Controlling distortion in robotic structural-steel welding

Define the dimensional problem before changing the program

A robotic cell can repeat a welding sequence while the assembly still changes shape. When that happens, the first useful action is to define the distortion: where it occurs, when it appears and how it is measured. A photograph of a bowed member is useful context, but a decision about process or fixture changes needs a reference, measurement locations and an agreed final condition.

TWI explains that localized heating and subsequent cooling create nonuniform stresses and can produce permanent deformation. Its distortion guidance distinguishes longitudinal and transverse shrinkage, angular distortion, bowing or dishing, buckling and twisting. [1] These forms require different observations. A single overall length check cannot describe a twisted or locally buckled assembly.

Start from the product drawing and the project-specific dimensional requirements. Identify which deviation is unacceptable and which stage is being assessed. This article provides a review method, not a universal tolerance or a welding sequence suitable for every beam. The responsible engineering team must approve changes to design, procedure, restraint and any corrective work.

Measure the same features in defined states

Choose reference features that remain accessible and meaningful through the fabrication sequence. Record the assembly before welding, at selected intermediate stages if the trial requires them, and after the specified cooling and release condition. State whether the part is clamped, supported or free at each measurement. A reading taken under restraint cannot be treated as the final unrestrained dimension without explanation.

Use a measurement map. For a long member, it might identify ends, intermediate stations and selected cross-sections. For a stiffened plate, it might identify reference corners and local panel regions. The exact arrangement should follow the distortion question and the inspection plan. Record the instrument and method so repeated trials can be compared fairly.

Separate incoming variation from welding-induced change. If the initial assembly already differs between trials, the final dimensions alone will not explain whether a new sequence helped. Keep material, fit-up, supports and measurement state visible in the record. This discipline turns a workshop observation into usable engineering evidence and reduces the risk of crediting a program change for a difference that actually came from the starting workpiece.

Review weld size and placement before adding compensation

TWI's design guidance identifies weld placement, weld-metal volume and balanced welding as factors in distortion control. It also warns against unnecessary over-welding. [2] The practical starting point is to verify that the programmed weld corresponds to the drawing requirement. A larger visible bead is not automatically a better manufacturing result.

Check whether the drawing specifies leg length, throat or another defined weld dimension, and make sure the inspection method matches it. Review any gap or preparation variation that changes the required deposition. If the actual joint differs from the assumed joint, changing the robot path alone may not address the source of variation.

Design changes require the appropriate authority. Do not shorten a weld, substitute intermittent welding or reduce its size simply because those changes might reduce heat or deposited metal. The responsible designer must determine whether the revised joint meets the structural requirements. Keep proposed design changes separate from process adjustments within an already approved detail, so the team can see who must review each option.

Develop sequence options around the complete assembly

A welding sequence is more than an ordered list of seam identifiers. It includes travel direction, weld position, pass progression, any repositioning, intermediate checks and the condition of the assembly when each weld is made. A robot makes the chosen sequence repeatable, which is useful only after the sequence has been reviewed for the actual structure.

For a development study, draw alternative sequences on the same assembly and label them as candidates. Explain the intended reason for each change, such as distributing welding across opposite sides or avoiding concentration in one region. The general balancing concept has support in TWI's design guidance, but the exact order remains an application-specific engineering decision. [2]

Check whether the proposed sequence is physically possible in the cell. A theoretically attractive order may require a new torch approach, blocked scan, additional turning operation or unavailable inspection access. Include those consequences in the comparison. Do not accept a sequence solely because it produces a clean animation or appears symmetrical on a two-dimensional drawing.

Use restraint with a clear release plan

Fixtures and clamps influence how an assembly moves while welding. TWI describes restraint, presetting and prebending as possible distortion-control approaches and notes that restraint can introduce residual stresses and increase cracking concerns in susceptible situations. It also emphasizes representative trials when determining presetting. [3] More restraint is therefore not an automatic solution.

Document which features the fixture locates, which directions it restrains and when clamps are applied or released. Include removal access and the condition in which final dimensions will be assessed. A part that looks satisfactory while forced into a fixture may move when released. That change belongs in the trial evidence rather than being treated as an unexpected exception after acceptance.

Any proposed preset amount or prebending method needs an engineering basis and validation. This article provides no angle, force or compensation value. The chosen approach must respect the material, joint, process and product requirements. Where temporary attachments are used, their application and removal also need the appropriate controlled instructions, including any required inspection of the affected surface.

Keep process changes inside the approved framework

Distortion investigations often lead to proposals about heat input, travel speed, pass size or welding position. Such changes can affect more than shape. They may also affect the joint properties and the scope of the applicable procedure. Review them with the responsible welding engineer before applying them to production.

ISO 15609-1:2019 addresses the content of arc-welding procedure specifications. [4] Use the applicable WPS and its supporting qualification basis to determine what is authorized. A distortion trial is not permission to change a controlled variable beyond its approved range. Record whether an option is a process-development experiment or a production-approved adjustment.

Keep the controller record understandable. A changed power-source job, speed command or sequence revision should identify the reason for change and the joints affected. If an experiment uses several simultaneous changes, recognize that the resulting evidence may not identify which change caused the difference. Plan trials that produce an interpretable decision, rather than accumulating settings that happen to yield a better-looking specimen.

Build a practical trial comparison

Select a representative assembly and establish the baseline configuration. Record incoming dimensions, fit-up, material, consumables, fixture, procedure, robot program and measurement method. Then define the change under investigation and the required observations. Keep other relevant conditions consistent enough to support the intended comparison.

Evaluate weld acceptance and dimensional outcome together. A sequence that improves shape but produces unacceptable welds is not a successful solution. Likewise, a trial that passes weld inspection but requires an unreviewed corrective operation has not established the intended manufacturing route. State the complete outcome and any additional work required.

Record exceptions such as interruptions, repairs or fixture adjustments during the trial. They may explain why a specimen differs from the planned condition. Repeat or expand testing when those exceptions prevent a useful conclusion. The number of specimens and acceptance rules must be set for the application; this guide does not invent a universal sample size or statistical confidence claim.

Plan for the next weld after the assembly moves

In a model-led cell, deformation can affect the relationship between the nominal geometry and the remaining joints. Review when the workflow relies on initial registration and when it uses local seam information. A previously valid reference should not be assumed adequate after a substantial change in presentation or shape without the checks required by the configured process.

Include intermediate observations where the engineering study identifies a need. The decision might concern whether the next joint remains within the reviewed approach, whether a clamp has shifted or whether the remaining fit-up requires attention. Do not describe local tracking as a guarantee that any amount of deformation will be accommodated.

Keep motion and process responses coordinated. If the part has moved, an approach may need review; if the joint condition has changed, the WPS or manufacturing instruction may define additional action. These are connected questions with different owners. A clear trial record helps the automation and welding teams identify the appropriate response instead of treating every problem as a path offset.

Compare results without hiding the starting condition

Consider two illustrative trials of the same nominal assembly. One finishes straighter, but it began with a different initial bow and used a different support position. The final reading alone cannot establish that its welding sequence was better. Compare the measured change from the documented starting state and identify the support difference before drawing a conclusion.

A useful trial table includes initial dimensions, fixture state, sequence revision, interruptions and final measurement state. Add the weld-inspection outcome. Make an inconclusive comparison visible instead of forcing every trial into a success or failure label. Sometimes the correct next action is to improve the experiment's consistency.

When a method is accepted, retain the conditions that made the comparison meaningful. Production control should identify when incoming variation or fixture changes exceed the evaluated scope. A clear controlled record can be sufficient; what matters is that the team can relate the result to actual conditions. Record the responsible reviewer and the reason for the release. If a later part behaves differently, this evidence helps determine whether the method needs further development or whether the production setup has departed from the validated condition.

Release a repeatable method with a measured boundary

The accepted method should identify the part family, fixture, procedure, sequence, reference checks and dimensional measurement state. Include the range of incoming conditions actually evaluated. Describe what staff must do when a part differs from that scope, rather than assuming that the validated sequence covers any similar structure.

If corrective straightening is required, treat it as a separate controlled operation subject to the applicable engineering and product requirements. Do not hide it inside a robot performance claim. The buyer needs to understand the complete route from incoming assembly to accepted product, including any additional labor, inspection or handling that the route requires.

For a first project review, send the model, weld details, initial and final measurement records, fixture photographs and the sequence currently used. Those inputs allow the engineering team to identify the controlling uncertainty and plan a focused trial. Distortion control becomes more reliable when the team can explain what moved, which change was tested and under what conditions the resulting method is acceptable.

Project review checklist

• Record initial dimensions and the reference measurement state.
• Identify the expected distortion modes for the actual member.
• Document tack arrangement, restraint, sequence and procedure.
• Measure the final condition after the agreed release and handling steps.
• Review corrective work and changes before repeating or extending the method.