Robotic Welding Changeover Planning Checklist

Measure last-good to first-good, not fixture movement

A changeover includes more than removing a nest. Material identification, program selection, torch condition, checking the next assembly and waiting for a release decision can all keep the cell unavailable. State whether the finish event is first accepted part or stable production at an agreed condition; record both if they answer different planning questions.

Build a transition list from the actual product mix: shared-fixture recipe changes, insert changes and complete-pallet exchanges are different jobs. Group compatible parts using geometry, locating scheme, process and handling requirements, not the product name alone. Miller's manufacturer guidance includes part families and fixture-changeover downtime in the whole operation; the timing worksheet here turns that planning concern into named events.

Further reading: Miller — robotic welding planning and part families ↗

Separate offline preparation from stopped-line work

Before the current run ends, stage the correct material batch, inspect the replacement nest, check that required gauges are available and retrieve approved files. Record completion and any missing item. A nearby cart does not make a task ready if its fixture condition, lifting method or revision remains unresolved.

Removing the current fixture, connecting utilities and proving the installed configuration usually depend on an approved stopped-cell state. Follow the application's access, isolation and restart instructions; this planning sheet does not define an isolation procedure. Schedule shared technicians, cranes and inspection equipment explicitly. Work cannot overlap simply because two people are assigned to it.

Further reading: Premier Automation — welding fixtures and tooling ↗

Match fixture identity, locking and program revision

Define the allowed combination of part revision, fixture ID, insert set, robot program, weld recipe and positioner state. Check the locating surfaces separately from clamp or lock confirmation: a closed clamp does not identify the part revision. Include hose/cable routing, connections and the condition of locating wear parts in the setup record.

TTM's fixture-switching discussion connects installed fixture identity with control and program selection. For the buyer, the useful deliverable is a state/recipe table and a test for a mismatched or missing ID, conducted under the approved commissioning procedure. Specify who may load or change files, how backups are identified and which changes require another first-piece approval. A simulated path does not replace installed-cell verification.

Further reading: TTM — fixture changeover, identification and utilities ↗

A dependency-based changeover time model

For each action record start, finish, predecessor, assigned resource and whether it occupies the stopped-line window. Elapsed changeover is the duration of the longest valid dependency path between the start and release events, including queues and waiting. Add serial tasks; for genuinely independent tasks with separate resources, use the longest parallel branch. Total labor-minutes sum all people's task time and are a separate measure.

Hypothetical assumptions, not measured performance: preparation 8 min, fixture exchange 6 min, configuration checks 4 min, consumable check 2 min, trial part 5 min and inspection/release 3 min. If all run serially after stopping, elapsed time is 8 + 6 + 4 + 2 + 5 + 3 = 28 min. If preparation is completed beforehand and the 4-minute configuration check can safely overlap the 2-minute consumable check with separate resources, stopped-line time becomes 6 + max(4, 2) + 5 + 3 = 18 min. Labor remains 28 person-minutes under these assumptions; only its placement changes. Replace every value and validate the overlap before using the result.

Run sensitivities on the constraint: an extra 7-minute inspection queue adds 7 minutes to either route; an unavailable setup technician can remove the assumed overlap. Track the actual spread by transition and reason rather than publishing one average. The model does not predict output, safety performance or investment payback.

Include first-piece approval and held material

Agree which dimensions, weld features and process records the first B part must satisfy, the inspection method and the authorized decision maker. Identify the trial piece and all subsequent pieces made before release. If the inspection needs cooling, sectioning or an external service, include that dependency and waiting in the plan instead of ending the clock at arc-off.

A rejected first piece returns to a defined cause-and-recheck route. Keep the original program/recipe and observed defect linked to the corrected version. Product acceptance criteria come from the applicable drawing, specification and quality plan; there is no universal first-piece sample size or weld threshold for every cell.

Plan abnormal recovery and the next shift

List credible interruptions: fixture ID mismatch, missing utility confirmation, damaged locating surface, unapproved file, failed trial or interrupted cycle. Define the stop/hold decision, person to contact and the evidence needed before restart. Restoring a previous configuration requires its own verified state and part identity; reverting a filename alone is insufficient.

At handover record the installed configuration, approved first piece, outstanding issues, segregated stock and current operator instruction. After maintenance or a program edit, assess which location, path, quality and recovery checks are affected. Use the same controlled checklist across shifts while keeping actual observations attached to each transition.

Changeover action, time and responsibility matrix

For your A→B transition add planned/actual duration, predecessor and named person to each row. O = offline candidate; S = stopped-line dependency; Q = quality release. These categories are planning prompts, not permission to overlap work or enter a cell.

Copy the rows into your RFQ or investment worksheet.