Retrofitting torque sensing onto an existing FMCG packaging line is governed by the same five-phase commissioning logic as any new asset: a Factory Acceptance Test at the sensor vendor's facility, mechanical and electrical installation verification on site, a Site Acceptance Test, performance validation against URS, and an OEE/PM baseline with full traceability [S1].
The retrofit differs from greenfield in one operational respect: the line must stay producing, so most cap-torque evidence is collected at line speeds up to 1,200 units per minute, with applied torque estimated per capper head from cap height profile or read directly from instrumented spindles for precision-critical SKUs [S2].
Phase 1: FAT Documentation and Compliance Evidence
The first certification checkpoint is a complete document pack handed over at FAT and tied to the asset record from PO issuance: P&ID mark-ups, electrical schematics, sensor data sheets, material certificates, calibration certificates, CE/UL declarations, and the IEC 62381:2024 compliance matrix for automation systems integration [S1]. For a torque sensor specifically, the FAT pack must include the ISO 6789-1 calibration certificate traceable to a national standards body (NIST, NPL, PTB or equivalent), the proof of class-1 or class-2 accuracy, the test report showing the sensor's combined non-linearity, hysteresis and repeatability as a single percentage of full scale, and the EMC test report covering EN 61326-1 industrial environments [S3].
Missing certificates are a FAT hold point, not a punch-list item, and no release for shipment is signed off until the torque sensor's documentation set matches the Purchase Order line-by-line [S1]. A practical data point: fixing a defect identified at FAT costs roughly 1× the defect cost, while the same defect at SAT runs 10× and during production 50–100×, which is why the calibration certificate and EMC report must be in the FAT pack, not chased after delivery [S1].
Phase 2: Mechanical and Electrical Installation on an Operating Line
Installation verification on a retrofit line has two failure modes greenfield does not: the sensor must mount to an existing capper head without altering the capper's mechanical stack-up, and the cabling must route through existing cable trays without crossing VFD power runs that would inject common-mode noise into the low-level torque signal. Dimensional checks against the as-built capper drawing, with focus on the adapter plate concentricity (typically held under 0.05 mm runout to avoid bearing-load side effects on the strain-gauge bridge) and the connector pin-out verified against the wiring diagram, are the same hold-point items as a new build [S1].
For the electrical side, the rules diverge from greenfield: a retrofit must prove galvanic isolation (commonly 1,500 V AC for 1 minute per EN 61326-1) and shield grounding at the panel end only, with the sensor-end shield isolated, because the existing grounding scheme cannot be disturbed without re-certifying the whole panel [S3]. The mechanical-to-electrical handoff is a documented FAT-to-SAT gate: the physical install team cannot close the gate until the electrical team has logged shield continuity, supply voltage under load (typically 24 V DC ±10%), and a clean 4-20 mA loop reading at zero torque using a torque sensor tester or equivalent reference transducer [S3].
Phase 3: SAT, Performance Validation and Per-Unit Traceability
SAT is where the torque sensor proves it sees what the URS said it must see, and on a packaging line that means proving the applied torque value for every closure within a defined tolerance window, typically ±5% of the target torque for plastic-threaded caps and ±2% for metal caps on glass, with each result logged with a unit serial, capper head ID, timestamp and batch code [S2]. A representative retrofit case published in March 2026 documented a personal-care bottling line where induction sealer head 4 drifted 3°C over eight hours; the equivalent failure mode on the capper side is a spindle clutch that loses grip, and a SAT that does not include a per-head torque drift test across at least one full shift will not catch it [S2].
For retrofit acceptance, the SAT protocol must include a 30-minute minimum run at rated line speed with all capper heads active, comparison of each head's mean applied torque against a calibrated reference torque wrench on at least 30 pulled samples, and verification that the data export (OPC-UA, MQTT, or REST, depending on the line SCADA) lands in the historian with sub-second time-stamp resolution [S2][S3]. A camera-and-AI overlay can then estimate applied torque from cap height profile as a redundant channel, with vision inference completing in under 50 ms per unit at line speed, so neither the mechanical sensor nor the vision channel is a single point of audit failure [S2].
Phase 4: Standards Mapping and Audit-Ready Evidence
The audit evidence chain for a packaging-line torque retrofit typically spans four standards, and the certifier will trace each requirement to a deliverable: ISO 6789-1 covers the torque tool/sensor calibration, EN 61326-1 covers EMC, IEC 62381:2024 covers automation systems integration, and the customer-specific quality standard (BRC, IFS, or a retailer code of practice) covers the per-unit traceability record [S1][S2][S3]. For zones where the capper sits near a solvent or alcohol-based product filler, ATEX 2014/34/EU or IECEx certification may also apply if the sensor is mounted inside the classified area, and the FAT pack must then include the safety certification document plus the EU type-examination certificate number [S3].
Comparison of the three common evidence channels a retrofit project can present to an auditor:
Instrumented spindle (strain gauge on the capper head): highest accuracy, typically ±1% of full scale, requires line stop for installation, costs roughly 3–5× a vision-based retrofit and is the right choice for precision-critical SKUs or metal-on-glass lines [S2][S3].
Vision-based cap height inference (AI camera looking at cap profile post-closure): no mechanical contact, installs in under one shift, accuracy in the ±5–8% range depending on cap geometry and lighting, and suits FMCG plastic-threaded caps where the tolerance window itself is wide [S2].
Inline rotary torque transducer on the capper drive train: middle ground on accuracy (typically ±2% of full scale), fits between the servo and the capper chuck, and is the common pick for retrofitting a servo-controlled capper without disturbing the capping turret [S3].
Phase 5: Failure Modes Specific to a Retrofit
Four failure modes recur on retrofitted lines and rarely on greenfield: (1) the existing capper's clutch wear masks a sensor that reads correctly, leading to false-pass at SAT; (2) the existing cable tray routing injects VFD common-mode noise, which manifests as torque-reading jitter that looks like process variation; (3) the historian's time-stamp resolution is too coarse (1 s or worse) to correlate torque spikes with capper servo faults; and (4) the OEM's calibration certificate was issued for a standalone sensor, not for the sensor-plus-adapter-plate assembly actually mounted on the capper head, invalidating the traceability chain [S1][S3].
Each of these has a checklist countermeasure: clutch verification under simulated load at FAT, shield-continuity and earth-bond checks at install, OPC-UA or equivalent sub-second historian integration at SAT, and a system-level calibration at the assembled mounting point using a torque wrench tester as the reference, with the system-level certificate stored in the same folder as the OEM certificate [S1][S3].
Verifying a Supplier's Certificate is Real and In Scope
A torque sensor certificate from a vendor is only as good as the issuing body's accreditation: the calibration lab must be ISO/IEC 17025 accredited, the accreditation scope must explicitly list torque in the required range, and the certificate must carry the issuing lab's accreditation logo and number cross-checkable on the national accreditation body register [S3]. The sensor's safety certification, if claimed, must be checked against the EU notified body database for ATEX or the IECEx OD database for IECEx certificates, and the certificate number's scope must list the exact model designation, the Ex marking string (e.g. Ex db IIC T4 Gb), and the ambient temperature range actually present on the line [S3].
For the packaging-line context specifically, the certifier will also want to see the packaging machine builder's declaration that the retrofitted torque sensor does not invalidate the original CE/UL declaration on the line, or a new declaration covering the modified assembly, and a documented risk assessment covering the change under ISO 12100 [S1].
Comparable Retrofit Touchpoints and Cross-References
Torque sensing is one of several instrumentation retrofits that follow the same FAT-to-SAT-to-evidence pattern; a parallel retrofit on the same line is AI vision inspection for seal and cap integrity, which runs on an edge NVIDIA appliance, achieves 99.8% defect detection accuracy, and verifies seal integrity, cap presence, tamper evidence and closure torque on every unit at full line speed [S2]. Adjacent retrofit specifiers will also want the vacuum packaging machine sealing-bar temperature map and the packaging material lot traceability scheme lined up against the same SAT gate, because the auditor will read them as one evidence pack, not separate projects [S1].
For engineers specifying the spindle side of the retrofit, the choice between ceramic-hybrid bearings and standard bearings in the capper turret interacts directly with the torque sensor's noise floor, and the 2026 ceramic bearing spec map is a useful adjacent read for sizing the adapter-plate stack-up. Where the retrofit also includes a vision overlay for torque estimation, the industrial camera spec map for surface finish requirements gives the lighting and lensing floor for cap-height inference to hit the ±5% torque window.
Next checkpoint: lock the SAT protocol around a 30-minute full-speed per-head torque drift test, demand the system-level ISO/IEC 17025 calibration certificate covering the sensor-plus-adapter assembly, and require the historian to log torque with sub-second time-stamp resolution before the SAT sign-off is signed [S1][S2][S3]. Trackable signals over the next quarter: whether the capper OEM issues a torque-sensor retrofit kit with a pre-validated CE delta-declaration, and whether the IEC 62381:2024 update introduces a specific annex for retrofitted instrumentation on packaging lines [S1].