Folding displays are assembled in a vacuum. Once the OLED panel, the folding cover window and the adhesive layers are pulled into a chamber, there is no room for improvisation: every alignment step, every pressure profile and every wait time has already been decided on paper. That paper is the acceptance plan. A vacuum bonding cell that meets its datasheet but fails your production requirements is not a machine problem, it is an acceptance-plan problem, because the plan defined what “meets requirements” means in the first place.
If you are sourcing an OLED folding screen vacuum bonding machine for a new line or a capacity expansion, this is the point in the project where you have the most leverage and the least data. Below is the checklist we work from: what the plan must state, how each item gets demonstrated, and which clauses are routinely left vague enough to cause a dispute later.
Why the acceptance plan matters more for folding than for rigid OLED
A rigid module is bonded flat, and a defect usually stays flat. A folding module is bonded flat and then asked to bend. Any residual stress, trapped gas or bond-line thickness variation introduced during lamination does not simply show up as a visual defect — it shows up later, at the crease, after thousands of fold cycles, in a device that has already shipped.
That changes the character of acceptance. You are not confirming that the machine can produce one good sample. You are confirming that it can reproduce that sample on a shift basis, on your materials, on your cycle time, with defect patterns you can see and measure. The acceptance plan is the document that turns “it works” into “it works the way our contract says it works.”
Before the machine arrives: three items buyers skip
Most acceptance disputes start with a scope gap that existed on the purchase order, not with a machine defect.
- A written substrate and material definition. Panel size and thickness, substrate type, cover window material and coating, adhesive system and its storage and handling conditions. Vacuum bonding is sensitive to surface energy and outgassing. If the material set is not frozen in the plan, the supplier can legitimately claim that the trial condition was never defined.
- Your own incoming inspection standard. Substrates arrive with their own defect population. Define what you will reject at incoming, and record the baseline before bonding trials. Otherwise yield losses get attributed to the wrong station.
- Facility and utility conditions in writing. Chamber limit pressure, ambient class, temperature and humidity band, compressed-air quality, power quality. Vacuum process results drift with the environment, and a machine accepted in a dirty, humid room is an argument you will lose later.
Write these into the plan before shipping. It costs nothing at contract stage and is close to unfixable afterwards.
The static inspection section: what to check in the chamber
This is the standard part of the plan, and it should still be specific.
State the ultimate vacuum the chamber must reach and the leak rate threshold that proves the chamber holds it, together with the pump-down time to reach the working pressure. Record pumping curve data at the accepted condition; a pump-down that gets slower over the campaign tells you more than a single pass mark. Confirm chamber internal dimensions and fixture capacity against your largest intended format, including clearance for the folding fixture geometry, not just the flat panel.
For a semiconductor process equipment cell of this type, also inspect alignment capability as a measurable, not a claim: the alignment accuracy figure, the vision system used to measure it, and the method for verifying it after the machine has run for a while rather than only at commissioning. Specify pressure uniformity across the working area, heating uniformity if the process uses a heated platen, and the temperature ramp and soak profile the machine can actually hold.
Then the things that only matter in a vacuum chamber and are always expensive to fix afterwards: the particle and outgassing behaviour of the fixture materials inside the chamber, the accessibility of the fixture for cleaning without a full teardown, and whether the recipe structure lets you change profile parameters without vendor intervention.
Process capability: run it as a run, not as a demo
This is the heart of the plan and the section most often written as a single line. A bond that looks perfect on one substrate proves almost nothing. Define a trial matrix and hold it.
Run the same product at the stated cycle time across consecutive cycles, from a cold start and from a hot chamber, and record the results of every cycle rather than a summary. Specify the pass criteria in terms you can measure on your own equipment: bubble size and location limits, the maximum acceptable bubble area near the crease and the hinge region, bonding yield per cycle and cumulative, and the visual and functional inspection method for each.
For folding specifically, require the fold test as part of acceptance, not as a later reliability item. Fold the bonded samples to the target radius at the specified cycle count and inspect the crease for delamination, whitening and adhesive flow. A machine that passes flat inspection and fails after folding has not passed. Similarly, run the plan with the actual adhesive lot you will produce with, or document precisely the lot condition used in the trial.
Include one deliberately hostile condition: a deliberately misaligned input and an adhesive-margin edge case. You want to see whether the machine detects and rejects bad input, or whether it quietly produces a defective module that your downstream inspection has to catch. A rejection function that is never tested is a rejection function you cannot rely on.
Integration: the boundaries that get forgotten
A bonding cell does not run alone. Write the interface items into the plan explicitly.
- Load and unload. Whether handling is manual, semi-automatic or in-line, the hand-off dimensions, indexing signals and takt must be defined. Where the cell feeds a downstream high-pressure debubbling machine, the hand-off condition between stations — timing, temperature state, transport orientation — belongs in the plan, because rework between stations silently changes what your bonding yield means.
- Data and traceability. Which process values are logged, at what interval, in what format, and whether a recipe and its result can be tied to an individual unit. If you need per-unit traceability for warranty analysis, ask for it now; retrofitting it into a controller is not a small change.
- Recipe and permission structure. Who can edit a recipe, what is locked, and what change log exists. Production needs to be able to adjust; quality needs to be able to audit. Both requirements belong in the same document.
- Maintenance access and consumables. Chamber seals, fixture elements and heating components have finite life. The plan should state the wear parts, their rated life, and whether replacement requires vendor attendance or your own technician.
The most common integration failure is not a broken interface — it is two teams that each assumed the other side handled it.
Documentation, training and the spare-parts list
Acceptance is also a paperwork event, and the paperwork decides whether the machine is usable six months from now. Require the drawings and wiring diagrams for the delivered configuration rather than the generic catalogue version, the pressure and electrical safety documentation for your market, a calibrated instrument list showing which gauges and sensors were used to sign off the numbers in the plan, and a consumables and wear-parts list with part numbers you can order directly.
Training should be tied to the people who will run the machine and to the failure modes they will meet, not to a general overview session. A short written competency check on recipe handling, alarm response and emergency venting is worth including in the acceptance record. Ask what happens to the recipe set after a controller service, and confirm the backup and restore procedure once, in front of your own engineer.
What should not be left open at the end
Before you sign, check that the plan answers these four questions in its own text:
- Which numbers are pass/fail, and who measured them with what instrument?
- What happens if a criterion is missed — how many correction loops, on what schedule, and at whose cost?
- Which items are explicitly excluded from acceptance and therefore become your risk later?
- What is still outstanding, with a dated owner, and does that outstanding list block the payment milestone or not?
An acceptance plan is not adversarial. It is the shared record that lets both sides say the machine is finished. Vague plans do not make the supplier more flexible; they postpone the argument until the project has no slack left to absorb it.
Where PowerDe fits in this process
Shenzhen PowerDe Automation Precision Equipment has been building vacuum lamination, bonding and debubbling equipment since 2007, and its semiconductor process equipment range covers vacuum bonding for OLED folding-screen production, large-format and curved-surface lamination, and high-pressure debubbling. Because the company builds the chambers, fixtures and vision systems in-house, alignment accuracy, pressure profile and recipe structure are parameters that can be discussed during the trial rather than accepted or refused from a catalogue.
If you are at the specification stage rather than the acceptance stage, that is the better moment to talk. Send us your substrate definition, your target crease defect criteria and your required cycle time, and we will tell you which of them the machine can be signed off against and which need process development before the trial starts.
