Jig Design #07: Multi-Model Fixtures — Common Base, Change Parts and Error-Free Changeover

A multi-model jig is not simply a base drilled with enough holes to "mount anything".

If every locator and clamp is freely adjustable, changeover takes a long time, depends on operator skill and is hard to reproduce. If every model gets its own dedicated jig, stability is high but investment cost, storage space and the time needed to get a jig onto the machine all go up.

Good architecture usually sits between the two extremes:

A common base holds the stable interfaces.
Change parts carry the geometry that differs.
A quick-change mechanism restores the datum.
Recipe, part ID and hardware configuration stay in sync.
Wrong combinations are either prevented or detected before the cycle starts.

This article shows how to build that system, from product family to validation.

1. Choose the architecture from volume, mix and risk

Three common architectures:

Dedicated jig

One jig per model.

Suitable when:

Volume is high.
The product runs over a long life cycle.
Models differ substantially.
CTQ or safety demands a dedicated optimum.

The advantage is simplicity and stability. The drawback is investment cost, floor space and having to swap the entire jig.

Common base + change parts

Base, utilities, safety and machine interface are shared; nest, locator, clamp pad or sensor target change with the model.

Suitable when:

The product family shares datums and process.
Differences are concentrated in a few features.
Changeover happens often.

Modular fixture

Assembled from plates and standard elements into many configurations.

Suitable when:

Low volume, high mix.
The product is not yet stable.
Elements need to be reused.

There is no universally "best" architecture. Compare total cost: investment, changeover time, first-piece verification, storage, maintenance and the risk of a wrong setup.

2. A product family needs an engineering basis

Do not group models together just because their commercial names look similar.

Build a matrix of:

Functional datums.
Envelope dimensions.
CTQ.
Loading direction.
Process loads.
Clamping points.
Sensors and connectors.
Cycle time.
Cleanliness level and material.
Failure modes.

Models fit a common base when they share:

The datum system.
The force path.
The working space.
The machine interface.
The safety concept.

If one model needs a completely different datum or clamping force, forcing it into the common architecture can make every model worse.

3. Separate common and variable interfaces

Common interface

Should stay stable across models:

Mounting to the machine.
Datum base.
Air, electrical and vacuum utilities.
Safety interface.
PLC I/O or communication.
Robot approach zone.
Lifting points and storage cart.

Variable interface

Carries the differences:

Nest profile.
Locators tied to model features.
Clamp pads.
Sensor targets.
Guide funnels.
Connector inserts.
Go/no-go features.

The goal is to keep the variable part small, light, easy to preset and unable to disturb the alignment of the common base.

4. Locating and holding a change part are two different functions

A bolt should not both "find the position" and "hold it accurately" when the fit clearance is not controlled.

A quick-change interface needs:

A datum or seating face.
One feature that defines the center.
One feature that defines orientation while releasing redundant constraint.
Retention that pulls the part back onto the datum.
Resistance to lift-off and to process loads.
A controlled release.

A common configuration is a round pin combined with a diamond pin, or a zero-point interface. Carr Lane documents many quick-change solutions that use separate locating and retention elements to restore position.

Do not apply catalogue repeatability figures directly to the whole jig. The system result also includes:

Mounting tolerances.
Cleanliness.
Wear.
Deflection under clamping.
Temperature.
Measurement method.

5. The common base needs strong datums and the ability to stay clean

The base is the foundation for every model. If the datum base is dirty or worn, every configuration drifts.

Design for:

Datum faces that are just large enough and easy to clean.
Relief around non-functional areas.
Chip and fluid escape paths.
No pockets that trap debris.
Directed air blow that does not push chips onto another locator.
Replaceable wear inserts.
A check standard or master change part.

If air is used to confirm seating, design the orifice and the logic so that contamination, leakage or pressure loss is detected instead of producing a false pass.

6. Change parts must be light, stiff and self-locating

A good change part:

Can be handled safely by one person or by a robot.
Has handles in the right places.
Never puts hands near a clamping point.
Does not need dial-indicator alignment every time.
Has no loose shims that can go missing.
Is keyed against reversed installation.
Carries a unique ID.
Has a storage condition that protects its datums.

Reduce mass with sensible pockets and ribs, but never at the cost of stiffness along the force path. If a change part is heavy, use a cart, rail, lifter or docking arrangement instead of relying on human strength.

7. SMED: separate internal and external setup

The Lean Enterprise Institute describes the core insight of SMED as:

Separate the work that can only be done while the machine is stopped from the work that can be done while it runs.
Convert as much internal setup as possible into external setup.

Internal setup

Release and remove the old change part.
Install the new change part.
Make the required connections.
Confirm seating and lock.
Run the mandatory checks on the machine.

External setup

Prepare the new change part on a cart.
Verify the correct revision.
Clean it.
Preset locators and pads.
Fit sensors and connectors in advance.
Prepare tools and material.
Confirm the recipe and the order.

Do not measure changeover only from "last good part" to "first cycle start". A more meaningful indicator runs from the last good part of the old model to the first good part of the new one.

8. Remove steps, do not just perform them faster

Improvements worth making:

Captive fasteners.
One motion that locks several points.
Keyed and grouped connectors.
Color coding as support, never as the only cue.
Presetting away from the machine.
Fixed stops instead of adjustable scales.
Integrated handles.
Carts at the right working height.
Dropping special tools where they are not needed.

Every fastener or setting is an opportunity for error. Ask:

Can it be eliminated?
Can it be preset?
Can it be made self-locating?
Can it be confirmed automatically?

9. The configuration matrix is the source of truth for multiple models

Build a matrix:

Model	Base	Nest	Locator	Clamp pad	Sensor target	Recipe	Check standard

Every cell needs a clear ID and revision. Do not use names like "the tall block" or "the old-type pin".

The matrix supports:

BOM.
Work instructions.
PLC recipes.
Spares.
Validation.
Change control.

When a model changes, review the whole row and every interface it touches.

10. Recipe handshake: the production order must match the hardware

A robust system compares:

The model from the production order or MES.
The recipe that was called.
The ID of the change part.
Workpiece identification.
The state of locators, clamps and sensors.

The cycle is allowed only when all of them agree.

The ID can come from:

A mechanical key.
A sensor pattern.
A keyed plug.
Hardwired coding.
RFID or barcode.

QR and RFID are not automatically poka-yoke. If a misread, a lost connection or an operator skipping the step still lets the machine run, it is only information. The logic needs a reaction, diagnostics and controlled bypass.

11. Prevent hybrid configurations

One dangerous failure mode looks like this:

Nest from model A.
Clamp pad from model B.
Recipe for model C.

Each part fits on its own, but the combination is wrong.

Countermeasures:

Assemble change parts into a cassette that cannot be split.
Key families so that only compatible sets can be mounted together.
Encode the whole configuration in the connector.
Confirm every critical element with a sensor.
Store kits by model.
Require the kit remainder to be zero after changeover.
Check combinations in the matrix logic, not just "a part is present".

Challenge testing must cover the hybrid combinations that can realistically occur.

12. Mechanical poka-yoke first, electronic identification second

Give priority to:

Asymmetric pin patterns.
Pockets that only accept the correct family.
Connectors with different keys.
Stop lengths that prevent a wrong change part from seating.
Captive parts that cannot migrate to another model.

Only then add sensors to confirm.

The benefits:

No dependency on software for simple geometric errors.
Failures are visible.
Recovery is fast.
Fewer sensor-driven false passes.

But poka-yoke must account for tolerance and wear. A wrong change part must never be "almost able to go in" and then be forced by the operator.

13. Utility connections: fast, but correct and safe

Air, electrical, vacuum and coolant connections need:

Keyed connectors.
No possibility of cross-connection.
Self-sealing where required.
Strain relief.
No leakage in the disconnected state.
Enough hose and cable travel.
Controlled energy state while connecting.

If the model is changed with the machine stopped, define clearly:

Which energy sources must be isolated.
Which may stay live for setup.
Who is allowed to work on them.
Which logic prevents unintended actuation.

Quick change must never be shortened by dropping a safety step.

14. The storage cart is part of the system

Change parts are usually damaged away from the machine:

Datum faces set down on a bench.
Locators knocked against something.
Revisions mixed up.
Chips sticking to surfaces.
Fasteners lost.

A cart or rack needs:

A dedicated pocket per ID.
Protection for datum faces.
Captive positions for fasteners.
A safe picking posture.
Prevention of storing a part in the wrong slot.
A clean / awaiting-cleaning status.
Control of spares and revisions.

External setup is only fast when the change part reaches the machine correct, clean and ready.

15. Validate per model and across models

Each model needs:

Seating.
Repeatability.
CTQ.
Clamping force and deflection.
Sensors and interlocks.
A wrong-model challenge test.
Cycle time.

Cross-model testing is needed as well:

Does A → B → A return to the baseline?
Does the changeover sequence matter?
Do the largest and smallest models still have enough clearance?
Are hybrid combinations blocked?
Does a recipe mismatch lock the cycle?
Can a worn change part produce a false pass?

Do not validate only the "typical" model. Choose the worst case for:

Envelope.
Load.
Locator tolerance.
Mass.
Number of utilities.
The smallest difference between two models.

16. First-piece verification

After changeover, define:

Which CTQ is checked.
Which gauge is used.
Who approves.
When series production is released.
What happens to the product and the jig if it fails.

The goal of SMED is not to drop the first-piece check where it is needed. The goal is to standardize it and shorten it through presetting, check standards and automatic data.

Track first-piece yield. A fast changeover with frequent first-piece failures only moves time from setup into troubleshooting.

17. Change control for the product family

Every ECN should ask:

Does the datum change?
Does the envelope exceed the common base?
Do locators and pockets still act as poka-yoke?
Does the clamp pad still contact the right area?
Are the recipe and the ID mapping updated?
Can the new model slip into an old configuration?
Do storage, spares and work instructions change?
What is the validation scope?

Do not let a new model onto the line just because "it fits". The product-family matrix has to be approved again.

18. Operating indicators

Track:

Last-good-to-first-good changeover time.
Internal and external time.
First-piece yield.
Setup errors.
Recipe and configuration mismatches.
Recovery time.
Number of bypasses.
Drift versus number of changeovers.
Change-part damage.

Use video analysis or time observation to see searching for tools, walking, cleaning and waiting for confirmation. Do not simply ask the operator to move faster.

19. A design process for multi-model fixtures

Step 1 — Build the product-family matrix

Compare datums, CTQ, envelope, loads and process.

Step 2 — Choose the architecture

Dedicated, common-base or modular, based on volume, mix and risk.

Step 3 — Split common and variable

Keep the variable part small and unable to affect common alignment.

Step 4 — Design the quick-change interface

Datum, centering, orientation and retention clearly separated.

Step 5 — Design the poka-yoke

Against wrong model, wrong orientation, missing part and hybrid configuration.

Step 6 — Design the external setup

Presetting, carts, cleaning, IDs and tools.

Step 7 — Create the handshake

Order × recipe × change-part ID × part ID.

Step 8 — Validate

Per model, cross-model, worst case and challenge test.

Step 9 — Run the life cycle

Spares, storage, wear, revisions and requalification.

20. Review checklist

Architecture

The product family is based on datums and process, not only on model names.
Dedicated, common-base and modular options were compared on total cost and risk.
Common and variable interfaces are clearly separated.

Quick change

Datum, centering, orientation and retention each have their own function.
The change part is self-locating and needs no manual alignment.
Relief, cleaning and datum protection are provided.
Mass and handling are safe.

Configuration

The model–hardware–recipe matrix is the source of truth.
Hybrid combinations are impossible, or a wrong combination is detected.
The handshake locks the cycle on a mismatch.
Bypass and diagnostics are controlled.

SMED

Internal and external setup are separated.
Change parts are preset and prepared away from the machine.
Fasteners, connectors and tools are reduced to the minimum.
Last-good-to-first-good is measured.

Validation

Every model and the worst case are tested.
An A → B → A test exists.
Wrong-model and wrong-configuration challenge tests exist.
First-piece verification and release are defined.
ECNs include impact assessment and requalification.

Conclusion

A good multi-model fixture is not the one with the most adjustment capability. It is a controlled configuration system.

The common base keeps datums, utilities and safety stable. Change parts carry the differences, locate themselves and are prepared away from the machine. The recipe handshake makes sure the production order, the hardware and the workpiece all refer to the same model. Poka-yoke blocks the wrong combination before the cycle starts.

When architecture, SMED and configuration control are designed together, changeover becomes fast, repeatable and free of new quality risk.

View all MINATA technical articles