A project team buying flexible-display bonding equipment usually holds three meetings before it agrees what it is testing: one about the parts, one about the schedule, and one about the money. Testing and validation is where those three views have to agree on the same evidence, or the purchase decision rests on a supplier demonstration that answers nobody's question. The plan below is written in the order a team can actually use it: define the object, agree the pass criteria, run the checks in sequence, record what happened, and say what each result changes.
A word on scope first. The bonding machines in the semiconductor process equipment category are described by what they work on and how they hold and bond the film, not by a universal performance table. The category cards describe alignment, roller bonding, vacuum bonding and a dedicated mechanism for curved surfaces; those are the elements worth verifying, and the numbers attached to them belong to particular machines and particular test pieces. Treat published figures as a reference configuration and ask for them in writing against your own part.
First agree what the validation is for
Two different questions get confused in one test plan. The first is machine validation: does this equipment do what its own description says, repeatably, with the alignments holding across a run. The second is process validation: does this equipment, on your stack, produce parts your inspectors pass. A machine can pass the first and fail the second because your film, your substrate and your tolerance are different from the ones used to describe it. Write both lines in the plan, each with its own pass criterion, and never let a witnessed machine run be recorded as proof about your parts.
Second, write down the part before the protocol
The tests follow from the part. Before any protocol is drafted the team needs a single agreed sheet: substrate and thickness; whether the surface is flat, curved or three-dimensional; the film stack in the order it is laid, including hard coat, anti-glare, anti-shatter, polariser, adhesive and release liner; the bond area and the edge condition; the position tolerance you will accept; and whether the line runs one product all week or changes over several times a day. Flexible material behaves differently from rigid material at every stage from cutting to bonding, so a plan written for a rigid test piece says nothing about a flexible one, and geometry needs its own samples rather than being folded into the main run.
Agree the running conditions before the first trial
Before any block starts, the team has to agree the conditions the machine will be asked to run under, because a validation that ignores the site cannot be closed at the site. Write down where the machine sits, what supplies reach it, how it exchanges signals with the equipment upstream and downstream, and who may change a recipe once the machine is running. If a supply or an interface is missing, that is work the project team owns, not something the supplier will solve during installation, and it belongs in the schedule alongside the trial dates.
Set the pass criteria before any trial starts
A criterion an inspector cannot apply is not a criterion. For each characteristic decide three things in advance: what counts as a defect, the size or position at which it becomes a reject, and who has the final say when two people disagree. The characteristics worth fixing in writing are registration against the reference marks, bubbles and inclusions, dust and particles caught in the bond, edge lift or a bond that does not run to the line, and any change in appearance after the bond rather than at it. Write the wording the inspector will read, and keep a physical boundary sample of a borderline part; photographs alone drift between shifts.
Then state what you are measuring against. If the requirement is a position tolerance, say how it is measured, with which instrument, on which datum, and on how many consecutive parts. A single good part proves the machine can do it once.
Sequence the stages so each one answers a separate question
Run the validation in five blocks and keep the records separate.
The first block is a documentation review. Drawings for the delivered configuration, panel layout, terminal schedule, a parts list with manufacturer codes for the consumables and wear items, the input and output list where the machine talks to other equipment, and the manuals for the program actually installed. Missing documents are a finding, not an afterthought; they decide whether you can maintain the machine later.
The second block is a machine check at the supplier. What you witness there is the machine's own behaviour: cycle with its contents, alarms and the way stops are reported, alignment behaviour on consecutive parts, drift across a continuous run, and the changeover from one stored recipe to another. This block tells you how the equipment handles work. It does not tell you whether your product passes.
The third block is a process trial on your own materials: your stock, your substrate, your stack, your inspector, and the pass criteria agreed above. Trial only on the material you intend to buy, because film grade changes how a bond looks and how an edge behaves.
The fourth block is a site check after installation: power at the position, compressed air, floor space including access for maintenance, and the environment the machine will actually sit in. A result achieved in a supplier's conditioned area is not a site result.
The fifth block is repeatability and handover. Repeat the trial on a second batch after installation, then train the named operators and supervisors and keep the signed record. Include who clears a jam, who performs the first-off check after a changeover, and who is allowed to change a recipe.
The records the team needs to sign
Five documents close this out: the agreed part and stack sheet; the pass-criteria sheet with the boundary samples; the witnessed machine run, timed and noted by whoever attended; the process trial report with measured values, reject counts and the reject parts themselves kept; and the site plus handover record covering utilities, training and the responsibility split for maintenance and support. If a change was made to the delivered machine, record it against a drawing. The point of the set is that the person who signs the purchase order and the person who will run the machine in two years read the same evidence.
Split the work by role, not by meeting
A plan fails when one engineer owns all of it. Assign each block to the role whose judgement it tests: process engineering owns the part sheet and the process trial; quality owns the criteria, the boundary samples and the reject wording; production planning owns cycle time, changeover frequency and how much stoppage is tolerable; maintenance and facilities own utilities, access, spares and who services the machine; procurement owns the enquiry, the identical question list to every bidder, warranty, and the consumable price after the warranty period. One person then assembles the five records and calls the decision. Put that split in writing at the start; it is the cheapest part of the plan and the first thing to be dropped under schedule pressure.
Use third-party testing and custom-build routes on purpose
Not every team has the people to run a validation on its own. PowerDe publishes testing and enquiry routes on its engineering pages, although they are written for cleaning equipment rather than display equipment: steam cleaning equipment testing, and a customized steam cleaning equipment enquiry used to scope a machine against a stated process instead of a catalogue entry. Borrow the discipline from them, not the product claim: agree the trial method, the material and the pass line before any trial run, and have the trial recorded on your own part rather than on a sample chosen to show the machine at its best. If your stack is unusual, one published card describes a machine built for a single customer's flexible-display process, so a process-specific build is a normal answer: send the drawing and the stack sheet instead of asking which category name fits.
Decide in advance what each result changes
Write the consequence next to every block. A failed machine check is a question to the supplier and a re-run, not a reason to cancel the trial. A failed process trial splits into two very different actions: an alignment or bond defect points at machine adjustment and recipe, a material defect points at the film stack and the supplier of the film. Fix that branch before the trial so the team does not spend two weeks arguing about cause afterwards. And set one stop condition: if the pass line cannot be agreed in writing, the trial does not start, because a run without an agreed line produces a report nobody is obliged to accept.
Send the same part sheet, criteria sheet and question list to every bidder, insist on a trial on your own part and film before the order, and keep the five records together. That is what turns testing and validation from a schedule line into a decision the whole project team stands behind.
