Riveting machine selection cannot start and end with maximum forming force or a catalog cycle time. The joint may also need a defined pivot torque, retention strength, appearance and traceable result. This guide helps assembly plants in Thailand compare radial, orbital and direct press forming on the same parts, then connect the choice to an RFP, sample trials, factory acceptance testing (FAT) and site acceptance testing (SAT). Set actual tolerances and guarantees from the product design, material trials and an agreement with the equipment supplier.
Start riveting machine selection with the job of the finished joint
Riveting or staking plastically forms the projecting end of a rivet or tenon so that the mating parts remain together. An identical visible head diameter does not mean two joints perform alike. A fixed joint may require resistance to push-out, pull-out, play or rotation. A pivot must retain the components yet meet a specified running torque. Some products also prohibit a mark on the visible face. Define what the assembled product must do before reading a machine specification.
Provide a drawing and physical samples showing rivet material, hardness, shank diameter and projection, mating material and thickness, hole diameter and alignment, stack height, available support surfaces, permitted marks and joint functions. For the process, state orientation, how the part enters a nest, changeovers, required good output, inspection and reject handling, data destination, operating schedule and replenishment or cleaning tasks. If a drawing merely says “rivet” without retention or pivot criteria, suppliers will quote machines against different assumptions.
Orbitform discusses how to match an assembly method to functional requirements such as retention, articulation and finished height. This is a useful engineering framework, not a guarantee for a particular product. Product engineering should define the test, quality should review the decision rule, and manufacturing engineering should translate it into equipment settings.
Do not confuse processes that share the word “rivet”
Radial and orbital forming in this article shape the end of an already positioned rivet or tenon. Self-pierce riveting is a distinct process: the rivet pierces the upper sheet and forms an undercut in the lower one. Böllhoff’s description of self-pierce riveting shows why tooling, hole preparation and joint sections differ. A blind-rivet tool or impact riveter is different again. Do not treat their catalogs as interchangeable merely because all say “riveting.”
Compare radial, orbital and direct press forming on the same joint
BalTec’s forming and monitoring guide describes a rosette-like peen path for radial forming and a circular path for orbital forming. It identifies gentle forming and relatively low force as radial characteristics, and short cycles in simpler applications as one orbital characteristic. These are the manufacturer’s general comparisons, not a claim that radial is best for every part. Test with the same material, rivet projection, intended head shape and supporting fixture.
In Orbitform’s orbital forming explanation, an angled peen makes a sweeping contact line and progressively forms the head at room temperature. Peen geometry determines the final form; the manufacturer recommends trials with the actual application to establish the required force. A direct press applies axial force to displace material. Orbitform’s press-forming description presents it as an option where hole fill or contact area is central to the joint. For thin components, sensitive bearings or strict cosmetic surfaces, examine how each head and its fixture load the real part.
Use common test pieces and common evidence rather than comparing machine names alone.
| Evaluation | Put in the specification | Check in the trial |
|---|---|---|
| Formed head | Height, diameter, eccentricity, cracking, surface | Dimensions, section, magnified images |
| Fixed joint | Push-out, pull-out, slip and play | Suitable destructive and non-destructive tests |
| Pivot joint | Breakaway and running torque, axial play | Immediately and after specified durability |
| Part loading | Indentation, distortion, bearing load | Local dimensions, appearance and function |
| Process | Cycle, changeover, tool life, recovery | Continuous run under real feeding conditions |
The table does not impose universal numbers. Derive retention and torque limits from the product function. Tightening a pivot simply to increase retention can make it too stiff. Establish a processing window that satisfies all relevant limits at once.
Include part variation in the process comparison
Projection, hole size, stack thickness, material hardness, plating, lubrication and rivet lots can change the result at an unchanged machine stroke. Trials at nominal dimensions alone may miss head cracking or excessive clamp load at the tolerance edges. Make physically realistic samples at upper and lower permitted limits. If a full combination matrix is impractical, engineering, quality and the supplier should select the combinations most likely to fail and record why.
Fixture stiffness and the reaction-force path must also stay comparable. Changing the forming head while supporting a thin plate far from the rivet makes it hard to distinguish process effects from fixture effects. Fix the datum surface, clamp points, support beneath the rivet and removal path, and document the setup used for every trial. The production machine should detect whether the component is seated and stop before forming a reversed or floating part.
Connect visible quality to joint function
A head may meet height and diameter tolerances while the shank fails to provide retention. At the other extreme, over-forming may make a hinge too stiff. Treat appearance, geometry, function and process signals as separate evidence. Geometry can include height, diameter, concentricity and part distortion. Functional evidence may include push-out or pull-out force, pivot torque, play and a specified result after durability. Destructive tests cannot inspect every part, so establish correlation during development and an appropriate sampling plan in production.
Orbitform’s process intelligence guide describes logging force and distance against process limits. A force-displacement curve within a programmed window, however, is not itself proof that every finished joint meets its retention requirement. Process signals are proxies that may reveal tool wear, incoming variation and unseated parts. Test known good and known defective samples to learn which feature detects which defect. Keep another inspection where the signal cannot identify a failure mode.
A good-part window is more than a band around one average trace
First test the joint’s critical functions, then collect traces linked to passing and failing samples. Include tolerance edges, new and worn tools, temperature change, changeovers and restart after a stop. The result depends on whether the horizontal axis is time or stroke and whether alignment begins at first tool contact. Specify sampling, filtering, sensor calibration and trigger location in the RFP.
A window may use contact position, force rise, force at a defined position, final position, dwell time and abnormal stop status rather than maximum force alone. Explain which feature relates to which failure mode. If the rule becomes too complex for operators to diagnose, the monitoring system will be difficult to maintain. Do not repeatedly form a bad part just to make its trace fit the window; decide in advance whether rework is permitted, how often and how it will be identified.
Define what “trace every part” means
Where traceability is needed, specify the unit of record: product or lot ID, recipe version, machine and tool IDs, timestamp, result, key force and stroke values, reject reason, retries and maintenance intervention. For a network outage, test local buffering and duplicate-safe replay. SAT should reconcile physical reject confirmation with the event sent to MES, so that a recorded OK does not hide a part in the reject bin.
CSV export alone does not establish useful traceability. Decide whether the cycle can start without an ID, whether recipe changes are logged, whether a rejected trace can be retrieved and how long records remain searchable. The related custom machine design and build guide explains boundaries between feed, control, inspection and information. When a riveting station becomes part of a line, contractually assign responsibility for the record and physical segregation of rejects.
What to put in a riveting machine RFP
An RFP gives all bidders the same starting conditions and asks for comparable responses. “Radial machine, target cycle and quality monitoring” leaves the forming head, feed, fixture, safeguards, inspection and data scope open to interpretation. Even at concept stage, label unknowns with an owner, a sample and a date for resolution. Separate product performance specified by the buyer from the machine configuration proposed by the supplier.
Product, material and process inputs
Supply part numbers and likely future variants, assembly and component drawings with revisions, critical dimensions, tolerance stack, material and finish, rivet presentation, orientation rules and prohibited quality defects. For each required function, state the measurement method, point, speed, instrument and sampling rule. If customer specifications include durability, humidity or performance after transport, explicitly state that an in-process signal does not replace those tests. Add a gate to repeat trials on production material if early samples differ.
Scope, throughput and changeover
Show whether the offer includes manual loading, semiautomatic feeding or full automatic feeding; a rivet feeder and no-rivet protection; part presence, orientation and seating checks; forming; inspection; OK/NG discharge; data integration; downstream transfer; and fixture changes. Define the cycle-time boundary to include actual feeding, loading, clamping, judgment and discharge, not only the peen’s forming time. Specify the changeover time, people needed, recipe access and wrong-recipe prevention.
Communicate the required good output using the product mix, operating hours, planned stops, replenishment, tool changes and rejects. There is no universal seconds-per-part value: forming volume and feeding design change it materially. A short catalog machine cycle is not the plant’s guaranteed hourly output of good assemblies.
Begin with a semiautomatic station or integrate the line?
A manually loaded station may suit a high product mix, difficult orientation for automatic feeding or a product design that is still changing. It still requires safeguards and mistake prevention because the operator works near the forming point. Verify part orientation, rivet presence, seating and agreement between product and recipe, and make it difficult to return NG parts to OK flow. Measure the real cycle including loading and inspection. If one nest accepts several variants, identify which adjustments exist and who may change them.
Line integration adds handoff, part ID, stop signals, reject destination and network-failure behavior. If the rivet feeder stops, should the upstream process stop or send parts to a buffer, and how is an unprocessed part prevented from reaching downstream? Even if the riveting head meets its cycle, a transfer that disturbs part orientation can reduce good output. Use an interface table that describes not just signal names but the physical state each signal represents. “Process complete” could mean the head has formed or that inspection passed and discharge was confirmed; those are different promises.
A phased installation may reserve space and controls for later line integration. Quote cabinet capacity, communication method, additional sensors, software changes and future safety reassessment as separate items so they are not mistaken for current scope. Adding a feeder to a manual station does not automatically make the operation fully automatic; the feeder must replace orientation and defect judgments previously made by a person. The related guide to semiautomatic machines can help identify decisions that will remain with the operator in this riveting process.
Safety, maintenance and site conditions
ISO 12100:2010 sets out a method for hazard identification, risk assessment and risk reduction. ISO 13849-1:2023 covers design and integration of safety-related control-system parts, while ISO 13850:2015 covers emergency-stop principles. Naming these standards on a quotation does not certify a particular riveting machine. Assess crushing and trapping in loading, changeover, jam clearing, cleaning and maintenance, then design and validate guards, interlocks, safety controls and residual-energy handling.
Check the applicable edition and scope of IEC 60204-1 for machine electrical equipment. For a Thailand installation, give the supplier the power and grounding conditions, air quality, ambient conditions, delivery route, maintenance clearances, spare-part expectations, required languages and local response scope. Confirm applicable local law and customer safety rules for the project. Two-hand actuation in normal operation does not by itself resolve hazards during setup or jam clearing.
Ask bidders to answer in one common format
Request the proposed method and why it was selected; how the fixture carries reaction forces; head and tool concept; candidate process range; feeding and detection; reject segregation; safety functions; data fields; buyer-supplied items; guaranteed functions; test plan; and exclusions. “Can support” is not evidence for an untested function; ask for a trial plan, price and timing. Accept alternative concepts, but compare each against the same mandatory baseline.
Break quotations into machine, fixtures, tooling, rivet feed, gauges, controls, network, spares, installation, FAT/SAT, training and maintenance. The related automation quotation comparison guide explains why an excluded item should not be treated as zero cost. Include extra tools and nests for each rivet and product variant.
Make sample trials more than a supplier demonstration
The purpose of a trial is to find a process window that meets every product function across permitted variation, not to make one attractive sample. Use candidate production rivets and components, recording material lot, dimension, treatment and storage condition. A successful supplier-lab sample may behave differently with the real feeder, fixture, operator and environment. Ask for tool drawings, recipes, measured input parts, process traces, functional results and joint sections tied to each sample ID.
Combine nominal conditions, tolerance edges, difficult material combinations, required changeovers and the effects of continuous running on temperature and tooling. Set sample sizes and acceptance limits from product risk, customer requirements and the statistical confidence the project actually needs. A fixed sample count or capability index from a generic article would be a baseless guarantee. Keep results before and after a recipe change separate.
Build a trial matrix with input, process and outcome layers
Give each sample an ID. Record rivet diameter, projection, hardness and lot; hole size, stack thickness, finish, lubrication and fixture revision as inputs. Record method, tool ID, recipe version, set force, stroke and time, measured trace and stop status in the process layer. In the outcome layer record head size, cracking and tilt, retention force, pivot torque, play and part distortion as relevant. Linking all three layers by sample ID lets engineers work backward from a part whose head passed but whose pivot torque is too high. Even after a destructive test consumes the sample, retain its ID, photograph, section and trace.
Changing projection, peen shape and stroke simultaneously after a cracked head appears may produce a good sample, but it does not reveal which change fixed the problem for the next material lot. Separate single-factor exploration from interaction tests. For the production qualification run, lock the actual tool, fixture and recipe to be released. Confirm that the supplier laboratory’s instruments, calibration state and measurement methods match the plant’s acceptance measurements.
There may be several “worst” conditions. A hard rivet with a short projection might be difficult to form, while a long projection and soft material might risk an overflowing or cracked head. These are hypotheses to test, not universal outcomes. A thin mating plate supported at the wrong location might warp, and an over-formed bearing pivot might run too stiff. For each hypothesis, specify which input goes to its boundary and which output will be measured. If a drawing-limit sample cannot actually be made, send the discrepancy between drawing tolerance and process capability back to product engineering.
Check variation in the measurement itself
When a trial result lies near an acceptance limit, do not confuse product variation with measurement variation. For head height measured by a contact gauge, the datum face and choice of crown or rim contact point affect the reading. For pivot torque, align start angle, speed, treatment of the breakaway peak, running value and rest time before test. For push-out, specify punch diameter and alignment, support hole, loading speed and failure endpoint. Confirm that different inspectors reach the same conclusion and define retest and quarantine rules for borderline parts.
An instrument calibration certificate does not by itself validate the measurement method including the workpiece. Thin plate flexure under the gauge, an asymmetric head that changes the contact point or lubrication that changes torque can matter. Compare the supplier lab method with the buyer’s shop-floor acceptance method before FAT. For borderline results, keep each reading, condition and retest history rather than accepting an average alone.
Safely simulate foreseeable faults such as over-forming, unseated parts, missing or reversed rivets, double feed, damaged tools and sensor disconnection. Check where the machine stops and how affected parts are isolated. Clearly segregate deliberate bad samples from saleable parts. Retain an inspection method for defects that process signals cannot find.
FAT checks the built machine; SAT checks the installed process
At FAT, verify the contracted configuration, forming, inspection, cycle, safety, abnormal recovery, data output and maintenance at the supplier’s site. At SAT, verify actual power, air and network, real part flow, upstream and downstream processes, operators, reject destinations and the record system at the plant. Use shared test IDs and record why a test that cannot be completed at FAT moves to SAT. “Passed FAT, so only power it up on site” can miss wrong reject routing or behavior during a network outage.
Each case needs preconditions, part revision, actions, expected result, measurement, evidence, adjudicator and open issues. Measure head geometry and joint function using agreed instruments; FAT and SAT torque values are not comparable if the method changes. Feed normal and known faulty parts to verify that classification, reject movement and records agree. Verify safety functions on the real machine against the risk assessment, including emergency stop, guard opening, restart, jam clearing and power restoration where relevant.
Accept each fault by testing detection, stop, quarantine and restart
For a missing rivet, check more than whether the feed sensor reports absence: can the machine prevent an empty stroke, stop the cycle and identify the workpiece as unprocessed? For double feed, test feed-path status, clearing instructions and first-off inspection after recovery. For an unseated part, consider a check before clamping or forming rather than relying on a force trace after the damage. If the mechanism cannot detect it, document the residual risk and additional inspection.
A chipped or worn peen may leave an apparently acceptable head while changing joint function. Decide who orders replacement, how far back to quarantine products, and what test permits restart. A disconnected sensor must not let a zero reading pass as normal; test the diagnostic and machine stop. If MES is unavailable, check local records and physical segregation, then verify that replay pairs each product ID with exactly one outcome. An alarm-screen photograph alone is insufficient acceptance evidence. Reconcile the physical sample, machine log and destination record.
Recovery itself is a quality test. When the guard closes and start is pressed, does a part interrupted mid-forming get automatically formed again, rejected, or held for an operator decision? Specify the answer. A reset should not erase the earlier NG event, and physical organization should prevent accidental placement in the OK bin. On restoration of power, verify datum, tool position, unfinished parts and local data consistency. A test simulated at FAT should be repeated against actual upstream and downstream interlocks at SAT.
Put changes to acceptance criteria under control
If the functional limit or measurement method changes before FAT, agree who proposes it, whether product engineering or quality approves it, and how price and schedule are handled. Quietly relabeling a failed trial “provisionally accepted” invites the same dispute at SAT. Keep the former rule, reason for change, affected part numbers, retest scope and approvals together. A supplier’s machine correction and a buyer’s change to product requirements carry different responsibilities; distinguish them in the change record.
A FAT video may show sequence but not measured values. A score sheet may show values without proving that the sample and fixture match the contracted revision. Link each test ID to drawing revision, sample ID, machine and tool settings, measurements, logs, images or video, decision and sign-off. Update that package at SAT for changed utilities, communication and operator methods, while also reconfirming product function. Production release needs more than a signature: check work instructions, training, maintenance, spares, calibration, reject quarantine, recipe access and data ownership. During early production, review head geometry, function and their relationship to process signals; set the duration and exit criteria from product risk and customer requirements.
A continuous run should include cleaning, rivet replenishment, variant changes, quality checks and recovery. Agree beforehand which stoppages count and which external causes are excluded. Do not close a fault just because resetting the machine lets it run again: log cause, affected parts, correction and retest. If an open item remains at handover, distinguish safety or quality blockers and agree an interim control, owner, deadline, payment treatment and warranty treatment.
Control recipes and tools after production starts
Tool replacement, rivet lots and new variants can move the process window even after a stable launch. Link part number, tool, nest, forming settings, trace-window version and inspection version in each recipe. Define request, approval, test, release and rollback. Widening the signal window every time rejects rise defeats process monitoring. First examine incoming dimensions, seating, tool condition, sensors, air and machine stiffness.
Compare the burden during production, not only the purchase price
Over a common time horizon, compare initial cost alongside scheduled tool changes and regrinding, fixtures for added variants, feeder cleaning, sensor calibration, gauges, spare parts, maintenance support and rejected-part handling during launch. Electric, pneumatic and hydraulic approaches have different utility and maintenance needs; do not infer a universal operating cost from the method name. Ask suppliers for consumable part numbers, replacement criteria, local stock, lead-time terms and breakdown support. A low-priced unit whose replacement tooling is slow to obtain can carry greater downtime risk.
Before crediting labor or scrap savings, measure current work, defects and stoppages and assign an owner to each forecast. Eliminating manual staking may still leave rivet feeding, sample inspection, changeover and fault recovery. A shorter machine cycle increases line output only when upstream and downstream capacity and demand allow it. Calculate improvements that the station can reliably deliver separately from those that depend on the whole line. Adopt quality or uptime targets only after aligning the contractual measurement boundary and evidence.
Set tool replacement from the material formed, cumulative cycles, trend in head and functional results and tool-face wear rather than applying a supplier reference life blindly. After a tool change, verify the datum, first-off product and trace baseline, and identify the first-off units. Link tool and fixture identity to part lots where needed so an investigation can narrow the scope. Maintenance planning should also cover sensors, anvils and feed components.
When a quality issue appears, the priority is defining its affected population. Connect the last confirmed good part, tool changes, recipe changes and sensor alarms to the lot quarantine. Unless product engineering has explicitly approved rework, do not return a twice-formed part to normal flow simply because its appearance improved; its internal condition may have changed.
Common selection mistakes and their remedies
“The highest force capacity is safest” ignores thin-part loading, lower-fixture reaction forces, control resolution and the useful window. Prove function and machine margin on actual parts. “A compliant head diameter means every joint is good” ignores retention and pivot behavior; correlate geometry, function and traces. “Monitoring means zero defects” ignores undetectable faults and sensor failures; define detection limits and supplementary inspection.
“SAT is unnecessary after FAT” ignores plant power, feeding, discharge, MES integration and actual operator recovery. Keep site acceptance conditions in the contract. “Choose the cheapest quotation” may omit tools, nests, guarding, gauges or maintenance. Align scope line by line. These gaps cannot be solved by assigning every responsibility to the machine builder. The buyer must own the product’s acceptance criteria and connect them contractually to what the equipment supplier will deliver and warrant.
Summary: connect joint function, process, machine and acceptance evidence
Define what the finished joint must do, provide production-representative parts including tolerance edges, compare radial, orbital and direct press methods with common tests, establish how process signals relate to final function, and lock the evidence into the RFP and FAT/SAT. This order makes responsibilities for retention, articulation, appearance, safety and records explicit.
Even while moving from manual staking to equipment or comparing early supplier concepts for a Thailand plant, you can discuss part conditions, RFP structure and trials with TOMAS TECH. Share drawings, examples of current defects and target functions to the extent your organization permits.
Frequently asked questions
Should I choose a radial or orbital riveting machine?
Test both against the same joint function, materials, head shape, permitted distortion, cycle requirement and fixture support. Generic advantages do not decide a specific product’s acceptance. For a pivot, evaluate torque and retention together.
Can required forming force be selected from rivet diameter alone?
No. Material, hardness, projection, head shape, peen, friction, support and forming method all affect it. Use supplier sizing charts to shortlist equipment, then set conditions through trials and function tests on candidate production material.
Does saving every force-displacement trace eliminate destructive tests?
No. The trace describes the forming process; its relationship to retention and internal cracking must be established separately. Plan functional or destructive trials at development and change points, then risk-based sampling during operation.
What is the difference between riveting-machine FAT and SAT?
FAT checks configuration and core functions at the supplier. SAT checks actual utilities, part flow, operators, adjacent processes and record destinations in the plant. Manage shared test IDs and identify what can only be tested on site.
What if the RFP cannot yet state a retention force or pivot torque?
Assign product engineering a date to define the test and provisional decision rule. Ask suppliers for a trial proposal against that rule. Do not pretend an undecided value is a final equipment acceptance guarantee.
Which safety standards matter for a plant in Thailand?
Risk-assess the actual machine and tasks, then check customer rules, applicable local law and the relevant scope of ISO 12100, ISO 13849-1, ISO 13850 and IEC 60204-1. Listing a standard does not validate the installed safety function.
Sources
- BalTec, Riveting & Joining: https://baltec.com/sites/default/files/2022-12/202212-BALTEC_en1.pdf
- Orbitform, Process Intelligence: https://www.orbitform.com/processes-services/process-intelligence-control
- Orbitform, Radial Forming: https://www.orbitform.com/processes-services/fastening-forming/radial-forming
- Orbitform, Orbital Forming: https://www.orbitform.com/processes-services/fastening-forming/orbital-forming
- Orbitform, Specialty Press: https://www.orbitform.com/processes-services/fastening-forming/specialty-press
- Orbitform, Aligning Joint Design and Assembly Processes: https://www.orbitform.com/blog/aligning-joint-design-and-assembly-processes-to-achieve-functional-requirements
- Orbitform, Pivot Joint Testing Equipment: https://www.orbitform.com/processes-services/process-intelligence-control/pivot-joint-testing
- Böllhoff, RIVSET self-pierce riveting: https://www.boellhoff.com/us-en/products/additional-product-ranges/rivset-self-pierce-riveting-technology/
- ISO 12100:2010: https://www.iso.org/standard/51528.html
- ISO 13849-1:2023: https://www.iso.org/standard/73481.html
- ISO 13850:2015: https://www.iso.org/standard/59970.html
- IEC 60204-1:2016+AMD1:2021: https://webstore.iec.ch/en/publication/26037
Standards and products change; verify the current edition and scope before purchase. This guide does not guarantee strength, cycle time, yield or legal compliance for a specific product.
