production beat challenge (continuous 8-hour full load operation)
Week 4: Cost audit (consumable loss rate, gas consumption comparison)
Conclusion
The end point of welding intelligence is to bring technology back to the essence of the process! When serving a new energy vehicle supplier, we decisively recommended that the robot be retained for the box weld (due to the high consistency of the workpieces), while the non-teaching program was adopted for the shaped joints of the impact beam. This “hybrid intelligence” strategy helped the customer save 41% of the initial investment.
Translated with DeepL.com (free version)
“On the robot must be selected without teaching” ‘fully automated welding = the future of competitiveness’ - the anxiety of the manufacturing industry is being infinitely amplified by the marketing rhetoric. As a deep-rooted welding field for more than 20 years practitioners, I was saddened to see: 60% of the customers in the selection of the early stage of the “technology path dependence”, while ignoring the depth of their own process analysis. This article from the essence of the process, three steps to end the “pseudo-needs”, to find the optimal solution.
Welding scene “three-dimensional positioning method”: first know yourself, and then choose the technology
Dimension 1: process complexity - the starting point for determining “intelligence”.
Simple scene (suitable for traditional teaching robots):
✅ Single type of weld (straight line/ring)
✅ Consistency > 95% (e.g. mass production of automotive exhaust pipes)
✅ ≤ 3 material types (carbon steel/stainless steel/aluminum alloy)
✅ Cost Warning: The payback period for such scenarios can be extended by 2-3 times with strong no-tutorials.
Complex scenarios (no teaching value highlights):
✅ Multi-species and small batch (e.g. customized parts for construction machinery)
✅ Workpiece tolerance > ± 1.5mm (real-time correction)
✅ Dissimilar material welding (steel + copper, aluminum + titanium, etc.)
✅ Typical case: after the introduction of a no-demonstration program in an agricultural machinery enterprise, the commissioning time for production changeover was shortened from 8 hours to 15 minutes
Dimension 2: production volume - to calculate the “automation” of the economic accounts
Formula: Break-even point = equipment cost / (single piece of labor savings × annual output)
When the production volume <5000 pieces/year, give priority to collaborative robot + simple teaching
When the output is >20,000 pieces/year and the product life cycle is >3 years, the teaching-free solution is more cost-effective.
Dimension 3: Environmental constraints - the “invisible threshold” of technology implementation
Four major constraints that must be evaluated:
① Workshop dust/oil level (affecting vision system accuracy)
① Workshop dust/oil level (affects vision system accuracy)
② Grid fluctuation range (whether the equipment can work stably under ±15% voltage variation)
③ Spatial accessibility (pipelines/tight spaces require customized robotic arms)
③ Space accessibility (customized robotic arms for pipelines/narrow spaces)
④ Process certification requirements (automotive industry needs to comply with IATF 16949 process specifications)
Process selection of the five “fatal misunderstanding”: to avoid 90% of the customer procurement pit
Myth 1: “Fully automated = completely unmanned”.
Reality: no teaching still need process experts to set quality rules, the blind pursuit of unmanned may lead to a spike in scrap rate
Avoid the pit strategy: require suppliers to provide process parameters debugging interface, retain the key nodes of manual review rights
Myth 2: “The more functions the software has, the smarter it is.”
Truth: Functional redundancy will increase the complexity of operation, a customer purchased “all-in-one” equipment because the operator mistakenly touched the AI button, resulting in batch rework.
Core principle: choose a system that supports modular subscription (e.g., purchase basic positioning functions first, then upgrade as needed).
Myth 3: “Hardware parameters equal actual performance”.
Key indicators disassembled:
Repeat positioning accuracy ± 0.05mm ≠ welding trajectory accuracy (affected by torch deformation, heat input deformation)
Maximum speed 2m/s ≠ effective welding speed (need to consider the acceleration and deceleration process energy stability)
Suggestion: Use the actual workpiece to carry out zigzag trajectory welding, and test the consistency of the depth of fusion at the inflection point.
Myth 4: “One-time investment to end the battle”
Long-term cost list:
Annual fee for software licenses (some vendors charge by number of robots)
Process database update fee (new material adaptation requires the purchase of data packages)
Four Steps to Scientific Decision Making: A Complete Map from Requirements to Landing
Step 1: Digital modeling of the process
Toolkit:
✅ 3D scans of welded seams (to assess trajectory complexity)
✅ Material heat input sensitivity analysis (to determine control accuracy requirements)
✅ Welding process evaluation report (to define certification criteria)
Output: “Digital Portrait of Welding Process” (with 9 dimensions of scoring)
Step 2: Technology Path AB Test
Comparison of program design:
Program A: high-precision demonstration teaching robot + expert process package
Scheme B: Teach-free robot + adaptive algorithm
Test metrics:
✅ First-piece pass rate ✅ Changeover time ✅ Consumables cost/meter welded seam
Step 3: Supplier Capability Penetration Assessment
Soul six-question checklist:
① Can you provide test weldments of the same material? (Rejected generic demo parts)
② Is the algorithm open to process weight adjustment? (Prevent “black box” decision-making)
① Can you provide test weldments of the same material (reject generic demo parts)?
④ Is the after-sales service response time less than 4 hours?
⑤ Does it support acceptance by third-party testing organizations?
⑤ Does it support acceptance by third-party testing organizations?
⑥ Is the sovereignty of data clearly attributed? (Prevent process data from being locked)
Step 4: Small Scale Validation → Rapid Iteration
30-day validation plan template:
Week 1: Basic function acceptance (positioning accuracy, arc stability)
Week 2: Extreme working condition test (large angle climbing welding, strong electromagnetic interference)
Week 3: Production beat challenge (continuous 8-hour full load operation)
Week 4: Cost audit (consumable loss rate, gas consumption comparison)
Conclusion
The end point of welding intelligence is to bring technology back to the essence of the process! When serving a new energy vehicle supplier, we decisively recommended that the robot be retained for the box weld (due to the high consistency of the workpieces), while the non-teaching program was adopted for the shaped joints of the impact beam. This “hybrid intelligence” strategy helped the customer save 41% of the initial investment.
Translated with DeepL.com (free version)
