Purpose of this guide Fused Granulate Fabrication extrudes large polymer parts directly from pellets. High throughput and lower feedstock cost can suit tooling, enclosures and low-volume spares, but scale amplifies thermal, shrinkage, feeding and machine-state variation. Validation is not a dramatic demonstrator; it is evidence that a defined part family repeatedly conforms.

This guide translates Fraunhofer IMWS AddiQ results into measurement logic. A sensor is not evidence by itself. The measurement chain, process signal, coupon, finished-part inspection and decision rule must be controlled as one versioned system.

1. Start from requirements Record intended use, maximum size, loads, environment, fire and chemical requirements, surface and dimensional tolerances, quantity, finishing and inspection. 'Rail spare part' is not a sufficient specification: an interior trim, a bracket and a safety-critical component require different evidence.

Select critical quality characteristics such as mass, interlayer strength, void content, maximum warpage, hole position, machining allowance or surface defects. Each needs a measurement method, uncertainty and acceptance limit.

2. Lock material state Identify pellets by manufacturer, grade, lot, recycled fraction, drying, storage, particle-size distribution and moisture. Nominally identical polymers with different additives, fillers or particle shapes can feed and melt differently. Verify the drying recipe with measured moisture, not time and temperature alone.

Reprocessed feedstock needs mass balance and a mixing rule. Test repeated thermal history, contamination and particle-shape effects. A change in blend ratio is a process change, not an administrative detail.

3. Qualify machine and feed Check axis or robot coordinates, tool centre point, platform flatness, thermal zones, feeder repeatability and emergency functions. Measure through the full work envelope; accuracy at the centre of a large system does not establish accuracy at its edges.

Run mass-flow tests at several screw speeds and back pressures. Measure delivered material rather than trusting the command. Calibration must cover intended materials, nozzles and temperatures; screw and nozzle wear will move the result over time.

4. Sensor strategy with a physical hypothesis Collect a signal only when a named failure mechanism and response exist. Pellet moisture may support gas and bond monitoring; melt pressure may indicate feed and blockage; acceleration can reveal mechanical anomalies; platform mass provides material balance; ambient and feed temperature support thermal-history control.

Record sensor location, range, accuracy, calibration, sampling, time synchronisation and behaviour during data loss. If clocks drift, an event can be assigned to the wrong path segment. Time base is therefore as critical an interface as the geometric coordinate system.

5. Link signal to property Use designed experiments to vary controlled factors and manufacture repeats. Coupons must represent real layer time, path direction, bead section, corners, stops and restarts. A small tensile bar alone does not model heat accumulation in a long wall.

Pair process data with geometry, mass, nondestructive and mechanical results. Evaluate models on separate training and verification data. Report sensitivity, false alarms and domain of use; high correlation does not prove causation or necessarily support release.

6. Nozzle and wear Create a new-condition reference image, then follow outlet edge and bore with fixed optics, lighting and position. Link image metrics to bead width, mass flow and finished-part defects. The replacement limit should represent validated quality risk, not merely pixel difference.

Also test repeatability of cleaning and installation. A badly reseated nozzle or changed tool centre point can cause more error than slow wear. A short confirmation coupon is required after maintenance.

7. Representative demonstrator and full scale The demonstrator should include the longest continuous path, largest section, critical corner, hole, transition, start and stop. If scaled, analyse the scale effect separately: cooling time, self-weight, path length and residual stress do not change proportionally.

Inspect the first full-size article with its real finishing and fixturing. Machining may hide a surface defect but expose an internal void, so evidence is needed in both as-built and finished states.

8. Capability and release Use at least three consecutive, separately started builds for initial repeatability, and more for critical applications. Match sample size to risk and the statistical claim. Evaluate spread, trend and position dependence as well as the mean.

The release rule must state which process alarm stops production, what triggers reinspection and when an engineering deviation is allowed. Sensor data must not become an automatic conformity stamp until its relationship to finished-part properties is validated.

9. Change control A new material lot, supplier, recycled ratio, dryer, screw, nozzle, firmware, software, path generator, sensor or finishing step needs impact analysis. Define in advance what requires a short regression coupon and what requires full requalification.

Preserve raw signals, processing algorithms, thresholds and versions. If the detector later changes, earlier data must be replayable so the organisation can identify affected release decisions.

Minimum evidence package The package contains requirements; material and machine state; calibrated measurement chains; experimental design; process-to-property relationship; representative and full-size trials; capability results; deviation and release rules; and change control.

10. Receiving small-batch pellets Small packs reduce inventory risk but must not weaken material identification. At receipt record manufacturer, exact grade, lot, pack mass, seal condition, SDS and TDS revisions, and any shipping deviation. 'The same material used for filament' does not replace a pellet lot identifier and a defined formulation scope.

Treat pre-dried status as an initial condition, not a permanent guarantee. Measure moisture on opening and after extended storage according to risk; define resealing, dry storage and re-drying rules. A smaller canister only retains its advantage if the workshop preserves condition through final use.

11. Laboratory compounding and in-house filament Adding colour, stabiliser or filler makes mixing order, dosing accuracy and homogeneity critical process parameters. Visual uniformity in a dry blend does not establish dispersion. Sample the beginning, middle and end of a batch and connect those samples to colour, flow, diameter or mechanical results.

For in-house filament, continuously measure diameter and ovality, track winding tension and separately validate the resulting print. In FGF, compounding and final extrusion may occur in one screw; include mixing length, residence time and cleaning in the validation.

12. Pack-size decision gate Do not choose pack size from price per kilogram alone. Estimate coupons, setup waste, purging, the representative demonstrator and reserve, then compare that amount with allowed open-pack storage time. A smaller pack may be worth more during introduction because it limits inventory exposure to a wrong choice; stable production may benefit from fewer lot changes with a larger pack.

Prusament's pellet range announced on 13 October 2026 illustrates an intermediate 3.5–14 kg purchasing tier. It does not establish equal performance on every pellet extruder, but it allows laboratories and workshops to run a genuine multi-kilogram validation without an industrial minimum order.

Editorial conclusion FGF quality assurance depends not on data volume but on a defensible link among signal, failure mechanism and part requirement. AddiQ provides a useful pattern connecting sensors, nozzle inspection and test parts. It becomes an industrial system only when uncertainty, decision errors, full scale and change are also validated.