An extrusion traceability implementation cannot stop at linking one resin lot to one finished roll. Virgin resin, regrind and additives may mix in a hopper; a recipe can change during a run; and quality at the beginning of a roll may differ from quality at its end. This guide helps production, quality, maintenance and IT teams at Thailand plants specify a system that connects material receipt and feeding, extruder conditions, die and haul-off, winding, inspection, converting and shipment. It is written for an RFP, vendor demonstration and FAT/SAT.
Choose a traceable unit before collecting data
Extrusion is continuous. It does not automatically produce the discrete shot of injection molding or billet of forging. A production order, operating interval, feed interval, parent roll and metre segment are different units. Decide which one a customer will complain about and which one quality will quarantine. Film, sheet, pipe, profiles and cable coating also have different physical finished units. The roll is our main example; for pipe substitute length and cut bundles, and for profiles substitute bars and bundles.
Suppose a customer reports a thickness defect in the first 300 metres of roll R. A daily average barrel temperature cannot explain that segment. The team needs its winding time, line, candidate material lots, recipe revision, die and haul-off conditions, and inspection evidence. Conversely, if supplier resin lot A is suspect, the team must trace every potentially affected roll, cut item and consignee, accounting for hopper residue and purge. These backward and forward queries should be the first acceptance scenarios.
The GS1 Global Traceability Standard describes critical tracking events (CTEs) and key data elements (KDEs) for receipt, transformation, packing and shipping. A manufacturing transformation should preserve the relationship between inputs and outputs. GS1 does not prescribe a universal metre resolution or a sensor list for an extruder. Those choices depend on customer requirements, product risk and what the machine can measure. EPCIS 2.0 TransformationEvent is an optional way to express the relationship for interoperability; first make the plant’s internal event model work.
| Unit | Typical question it answers | What it can miss |
|---|---|---|
| Production order | Which customer order? | Changes within the order |
| Run interval | Which line and shift? | Feed changes or local roll defects |
| Material feed event | Which incoming lots were candidates? | Residue and transition boundary |
| Parent roll | Which labelled physical roll? | Difference between head and tail |
| Metre segment | Which local part of a roll? | Encoder slip and time alignment |
| Child roll or cut bundle | What was delivered? | Parent split and merge if unrecorded |
Start with the complaint and isolation unit instead of promising the smallest conceivable resolution. A roll may have one laboratory disposition while an online gauge records local thickness. If encoder slip, reel changes or speed changes make the mapped position approximate, store that uncertainty and widen the suspect interval. False precision is less useful than an honest candidate range.
Connect virgin resin, regrind and additive lots through feed events
At receipt, distinguish supplier, grade, supplier lot, internal receipt ID, quantity, inspection disposition and storage location. Transfer from warehouse is not proof of actual feeding. Record opening, weighing, blending, loading, return and scrap as separate events. A bag, bulk container, silo and central feeder may require different capture points. Keep the physical container and its lot identity together when material moves between them.
Regrind creates another parent-child relationship. For in-house regrind, link the original roll or scrap container to a regrind batch and then to the later feed. For purchased regrind, record the supplier lot and incoming inspection. Do not collapse both into a free-text field called “regrind.” Keep target blend ratio in the approved recipe separate from actual dosing and feed records. A feeder stop or top-up can change the actual blend without changing the recipe.
Small additive quantities can still be relevant to quality. Record the lot, quantity, destination and container change time of masterbatch or additives. If equipment cannot distinguish a precise one-metre section physically, retain a candidate set and uncertainty. Assigning every metre to a single resin lot requires evidence about hopper mixing, residence in the screw and die, and purge. Copying a formulation into a roll record does not prove what was fed during changeover.
Receiving quality should control use as well as record history. Can a line terminal reject a held lot? Who may approve exceptional use and why? Corrections to a supplier lot must retain original value, replacement, reason, actor and timestamp. If a label cannot be read, hold and reconcile the physical material rather than guessing.
| Event | Relationship to preserve | Common exception |
|---|---|---|
| Receipt | Internal ID to supplier lot | Several lots in one delivery |
| Weigh and blend | Multiple inputs to one blend batch | Extra material after initial weighing |
| Line feed | Blend to feeder and line | Silo switch and hopper residue |
| Regrind | Roll or scrap to regrind batch | Mixing several source products |
| End feed | Stop and remaining quantity | Material left in pipe or hopper |
The plant’s general lot management design provides a foundation for identification, issue and reconciliation. Extrusion adds the difficult mapping from continuously moving material to winding length. In contrast, the site’s forging traceability guide deals with discrete billets and press operations; do not carry its natural unit into a continuous line.
Keep recipe, setpoint and measured actual separate
Separate an approved product recipe, the revision selected for a run, the setpoint entered by an operator and the measured actual from the machine. Overwriting the recipe with a measured value destroys evidence of deviation. Model die, cooling, haul-off and winding conditions with their equipment IDs and timestamps rather than burying them under one “extruder settings” field. Record approval, trial and release-to-production states.
Possible signals include barrel zone and die temperature, screw speed, melt pressure, feeder rate, haul-off speed, thickness or diameter, cooling condition and winding tension. The right list varies by product and equipment. Before specifying “store every tag every second,” decide which variables affect quality, what update rate the control system exposes, how accurate its clock is, how much history is needed and how missing data will be handled. Preserve measurement point, engineering unit and calibration status. A dense record from an uncalibrated instrument is weak quality evidence.
Reifenhäuser describes Production Compliance tracking process parameters from the first to the last metre in an extrusion example. This shows a metre-based design is possible in a supported setup. It does not establish that every existing line or vendor product has that capability, nor that resin lots, QA and shipment are automatically joined. Verify the local signals, encoder, roll changes and MES interface on the target line.
Create explicit events for start and stop, recipe revision, manual setpoint edit, feed change, die change, speed change, alarm, quality hold and release, and reel change. Retain continuous values as time series and link them to events. A roll segment can refer to the source time series and its calculated disposition rather than copying every tag into every segment row. Keep both alarm start and recovery, and never infer “normal” from missing samples.
Synchronise machine, server and operator terminal clocks. Store a timezone or UTC offset even when operators see Asia/Bangkok local time. Keep occurrence time separate from receipt time for events arriving after an offline period. Give events stable IDs so replay does not create duplicate feeds or rolls. Missing data should appear as a gap, not as a passing value.
Map die, cooling, haul-off and winding conditions to product length
Material leaving the die reaches the winder later. Path length, speed, slip, buffer devices and stops change the delay. Simply assigning a 10:00 melt pressure to material wound at 10:00 can be wrong. Record the physical locations of measurement points, the downstream length counter and the alignment method. Validate under relevant operating conditions.
For example, a melt pressure excursion may occur before that material reaches the online gauge. Compare the two records using estimated travel time or length, and validate the mapping through physical marks or a known process change. Where slip is significant, compare length count with roll weight or another independent observation. Mark uncertain intervals rather than displaying a neat but unsupported genealogy.
Die replacement, lip adjustment, screen change and cooling adjustment are quality-relevant events. Preserve die ID and revision, installation time, operator, adjustment, first-piece check and release approval. Haul-off speed, tension and reel change influence dimensions, appearance and roll build. The equipment log must be linked to what people changed and when quality authorised renewed production.
Assign the roll ID when the winder creates it. Preserve start and end timestamps and counters, width, weight, item, line, order and disposition. If the machine resets its counter at each change, retain a link to an absolute or continuous counter. Reprinting a label should create a label revision for the same roll, not a new roll. Choose a barcode or QR encoding the floor can reliably scan; map to GS1 identifiers when external exchange requires them.
Represent changeover, blending and purge as uncertain boundaries
Stopping resin lot A and starting B does not mean that the product instantly becomes B-only. The hopper, screw, die and feed pipes may retain material, and a blend interval follows. Represent states such as “A only,” “A and B candidates,” “B only,” and “unresolved,” with the evidence and approver used to set their boundaries. Purge quantity, tests, colour observation and process knowledge may support a boundary; no universal number of metres applies to all materials and machines.
If changeover output is scrapped, link the physical section to the scrap container and show that it was not shipped. If it later becomes regrind, link that container to the regrind batch and its later feed. When physical evidence is insufficient, retain a candidate set and let quality widen the hold. The same principle applies to a recipe edit: operator command time, controller acceptance, stable actual value and quality release can be four different times. Do not use order change time alone as the physical product boundary.
The point of traceability is to identify all potentially affected units and remove only those excluded by evidence. Representing uncertainty also identifies where another sensor, procedure or validation trial would improve precision. It is more actionable than hiding uncertainty or automatically holding everything forever.
Close the genealogy through slitting, QA and shipping
When one parent roll is slit into child rolls, record the parent, its start and end length, blade position and width, child IDs, yield and edge trim in a transformation event. If a defect occupies one parent segment, the system should return every child that overlaps it. Joining material from several parents, splicing a web or mixing cut pieces in one box requires a many-to-many model. A single parent-ID field cannot describe all real converting operations.
For a laboratory sample, store its sampling position, roll segment, method, instrument ID and calibration, measured result, decision, retest and approver. Whether an edge sample represents a whole roll is a quality rule, not an IT assumption. An online gauge or vision alarm needs a start and end position and disposition. If a failed section is removed, identify the physical removed item and quantity. Retain the first failure as well as later retest and final release.
Connect child roll or bundle to carton, pallet, delivery note, customer and destination. If ERP owns orders, inventory and shipment, define the interface that passes manufacturing IDs to shipment lines. A shipment date alone cannot distinguish several rolls sent on the same day. Returns, repacking, relabelling and customer-directed sorting should add state changes without losing the original product ID.
Backward search should follow shipment line → package → child roll → parent interval → run → candidate feeds → supplier lots. Forward search should follow material lot → feed event → run and roll interval → children → stock and consignees. Use the same relationships in both directions. Prevent a held roll from being offered for shipment; record any exceptional release and its authority.
Put concrete extrusion scenarios in the RFP
“Traceability from raw material to shipment” is too broad. Attach the line drawing, controller and communication inventory, existing tag list, current feed/reel/slitting procedures, quality specifications, customer requirements and ERP/MES/WMS responsibilities. Ask each vendor which data it captures automatically, what an operator enters, and which exception blocks release. If signals are unavailable on an existing machine, price sensor addition, manual capture and staged implementation separately.
| RFP area | Specific request | Demonstration |
|---|---|---|
| Identity | Separate IDs for feed, blend, run, parent and child rolls | Duplicate label and reprint handling |
| Genealogy | Many-to-many inputs, splits, joins and overlapping intervals | Child from several parents |
| Equipment | Tag, unit, period, clock and gap handling | Stop, replay and skew |
| Change control | Recipe revision, die change and approval | Before/after values and actor |
| Quality | Fail, hold, retest, release, exceptional shipment | Original fail retained |
| Search | Both directions, candidate range and evidence | Suspect resin and defective roll |
| Operation | Thai UI, roles, offline procedure and audit log | Terminal use on the floor |
Compare more than licence cost: signal extraction, length-counter validation, labels, network, terminals, migration, training and support also matter. Separate the initial line from future rollout in the quotation. A generic demo must reproduce the site’s feed change, reel change and slitting cases. Clarify who owns each master and identifier. Production owns actual feed and changeover, quality owns hold and release, maintenance owns signal and calibration, IT owns interfaces and recovery, and logistics owns dispatch identity. The vendor must explain the exception data model as well as screens.
FAT and SAT should exercise failure cases
At FAT, load representative plant data and test the relationships and search. At SAT, include actual machine signals, labels, terminal, operator and shipment interface. Test a complaint backward and a suspect resin lot forward. Inspect not just result count but the evidence for each candidate and why other units cannot safely be excluded.
Scenarios should include A-to-B resin change, regrind addition, feeder stop, recipe edit, die adjustment, thickness alarm, reel change, slitting, splice, QA failure and retest, label misread, terminal offline, clock skew and ERP shipment replay. Test logic at FAT when a physical disturbance cannot be induced safely, then verify the relevant live signals at SAT. With missing data, the system should expose a gap, widen candidate scope or trigger hold, never silently pass the product.
| Test | Example pass criterion |
|---|---|
| Feed change | Shows A/B candidate interval; does not assert B-only without approval |
| Split | Returns each child overlapping a parent defect |
| Retest | Displays original failure, retest and final disposition |
| Outage/replay | Exposes gaps without duplicate feed or roll IDs |
| Forward search | Returns affected stock, WIP and consignees |
| Backward search | Reconstructs segment, recipe, candidate inputs and QA |
Set permitted response time, position accuracy, retention and fallback procedure to the plant’s and customer’s actual requirements. Name the acceptance authority and evidence repository. Try Thai terminology, gloves, dirty labels, shift handover and network loss in an operational test; a single successful screen demonstration does not prove sustained use.
Introduce the system in stages and measure its reliability
Choose one representative product and line. Walk the material physically from receiving label through feeding, die, gauge, winder, inspection, slitting, warehouse and dispatch. Mark each place where the ID changes. Separate information available from machines from information known only on paper. Perform a manual backward and forward mock investigation; the first broken link becomes the first implementation target.
An initial stage may connect material receipt, actual feed, parent roll, inspection and shipment. Next add recipe version, key measured variables, changeover boundaries and child segments. Later refine position alignment and external data sharing. The sequence can change with customer needs. Adding sensors before the physical IDs and operator practice are sound merely creates more unconnected time-series data.
Monitor scan success, unmatched feeds, equipment data gaps, held intervals, missing parent-child links and investigation time. Set targets from baseline observation and customer requirements. Do not improve a search-time KPI by excluding unproven suspect material. Record quality’s reason for each hold decision and invest in the steps that repeatedly produce broad uncertainty.
Make integration, revisions and recovery operational
Determine which existing ERP, MES, inspection, historian and warehouse system owns each record. An example split is product and order master in ERP, execution in MES, fast measurements in a historian and dispositions in quality software. Specify when each ID is created, who can correct it and when it reaches another system. The ISA-95 standard family provides a framework for enterprise-manufacturing integration; it does not decide a particular plant’s interfaces.
Exchange meaningful events rather than sending every sensor sample to ERP. Shared keys must still allow a roll search to reach its raw signals and candidate feeds. Choose API, file or messaging based on existing constraints, and specify retry, idempotency, error monitoring and reconciliation owner. Give material grades, recipes, dies and test rules effective dates. A historic roll must display the revision effective when it was made, not today’s revision. Audit changes to IDs and quality status with old/new value, actor, reason and timestamp.
Restore tests should prove that genealogy, time series and shipment lines can be searched together after recovery. Define retention from product, contract and quality requirements. Restrict what a customer-facing certificate exposes so it does not reveal another customer’s order or confidential formulation.
Do not confuse physical genealogy with recycled-content attribution
“This roll physically consumed regrind batch X” and “this product carries an attributed recycled-content percentage” are different claims. A chain-of-custody or mass-balance scheme may be relevant to the second. ISO’s cross-sector chain-of-custody update, which also applies to plastics, should not be read as proof that a site’s physical metre-level genealogy is complete, nor as a universal certification requirement for Thai plants. Define the numerator, denominator, trim and evidence for any percentage separately from feed-to-roll traceability. Label measured, estimated and certificate-based values distinctly.
Is roll-level extrusion traceability enough?
It can be a first stage if customers accept the entire roll as one decision unit. Local thickness, diameter or inclusion defects and narrower containment require a validated within-roll position map. If mapping precision is unverified, hold the full roll or another conservative interval.
Can resin lots be traced after regrind is blended?
Yes, as a documented candidate set. Link source rolls, regrind batches, quantities, feed events and residual/purge boundaries. Do not assign a single lot to a blended interval without physical evidence.
Can implementation start without machine data?
Manual scans can connect material, roll, QA and shipment. State the limits on setpoint-versus-actual analysis and metre mapping. Price sensor or PLC connectivity separately and test those limits at FAT/SAT.
What drives extrusion traceability cost?
Line count, PLC connectivity, encoder and gauge availability, labels and terminals, slitting, ERP integration, retention and Thai-language training. Compare proposals against the same traceable units and acceptance cases, separating essential scope from later expansion.
Summary
Extrusion genealogy is not a single “material lot → finished lot” arrow. It links virgin resin, regrind and additives, versioned recipes and actuals, transport from die to winder, within-roll intervals, splits and joins, quality decisions and consignee. Preserve candidate ranges and their evidence when a boundary is uncertain. Put site-specific exceptions in the RFP and prove both complaint-to-input and suspect-input-to-shipment searches at FAT/SAT.
Even if your Thailand plant is still deciding which feed, reel-change, slitting or inspection point to connect first, you can contact TOMAS TECH. We can review the physical units and signals with your team and turn them into an RFP and acceptance plan.
