A machine can complete one clean cycle in a second and still miss your daily output. That is not a contradiction. The one-second figure may describe only the press motion, while your production result also depends on feeding, material presentation, operator movement, inspection, minor stops, refills, and rejected settings.
You need two measurements, not one. First, verify how long the core machine takes to repeat its own cycle. Then verify how many accepted eyelet settings the complete station or machine cluster delivers during a representative production window. Keep those results separate.
If you merge them into one “speed” figure, you will not know whether a shortfall comes from the machine, the operator method, the feeding system, the material, or the test calculation.
This guide gives you a practical measurement standard you can use for a factory trial, supplier acceptance test, or production audit. It is designed for automatic and semi-automatic eyelet machines, including single-head and synchronized multi-head configurations.
Quick Answer
To verify eyelet machine cycle time, choose one repeatable machine event as the start and the same event on the next cycle as the finish. Time a block of consecutive cycles after warm-up, then divide the elapsed time by the completed cycles.
Machine cycle time = elapsed run time / completed machine cycles
To verify real output, run the actual eyelet, washer, material, tooling, and operator method for a fixed scheduled window. Count accepted settings at the end of the window, without deleting normal jams, refills, small adjustments, or rejects from the clock.
Observed good-output rate = accepted eyelet settings / scheduled test minutes
For a core-machine cluster, run the required machines at the same time and add their accepted settings. Do not multiply a single-machine result by the number of machines unless operator coverage, utilities, material flow, and inspection have also been proven at that cluster size.
Decide What You Are Measuring Before You Start
The phrase “cycle time” is often used for different things. You can prevent most disputes by naming each result clearly.
| Measurement | Start and finish | What it tells you |
| Core-machine cycle time | The same repeatable machine event on two consecutive cycles | Whether the mechanism, controls, and feeder repeat at the expected pace |
| Station cycle time | One completed setting to the next completed setting at the workstation | The pace with normal loading, positioning, activation, and removal |
| Run-time output | Total or good settings divided by actual running minutes | Speed while the station is running, with recorded stop time removed |
| Scheduled-window good output | Accepted settings divided by the full scheduled test window | What the station actually delivered, including normal losses |
| Cluster good output | Accepted settings from all core machines during the same scheduled window | Whether the planned cluster can cover demand under shared conditions |
These figures answer different questions. Core-machine cycle time helps you diagnose the equipment. Scheduled-window good output helps you plan production. A buyer who asks only for “pieces per minute” may receive either one without knowing which it is.
Also define the unit. One machine cycle is not always one finished product. A single-head machine may make one eyelet setting per cycle. A synchronized double-head machine may make two accepted settings per cycle when both heads are active, the spacing matches the product, and both results pass inspection. One curtain panel or banner may require several settings. Write “cycles,” “settings,” and “finished items” as separate fields in your test sheet.
Use This Nine-Step Measurement Standard
1. Freeze the Test Configuration
You cannot compare two runs if the application changes between them. Record the complete test identity before timing anything:
- machine model, serial number, software or control version, and head configuration;
- eyelet and washer supplier, part number, material, dimensions, finish, and lot;
- workpiece material, thickness, layer count, hole condition, and product size;
- die set, punch, feeder track, bowl settings, pressure or force setting, and sensor setup;
- operating mode, safety devices in service, and whether punching and setting occur together;
- operator, material-presentation method, inspection method, and pass/fail standard;
- air pressure at the machine while cycling, electrical supply, and any shared utilities.
This matters because eyelet machines do not process an abstract “piece.” They process a specific combination of hardware, material, tooling, and handling. For example, HANG describes its 101-70 and 101-80 as automatic-infeed machines for paper, cardboard, and plastic, with single and double-eyelet configurations and different limits for self-punching and pre-punched work. That is a useful reminder: you should attach every measured rate to the tested configuration, not to the model name alone.
2. Define One Observable Cycle Boundary
Choose an event that you can see or extract reliably. Suitable boundaries include:
- start signal to the next start signal;
- ram top position to the next ram top position;
- part-present sensor transition to the next identical transition;
- accepted-setting counter pulse to the next counter pulse.
For a foot-pedal semi-automatic machine, “pedal pressed to ram returned and ready” isolates the machine response. “Completed setting to next completed setting” includes operator handling and should be labeled station cycle time.
Do not start at “operator reaches for the material” in one run and at “ram begins to move” in another. A measurement boundary must be observable, repeatable, and written on the test record.
3. Choose a Timing Method That Fits the Cycle
A handheld stopwatch is acceptable for a long block of cycles. It is poor for one fast cycle because the person timing can add more error than the difference you are trying to measure.
Use one of these methods:
- PLC or controller timestamps for the most repeatable event-to-event measurement;
- a machine counter plus an independent clock for a timed block;
- 120 fps or 240 fps video when you need to verify the exact motion boundary;
- a stopwatch for a block of at least 30 cycles when electronic data is unavailable.
Time a block, not one favorite cycle. If 50 cycles take 75 seconds, the measured average is 1.50 seconds per cycle. You avoid most reaction-time error because you start and stop only once.
Keep the counter definition visible. A controller may count activation signals, completed strokes, feeder releases, or accepted settings. Confirm which event increments the number before you use it.
4. Warm Up and Stabilize the Process
Do not use the first few cycles after setup as your speed proof. First, confirm guarding and safe operation. Then run the machine until the feeder, air supply, drive, tooling, and operator method have settled.
As a practical starting point, warm up for 5 to 10 minutes and make enough trial settings to establish a stable feed and acceptable closure. The exact period depends on the machine and application. Restart the warm-up after a major die adjustment, feeder change, or hardware-lot change.
Inspect the first pieces before the timed run. A fast sequence of loose, cracked, off-center, or missing-washer settings is not valid output.
5. Measure Core-Machine Cycle Time in Blocks
Record at least three timed blocks under the same configuration. Use 30 to 100 cycles per block for a fast machine, or a shorter count if the individual cycle is long. For each block, capture:
- start and finish timestamp;
- completed machine cycles;
- total elapsed seconds;
- average seconds per cycle;
- minimum and maximum cycle time if controller data is available;
- jams, alarms, incomplete strokes, or manual interventions.
Calculate each block separately. Then report the median block result as the typical core-machine cycle time and the slowest valid block as a conservative check. If the blocks differ materially, do not average the instability away. Find the cause.
You can also convert the result:
Cycles per minute = 60 / seconds per cycle
A 1.50-second cycle equals 40 cycles per minute. That is a machine result, not yet a promise of 40 accepted settings per minute for the full station.
6. Run a Scheduled-Window Output Test
Now test what production actually receives. Use the real workpiece, normal containers, normal feeder fills, the intended operator method, and the agreed inspection rule. Keep the clock running through normal minor stops, eyelet or washer refills, material repositioning, sensor resets, and rejected settings.
A useful test should be long enough to expose the events that a short demonstration hides. For many applications, 60 to 120 minutes is a reasonable starting window. Extend it when a feeder refill, material-roll change, container exchange, or operator rotation would not otherwise occur.
For final capacity approval, a representative shift or repeated runs on different days give you stronger evidence than one uninterrupted hour.
During the run, record:
- scheduled start and finish time;
- total cycles or total attempted settings;
- accepted settings;
- rejected settings by defect type;
- stop start, stop finish, and reason;
- feeder refills and hardware-lot changes;
- manual interventions that do not create a formal alarm;
- operator changes, material changes, and quality checks.
Do not pause the scheduled clock because the result looks untidy. The purpose is to expose the losses. You can calculate a separate run-time rate afterward.
7. Test the Core-Machine Cluster at the Same Time
If your demand plan requires two, three, or more core machines, test that cluster as a system. A multiplication exercise is not enough.
Run the machines simultaneously with the planned operator coverage. Use the same air network, power supply, material staging, hardware replenishment, inspection route, and downstream handling that production will use. Record each machine separately and the cluster total.
This test can reveal losses that a single-machine trial cannot:
- air pressure falls when several cylinders cycle together;
- one operator cannot present material or clear minor stops fast enough;
- inspection queues delay removal of completed work;
- the same tooling technician or material handler is needed at two machines;
- upstream cutting or marking starves the cluster;
- finished work blocks the station because packing cannot keep up.
Your core-machine cluster is the number of production machines required to meet the order plan. Keep a standby or maintenance spare outside the core result unless you intentionally plan to use it every shift.
8. Count Good Output With a Written Acceptance Rule
Before the run, agree what makes one setting acceptable. Your rule may include:
- correct eyelet and washer present;
- eyelet seated flat without flange damage;
- even flare or roll on the reverse side;
- no cracks, splits, sharp edges, or material tearing;
- position and spacing within tolerance;
- required pull-out, torque, spin, or functional test passed.
Classify rework separately. If a setting needs a second press, manual correction, or replacement, it is not first-pass good output. You may recover the product later, but the original cycle did not deliver an accepted setting.
Where destructive testing is required, use a defined sampling plan and subtract confirmed failures according to your quality procedure. Do not inspect only the easiest pieces at the top of the bin.
9. Calculate the Results Without Double-Counting Losses
Use the full scheduled window for the result you will compare with demand:
Scheduled-window good rate = accepted settings / scheduled minutes
Run time = scheduled minutes – recorded stop minutes
Run-time total rate = total attempted settings / run minutes
First-pass yield = accepted settings / total attempted settings
Stop ratio = recorded stop minutes / scheduled minutes
For a simultaneous cluster test:
Cluster good rate = total accepted settings from all core machines / common scheduled minutes
Projected cluster good output = cluster good rate x planned scheduled minutes
If your observed good rate already includes the normal stops and rejects from the scheduled test, do not multiply it again by availability, performance, or quality factors. That would count the same losses twice. You may apply a separate planning-utilization guardrail if you want deliberate schedule headroom, but label it as a planning decision, not another physical loss.
Worked Example: Two-Machine Core Cluster
Suppose your demand plan calls for 32,400 accepted eyelet settings during 450 scheduled production minutes. You test two automatic machines simultaneously for 90 minutes with the planned operator coverage.
Machine A produces 3,546 accepted settings. Machine B produces 3,371. The cluster produces 6,917 accepted settings in the common 90-minute window.
Cluster good rate = 6,917 / 90 = 76.86 accepted settings per minute
Projected cluster good output = 6,917 / 90 x 450 = 34,585 accepted settings
Unreserved demand margin = (34,585 – 32,400) / 32,400 = 6.7%
The cluster clears the arithmetic, but the margin is narrow. If you require 10% protected time for recovery, training, or schedule variation, test the guarded result explicitly:
Guarded output = 34,585 x 0.90 = 31,127 accepted settings (rounded)
Under that guardrail, the same cluster does not cover the demand. You now have a useful decision: improve the measured losses, change the schedule, reduce the reserve with evidence, or add capacity. You are no longer debating a brochure speed.
Set Pass/Fail Rules Before the Supplier Trial
Write the acceptance criteria before the machine runs. A practical sign-off can include all of the following:
- the median core-machine cycle time does not exceed the agreed value;
- no valid timed block exceeds the agreed slow-cycle limit;
- the scheduled-window good rate meets the demand-based target;
- the full core-machine cluster meets the target while operating simultaneously;
- first-pass yield meets the product quality requirement;
- no critical defect occurs;
- stop time and interventions stay within agreed limits;
- air pressure, tooling, hardware, material, and operator method match the test record;
- results are repeatable across the agreed number of runs.
Avoid a pass rule based only on the average. One exceptional run can hide two weak runs. You can require the lowest of three representative runs to meet the target, or require all three runs to pass. Choose the rule that matches your delivery risk and write it down in advance.
How to Diagnose a Gap Between Machine Speed and Real Output？
When the core-machine cycle passes but scheduled good output fails, the press is not automatically the problem. Use the run log to locate the loss.
If station cycle time is much longer than core-machine cycle time, review material presentation, fixture design, reach distance, pedal position, removal, and inspection. If run-time speed falls as the test continues, check feeder stability, air recovery, heat, sensor contamination, and tooling wear.
If total rate is acceptable but good rate is low, examine the eyelet-washer-die match, alignment, material variation, closure force, and the defect pattern. If each machine passes alone but the cluster fails, investigate shared operators, utilities, material flow, inspection, and packing.
Do not raise pressure or speed simply to recover the number. A change is valid only when the setting remains safe, repeatable, and within the agreed quality standard.
What Your Final Test Report Should Contain？
Your report should let another person reproduce the test without asking what you meant. Include:
- test objective and demand target;
- machine and application configuration;
- cycle boundary and timing method;
- warm-up and stabilization method;
- block-level core-machine cycle results;
- scheduled-window counts, rejects, stops, and interventions;
- per-machine and cluster calculations;
- quality criteria and inspection results;
- deviations from the planned method;
- pass/fail conclusion, names, dates, and signatures.
Save the raw counter export, video, stopwatch sheet, and defect log with the report. A calculated rate without its raw evidence is difficult to audit later.
FAQ
How many cycles should you time?
For a fast eyelet machine, time at least three blocks of 30 to 100 cycles after warm-up. Longer blocks reduce stopwatch error. If controller timestamps are available, collect more cycles and report the distribution, not only the fastest value.
Should you exclude jams and feeder refills?
Exclude them from the isolated core-machine cycle calculation and record them as events. Keep normal jams, refills, and interventions inside the scheduled-window output test. That separation lets you see both machine capability and production reality.
Should you count output in cycles or eyelets?
Record both. Use cycles to verify the machine mechanism. Use accepted eyelet settings to compare with demand. Convert to finished items only after you apply the correct number of eyelets required by each SKU.
Can you verify a double-head machine by doubling a single-head result?
No. Verify both heads operating together on the real spacing and material. Count only accepted settings. If one head is inactive, misfeeds, or creates a reject, the cycle did not produce two good settings.
Is a 10-minute demonstration enough?
It can confirm basic operation, but it rarely proves sustained output. Use a longer window that includes normal replenishment, handling, inspection, and small stops. Repeat the test when production risk is high.
What is the best number to use for capacity planning?
Use accepted settings per scheduled minute from a representative test. For a multi-machine plan, use the simultaneous cluster result. Apply only a clearly labeled planning reserve afterward; do not deduct the same losses twice.
Turn a Speed Claim Into Reproducible Evidence
You do not need another generic comparison between manual, semi-automatic, and automatic machines. You need a test that separates core-machine motion from station losses, counts accepted settings, and proves the required machine cluster under shared production conditions.
Send QC Machinery your eyelets, washers, representative materials, product drawings, quality limits, planned operator coverage, and demand target. You can then agree the cycle boundary, scheduled test window, cluster size, and pass/fail rules before the trial. The result will be a rate you can audit and use in your capacity plan, not a number that works only during a short demonstration.