You want to attach UHF RFID tags to metal parts and automatically record process movements and component lots. But choosing a tag by its advertised “maximum read range” can lead to missed reads on the actual metal surface, in stacked loads, in different transport orientations, or after washing. A successful deployment specifies what object is identified, where and how the tag is mounted, and what counts as a pass at each read zone. This guide takes manufacturers in Thailand from candidate selection through factory acceptance testing (FAT) and site acceptance testing (SAT).
Define the identification unit before choosing a tag
“Metal-part traceability” can mean tracking each product throughout its life, tracking a returnable container holding one lot, or locating a production fixture. Each requires a different design. A single tag on a box cannot follow individual parts after they leave the box. Conversely, if lot-level tracking is sufficient, attaching a durable tag to every small part adds mounting and replacement work without delivering the needed benefit.
At the start of the project, state whether the identifier belongs to the individual product, component lot, container, fixture, or equipment. Then draw one process map showing the tagged object from the as-cast blank through machining, washing, assembly, inspection, and shipment. If the tagged object moves to another container mid-process, record that transfer event. If tags are removed and reused, the procedure must first break the link to the old part before assigning a tag to a new one.
Here, “readable” does not mean that a handheld reader can detect the tag once at close range. It means that, at the required process step, the correct ID is captured as the object passes at normal speed, orientation, and density; nearby objects are excluded; and the correct event reaches the host system. If the respective roles of barcodes, QR codes, and direct-part marks are still undecided, first review our comparison of identification methods. This article focuses on making UHF RFID work on metal after it has become a candidate.
Why ordinary labels often fail on metal parts
A passive UHF RFID tag draws operating energy from the reader’s radio waves and responds. Metal changes the nearby electromagnetic field and the antenna’s operating conditions. When an ordinary thin inlay is placed directly on metal, the range quoted for free space or cardboard may not be reproduced. HID explains that proximity to metal reduces the performance of many tags, while on-metal tags use structures that create controlled spacing or are designed for that environment. Zebra also documents the performance loss of conventional labels on metal and the construction of metal-compatible labels.
Products labelled “on-metal” are not interchangeable. Options include thin on-metal labels, labels with a foam layer, resin-encapsulated hard tags, tags with screw holes, and suspended tags. Performance depends on the combined tag construction, distance from the metal, mounting location, and orientation. An “on-metal” claim does not imply unconditional suitability for a rough curved casting, a part just out of heat treatment, an oily surface, metal chips, wash chemicals, or stacked transport boxes.
A smaller tag fits in more places, but miniaturization can narrow the read margin. Check on the actual part that adequate margin remains for the required distance and orientation. Impinj’s selection guide likewise treats the mounting surface, attachment method, size, and required distance as separate considerations. Choose a position where the tag remains attached through the process and can be read at the intended point, not simply a place where it can be stuck to metal.

Put the requirements for UHF RFID tags on metal parts on one sheet
Before requesting quotations, prepare a specification for each part family. One row for all parts can conceal the hardest workpiece. Representative samples should include at least the smallest and largest parts, curved and recessed surfaces, different surface finishes, and parts exposed to washing or heat. Record the actual metal and surface condition used in the process, and retain the samples until testing is complete.
| Item | What to record | Why it matters |
|---|---|---|
| Identification unit | Individual part, lot, container, fixture | Defines the link between EPC and history |
| Target surface | Material, curvature, texture, paint, oil film | Affects tag construction and adhesive choice |
| Process environment | Temperature, washing, chemicals, impact, vibration | Affects durability tests and replacement intervals |
| Mounting constraints | Machined faces, sliding faces, fasteners, safety clearance | Limits mounting location and protrusion |
| Read scenario | Handheld, single-part fixture, conveyor gate | Changes required range and stray-read conditions |
| Data | ID scheme, lot relationships, write location | Prevents duplicate IDs and broken histories |
| Maintenance | Reattachment, replacement, disposal, reuse | Keeps the process workable on the shop floor |
Even with the same metal, a polished face differs from a rough one, a dry surface from an oily one, and a bare part from one in a plastic tray. Do not leave unresolved conditions to “supplier discretion.” State them as assumptions to test. For example, “include residual water droplets after washing in SAT,” “stack no more than two layers,” and “attach tags after heat treatment” make candidates comparable under the same conditions.
Work backwards from process interference to choose the mounting point
Avoid machined surfaces, sliding faces, fastening holes, and inspection contact points. Also avoid positions from which a detached tag could fall into equipment. If the workpiece always travels in the same orientation, the tag can face the reader. If it rotates, change the read-point arrangement or inspection procedure as needed; tag orientation alone may not solve the problem. A deep recess in a metal housing may protect the tag physically, but the surrounding walls can obstruct the RF path. Prepare at least two candidate locations that meet both safety and readability needs, and measure them on the actual part.
Choose the fixing method with detachment and replacement in mind
Adhesive labels are quick to apply, but oil, surface roughness, temperature, chemicals, washing, and curing conditions must be checked. Screws or rivets may make replacement easier, but assess whether holes are permitted, whether fasteners can loosen, the required torque, and interference from protrusions. Cable ties or suspended tags can create distance from metal, but are unsuitable if they can snag on rotating equipment or conveyors. Embedding can reduce detachment while increasing machining and repair costs. HID’s guidance on tag fixation emphasizes the combination of adhesive and substrate as well as position and orientation. The approved work instruction should cover cleaning, drying, application pressure, curing, the first read check, and who records it.
Do not make catalogue “read range” the purchase specification
A product sheet’s maximum read range is a value obtained under its stated test configuration. Results change with the real part’s material, tag position, reader power, antenna, distance, orientation, surrounding metal, and number of simultaneous tags. Zebra publishes upper-range figures for specific on-metal configurations; these cannot serve as a design guarantee for every factory.
Instead of specifying “a tag readable at three metres,” specify that the target workpiece must pass at a defined orientation and speed, its ID must be captured inside the designated zone, and IDs outside that zone must be excluded. Reading a loose tag on a bench is useful for initial screening. An adoption decision requires fixing it to the actual object and testing the plant’s real load arrangement and passage conditions. Repeated successes and failures, with their conditions recorded, matter more than the distance of one successful read.
Increasing range alone can also increase stray reads. If the reader captures the next aisle, a neighbouring tray, or goods waiting before the gate, fewer IDs may be missed while lot history becomes unreliable. The goal is a read zone that matches the business operation, not simply the longest range. When you need to design fixed-reader and antenna placement, power, polarization, and shielding in detail, see our guide to fixed readers and antennas. This article concentrates on the tag and its attachment to the workpiece.
Narrow tag candidates in three stages
First, eliminate candidates that cannot meet the basic constraints. A tag unsuitable for the temperature, wash process, chemicals, impact, required fixing method, or available space cannot be selected even if it reads well. Check guaranteed and test conditions in the supplier’s data sheet, and keep blanks as explicit questions. If a product merely says “heat resistant,” ask for exposure duration, number of cycles, and whether the claim includes the attachment material.
Second, mount candidates on actual parts and test static reads. Overlay the proposed length and thickness on the part drawing, temporarily fit tags in several locations, and compare normal and difficult orientations. Make changes one at a time: rotate the tag’s long axis, move it relative to a metal edge, or try a candidate with a spacer. Save photographs and equipment settings with each result. Changing several factors at once without identifying why performance improved makes production results hard to reproduce.
Third, test the real flow. Include plausible process conditions: adjacent parts in a tray, stacked returnable containers, conveyor speed, variation in stop position, parts just washed, and parts held by an operator. Treat the candidate tag, mounting point, fixing method, and reader settings as one “test configuration.” Record a new result row whenever that configuration changes. Compare candidates on read stability, installation labour, ease of replacement, and continuity of supply as well as price.

Keep individual IDs, lots, and process history distinct
A number written to a tag does not, by itself, create a quality history. Use the tag ID as a key in the MES or ERP to manage relationships with part number, serial or individual ID, production lot, material lot, process movements, and inspection results. The GS1 EPC Tag Data Standard defines EPC encoding and Gen2 tag memory content. If external trading or data exchange with other companies is expected, decide which GS1 identification key and encoding scheme to use first. Even a purely internal ID needs unique issuance, a no-reuse rule or a procedure to unlink it on reuse, an agreed length, and verification after writing.
EPC and TID also have different meanings. EPC is the value assigned to identify the object for business purposes. TID is information associated with the tag chip; it does not automatically represent the product’s lot number. When a tag is replaced, record the relationship between the EPC and new TID, or at least retain a history of disabling the old tag ID and assigning the new one. GS1 publishes tag data structures and identifiers managed under its standards. For internal identifiers, design issuance to prevent duplicates across the company’s other uses.
For example, if ten individually identified parts are made from material lot A, record in the process system that all ten IDs belong to material lot A. Update the container-ID relationship when the parts move to another container. When tracing a nonconforming lot later, what matters is not a large store of raw reads, but meaningful events: what was handled, where and when, and under which lot relationship. When designing the link to the material lot, use our guide to choosing traceability identifiers and state the identification unit clearly.
Make writing and reconciliation part of the attachment step
At attachment, reconcile the printed tag number, EPC, actual part label, and work order. Even if the writer reports success, reread the tag on the mounted part and check that the same EPC has not been used twice. Send unreadable tags, wrong EPCs, and labels whose adhesive has weakened during reapplication to a quarantine and reissue procedure, not back into the good-product flow. If duplication is first found at the end of the line, the team must reconstruct which workpiece should receive which history.
Be cautious about packing the lot, inspection values, and operator name into tag memory. That design requires decisions about capacity, write time, recovery from a wrong write, updates, and access control. In many applications, keeping a stable identifier on the tag and changing business information in the system is more flexible. If confidential data is written to the tag, design permissions and protection separately. GS1 explains Gen2 locking and memory protection mechanisms, but the existence of a feature does not by itself establish effective operational data protection.
Define the read zone as the space where reads are correct
A read zone is not a circle painted on the floor. It is a three-dimensional set of conditions including the relative positions of workpiece and antenna, orientation, and passage time. A station where one tagged part is placed in a fixture needs different tests from a gate through which boxes of multiple parts pass. The first must detect a wrong or missing part; the second must avoid both missed parts in the target box and stray reads from neighbouring boxes.
First, define the business event. Is a read meant to confirm “received,” “loaded into the machine,” “passed inspection,” or “shipped”? Define how several seconds of repeated reads become one event, what happens on a second passage or reverse movement, and how to cancel an event when an operator moves the part back. Sending raw RF reads straight to production records can make one workpiece appear to have gone through a process many times.
Next, draw target and exclusion zones on the layout. The defined load must be read in the target zone. Objects in the adjacent aisle or waiting area must not become business events. Sometimes an ID seen by the reader should become an event only when combined with a direction sensor or PLC conveyor signal. Reflections from metal can make actual RF paths differ from distance estimates made at a desk. Impinj notes that metal, dense arrangements, and surrounding structures affect tag power and performance. Measure in the plant and retain both the settings and installation layout.
FAT checks whether candidate tags work on the real part
Use actual metal parts and the proposed production tag in FAT. Do not approve a tag solely on a flat test plate. At a minimum, match the representative part’s material, surface, curvature, tag position, and fixing material. Before mass production, check that the attachment work instruction is repeatable and that EPC writing and reconciliation work. A test facility may not reproduce the site’s antenna layout exactly; record such differences and carry the unresolved checks into SAT.
The FAT plan should define failures as well as successes. “The target tag is not read,” “an out-of-scope tag becomes a business event,” “the tag detaches during washing,” and “a wrong EPC is not detected” are different faults. Reducing them all to one “read rate” obscures whether the remedy is a different tag, mounting point, or event rule.
| FAT aspect | Example test | Evidence to retain |
|---|---|---|
| Physical fit | Attach at the drawing position; check clearance from fixture, machining, and transport | Photos, dimensions, work instruction |
| Environmental durability | Inspect after process-representative washing, oil, heat, and vibration | Conditions, cycles, appearance, reread results |
| Encoding | Write EPC and reconcile it to print and workpiece | ID register, duplicate-check log |
| Static read | Repeat in normal and difficult orientations | Configuration, passes and failures, positions |
| Dynamic read | Run at process speed, density, and load arrangement | Raw log and confirmed events |
| Fault cases | Mix in a missing tag, wrong ID, or out-of-scope tag | Detection results, quarantine procedure |
The purchaser and supplier should agree numerical pass criteria before testing, based on process capability and site risk. Instead of an ambiguous phrase such as “always 100%,” define sample scope, repetition count, speed, placement, allowable errors, and retest conditions. For critical traceability, also test how a missed workpiece is routed to manual confirmation. If uncertainty remains, exclude the condition from production scope or use another identification method alongside RFID.
SAT checks whether reads become correct events in the real factory
SAT uses the installed readers, antennas, conveyors, metal equipment, and real operator movements. A tag that worked in FAT may behave differently near a rack, floor, or other gate, or when an operator holds the part differently. Check continuous running, shift changes, before and after washing, and both normal and crowded load arrangements. Raising antenna power to increase received reads can also increase stray reads in exclusion zones, so measure both.
The test log should retain time, target workpiece ID, lot, tag model and lot, mounting point, fixing method, read point, reader settings, test orientation, pass or fail, and fault handling. Photos or video help another shift reproduce the result later. Confirm that the event passed to MES or ERP carries the correct process, lot, and time. An RF log alone is not sufficient to close SAT.
Also test what happens when a tag fails in the working plant. Does an unidentifiable workpiece continue, or is it stopped and quarantined? If manual entry is allowed, who approves it and preserves the audit trail? If the same part receives a new EPC after tag replacement, how are the old ID relationship and incomplete process steps carried forward? These are operational rules more than equipment adjustments, but leaving them untested until the first day of production creates avoidable problems.

Compare deployment cost, not just tag unit price
Tags with the same unit price can require different cleaning and drying, spacers, printing and encoding equipment, fixture changes, installation time, reattachment rates, replacement labour, disposal, and spare stock. Fixed readers, antennas, cabling, networks, PLC signals, MES/ERP integration, FAT/SAT, and training are separate cost items. A cheap tag that takes longer to fit or detaches frequently after washing can cost more overall.
For a rough estimate, separate upfront cost, costs proportional to throughput, and maintenance cost. Upfront costs include site survey, proof of concept (PoC), equipment, and system integration. Volume-related costs include tags, attachment, printing and encoding, and inspection. Maintenance includes replacement, recalibration, spares, and change control. A permanent tag on every part and a reusable tag on a returnable box identify different units, so comparing only the lowest total in one table is misleading. See our RFID implementation cost guide for the wider investment picture.
Ask suppliers for several models for testing and confirm lot-to-lot performance control, production availability, alternatives, compatible printers/encoders, and replacement procedures. Testing a small number of the most difficult real parts before a large order reduces rework. Choosing a production tag only by a manufacturer’s maximum range or durability claim invites reselection when installation conditions change on the shop floor.
Ask suppliers to propose comparable tests in the RFP
To receive comparable proposals, attach the workpiece and process conditions to the RFP, not just a request for a tag model. Supply part photos and dimensions, materials, surface treatments, measured minimum and maximum temperatures, cleaning agents, impacts, load arrangement, conveyor speed, prohibited mounting locations, read-point drawings, and network and integration requirements. Share confidential drawings only to the extent needed, and invite suppliers to identify missing conditions.
Require each response to describe candidate tag construction and fixing method, the reasoning for its mounting point, the data sheet’s test conditions, an actual-part PoC method, proposed FAT/SAT pass criteria, exclusion-zone controls, the procedure for missing tags, and post-delivery replacement and supply. A bare “read range of X metres” does not establish factory suitability. Ask suppliers to state their assumptions about part count and layout, simultaneous tag reads, and who owns the tag attachment step.
Agree change control before contracting. What tests must be repeated if a part’s surface finish changes, the line moves, an antenna is added, the washing agent changes, or the tag supplier proposes a substitute model? Define who approves the change and when production switches over. RFID problems can arise during process changes as well as immediately after installation.
Include a decision to stop deployment in the design
If a metal part is too small for a safe mounting surface, suitable heat- or chemical-resistant tags cost more than the process justifies, or densely packed parts cannot be distinguished at the required speed, reconsider the identification unit or read step. A returnable container tag linked to a part lot, an identified fixture, or a direct mark or 2D code at selected points may preserve the traceability facts you need. Using RFID is not itself the goal.
A staged rollout should include difficult representative conditions early, not only the easiest parts. If a demanding workpiece fails in the initial PoC, decide whether the tag, mounting point, read point, or identification unit must change before expanding production scope. Retaining both passing and failing conditions helps prevent accidental rollout to unsupported parts.
Define the documents for production handover early
Production cannot reproduce a PoC if only the engineer who ran it knows the settings. The handover package should include the approved tag model and revision, photos of the attachment surface, prohibited positions, preparation and fixing method, encoding format, writer settings, read-point layout, power and filter settings, test results, and fault contacts. List what each process change could affect, and apply a revision number and approver. Even a tag-only substitution needs a comparison test: writing the same EPC does not guarantee reading it at the same point.
Use photos in the shop-floor work standard to show which face, position, and orientation to use. Separate a read immediately after attachment from a read after demanding steps such as washing or heat treatment; this helps distinguish poor fixing from changed RF conditions. Define a substitute label, quarantine location, and re-registration approver if parts must temporarily travel without tags. A process that prevents untagged workpieces from silently continuing protects the history’s reliability.
Maintenance staff should inspect more than broken tags: dirt at the read point, antenna orientation, damaged cables, and added neighbouring equipment can all matter. Set an inspection interval, such as monthly, according to operating conditions and retain a representative-workpiece check procedure. On the data side, monitor counts of missed reads, duplicates, out-of-scope IDs, and manual corrections. Investigate changes in both RF and work procedures. A good read rate cannot compensate for events linked to the wrong lot.
What to agree when starting small
The first deployment may limit part types and process steps, but it should not remove every hard representative case. Before the PoC, agree what will be measured and what results mean “adopt,” “redesign,” or “stop.” Changing sample counts or test repetitions after seeing the results makes candidate comparison unreliable. When a test fails, classify the cause as tag damage, detachment, RF, ID writing, or business-event logic. Change one factor and retest.
At the production-release review, confirm unresolved FAT/SAT items, exception-test results, supplier warranty scope, operator training, spare tags, and reissue procedures. Freeze the approved tag model + mounting point + fixing method + reader settings + business rules as one configuration so a new team can apply the same conditions. Approving the part side and reader side separately undermines repeatability.
Checks when changing a component lot
The relationship between a tag and a part is more likely to break during a changeover than during routine production. Plan for cases where the last pieces from a material lot remain on the line as parts from the next lot enter, a returnable container is reused while still associated with the previous lot, or only the tag printer job is left on the old setting. At changeover, reconcile the unfinished quantity from the old lot, container-to-part relationships, issued but unattached tags, and IDs of discarded tags. Record what was confirmed at the time instead of trying to infer it from logs later.
When a part receives a replacement tag, its earlier inspection history must remain linked to the new ID. If the old tag can still be read, capture and deactivate its ID, then transfer the history to the new one. If the old tag is damaged and unreadable, identify the physical part through other evidence such as a direct mark or work order, record the approver, and reissue the ID. An assignment based on a guess can leave two histories for the same part during a later nonconformance investigation. Demonstrate this reissue path at least once in FAT/SAT so the emergency procedure works beyond the written document.
Frequently asked questions
Which UHF RFID tag should I choose for a metal part?
Use an “on-metal” designation as a starting point, then narrow candidates by the actual part’s material, geometry and surface, attachment method, temperature and washing exposure, and required read zone. Decide between a thin label and a hard tag through mounted-part and post-process testing. A model name or maximum range alone is insufficient.
Can I use an ordinary tag if I add a spacer?
Spacing can improve performance, but the spacer thickness and material, attachment strength, safety of protrusions, and read orientation create a new configuration that must be validated. Compare it on the real part with a purpose-built on-metal tag. Do not assume that adding a spacer to a standard tag meets production requirements.
What read percentage should be the acceptance criterion?
There is no universal percentage. Set it according to the effects of a missed read on production or recall traceability, the availability of manual recovery, and the risk of stray reads. Agree the load arrangement, speed, repetitions, target and exclusion zones, and fault handling before testing, and retain them with the measured results.
Is EPC alone enough to trace material lots?
EPC is an identifier; its relationship to the material lot must be recorded separately. In MES/ERP, manage whether the EPC identifies a part, container, or fixture; which individual parts came from a material lot; and how relationships change when tags are reused or replaced.
If FAT passes, can we skip SAT?
No. FAT checks actual parts and repeatability of the proposed configuration. SAT checks business events amid the installed metal equipment, operator routes, neighbouring zones, and host systems. Carry conditions that could not be reproduced in FAT into the SAT plan.
Conclusion
Selecting UHF RFID tags for metal parts involves more than picking an on-metal model. The deployable specification connects the identification unit, attachment to actual parts, process durability, read zones, ID-to-lot relationships, and FAT/SAT evidence. Start with representative workpieces and a process map, test candidates under the same conditions, and agree how exceptions will be handled.
If you have parts and process drawings and are defining tag locations or FAT/SAT items, contact TOMAS TECH. We can help scope an actual-part PoC and define read-zone and lot-integration requirements.
References
- GS1 EPC UHF Gen2 and Tag Data Standard — Standards for the UHF air interface and tag data.
- GS1: How data is stored on a RAIN RFID tag — Roles of EPC, TID, and other memory.
- HID RFID FAQ — Performance near metal and mounting methods.
- HID Adhesive Tag Fixation — Considerations for adhesion, location, and orientation.
- Impinj: Choosing a RAIN RFID Tag — Material, mounting, size, and range considerations.
- Impinj AutoTune — Effects of metal, dense arrangements, and other conditions on performance.
- Zebra RFID On-Metal Tagging Solution — Example on-metal label and print/encode configuration.
Product specifications and regional applicability in public materials can change. Before adoption, confirm current data sheets for candidate models and the applicable conditions in Thailand with the supplier.