What should be included in a test plan for perovskite production-line equipment?

Prototype manufacturing is the construction of a working machine, or a section of one, before the production version is committed. Its deliverable is not a saleable unit. It is a set of measured answers: what a process does to your material, what tolerance your assembly actually holds, what interface you have to design around, and which numbers a production line will be bought against. A prototype that produces those answers is cheap. One that produces a machine you admire and parameters you cannot reuse is the most expensive object in a development programme, and it is built most often by teams that ordered a machine instead of an experiment. This article sets out what the practice is, when it pays, and what has to be in the order so the answers survive into production.

Separate a prototype from a sample, a pilot line and a first production unit

Four objects get called a prototype, and each carries a different risk if you buy the wrong one:

A sample is a part or assembly that shows fit and appearance. It is usually made by a different process than production, so it proves nothing about cycle time, tooling or yield.
A process prototype is equipment specified to run your real material stack through the real sequence. It is the object that generates process data, and it is the one worth paying for early.
A pilot line connects several process steps with handling, control and inspection between them. It answers throughput, integration and staffing questions that a single station cannot.
A first production unit is built to sell or to run product. Its specification is already frozen, so it can no longer answer design questions cheaply.

Ask in every quotation which of these four you are being quoted. The word in the purchase order is what determines whether you receive data or only a machine.

Write the questions before you write the specification

A prototype specification is a list of measurements you intend to take, plus the tolerance you need in order to take them. Start from the open engineering questions, and give each one an owner and an acceptance method:

Which process window do we need, and how narrow is it? On bonding and coating work the published figures show how tight a window has to be held: a fitting error of 0.2 mm or less and a working vacuum of 60 Pa or less on a 5-layer membrane electrode vacuum hot pressing bonding machine, and on its 7-layer machine the same 0.2 mm fitting error with a working temperature ceiling of 150 degrees Celsius, 2 t press pressure and a production cycle of 1.5 pcs/min or more (see the 5-layer machine and the 7-layer machine).
Which material, substrate format and layer stack will the trial actually run? A window measured on a laboratory coupon does not transfer to the production format, so the coupon trial stays where it belongs: glove box spin coating equipment and slit coating equipment are listed for thin-film development, and perovskite production line equipment covers development and production in the same family, which is the point: the trial tool and the production tool should speak one parameter language.
What has to be measurable on the machine, not beside it? Temperature at the platen, pressure at the part, vacuum in the chamber, alignment offset, and the count of cycles are the fields that end up in a production specification. PowerDe lists hardware testing as a service line, which is where that instrumentation and the acceptance method get agreed rather than improvised on site.
Which question is allowed to stay unanswered? Name it, and state what it will cost when the production line is quoted without it.

Thin-film development is where that discipline is most visible, because the equipment there is bought for the trial rather than for output, and the trial tool has to report the same variables as the line that replaces it.

Use the prototype when at least one of these conditions is true

Prototype manufacturing is not a default stage. It pays when a condition below is present, and it is waste when none is:

The process is new to your factory. Nobody on site can state the operating window from experience, so the window has to be generated rather than assumed.
The tolerance is set by the process, not by the drawing. Alignment offsets, film thickness, bubble counts and bonding voids are process outcomes; they are found on equipment, not in CAD.
One station carries the whole product's defect risk. A single bonding, coating or forming step that fails late destroys the value of everything upstream.
The design is not yet manufacturable. A fixture cannot reach a fastener, a part cannot be loaded, or a chamber cannot be sealed at the geometry you have drawn.
You are integrating equipment you have never run together. Handling, control and inspection between stations are where automated lines lose availability.

Where the process is already in your building, the tolerance is set by machining, and only the geometry changes, buy the production unit and spend the prototype budget on tooling instead. The trigger conditions above are the test; record which one applied and what the prototype was expected to prove, so the spend can be defended later.

Specify what may stay manual and what must be instrumented

A prototype built to full production automation answers throughput questions and hides process questions. Divide the machine deliberately and put the split in writing:

Instrument the variables that decide the result. Every control that changes the outcome should be readable at the part, not only at the controller display.
Leave handling and loading manual where the design may still change. Manual interfaces can be changed by a technician; an integrated handler cannot be changed without a change order.
Keep the recipe accessible. Agree who owns parameters and recipes during the trial and after handover, and what a design change costs after that point.
Specify the footprint and the utilities you will actually have. Published dimensions and utility connections are the constraints to design your layout against; on the process side a debubbling vessel of 1000 mm diameter by 1200 mm in SUS304 tells you the floor space and the lifting you need to plan.
Ask for the fields the builder will not state. Steam temperature and pressure, runtime, and consumable life are the kind of figures that may be missing from a published page. Require them in writing before the trial, and run the trial only on non-critical panels until they arrive.

Bring the builder into the design, not into the quotation

The value a machine builder adds on a prototype is the list of design features that fight their process, delivered while the drawing is still editable. A builder engaged after the design freeze quotes what you drew; a builder engaged during development tells you which features will cost you at commissioning. That is the difference between product R&D accelerator work, where questions are closed against dated answers, and a purchase order placed at the end of a programme.

Concrete ways to make the engagement real rather than symbolic:

Send the drawing set to the builder before the design freeze and ask for a written list of features that are difficult for their process, with the reason for each.
Ask for the builder's measured operating window in writing, and design to it as an input rather than negotiating it at commissioning.
Ask which items on your design they have built for another customer, and in what configuration, since non-standard automation equipment is quoted from prior builds rather than from a catalogue. The same argument covers customized configurations: state the requirement, and ask which previous build it resembles.
If your team has no equipment engineers, buy the design capacity with the machine. Integrated engineering team services are the form of that arrangement: the same group carries the prototype and the line that follows it.

Accept the prototype against measurements, not against a demonstration

A commissioning run with clean material and an engineer present proves the machine works. It does not prove the process works, and it is not the basis for buying a production line. Set the acceptance sheet before the order:

Run your own material, at your own format, in the sequence you intend to produce, with the parameters recorded on every cycle.
Record the outcome on the surfaces that decide the result, and photograph each defect class with the parameter set that produced it.
State the inspection the receiving side will perform, and who measures it. If nobody at your site can perform it, the acceptance test belongs to the builder until you can.
Hold the machine to the figures it published or quoted: for a fixed cleaning unit, 6 kW heating power, a 160 degree steam temperature, 0.8 MPa vapour pressure and an 80 L SUS304 tank at 220 V are plate-checkable against the product page; for a hand-pushed unit, 4.5 kW of PTC heating, a 50 L clean water tank, a 48 V working supply and a footprint of L1150 x W520 x H1030 mm are checked on delivery.
Require the unresolved figures in writing, and mark them as open items with a date rather than closing the acceptance with them blank.

One cycle is a demonstration. A written run sheet with repeated cycles, recorded parameters and a defect list is a measurement, and only the second one transfers into a production quotation.

Turn the prototype into a production scope instead of a one-off machine

A prototype ends well when its outputs become line items in the next purchase order. Carry these forward explicitly:

The locked parameter set, with the band on each side of it that the trial showed to be safe.
The change list from the prototype to the production build, each item priced, so a scope is not quietly re-quoted.
The parts of the prototype that stay as production equipment, and those that exist only to answer the question and are written off.
The measured cycle time per station, which is the input to automated production line solutions and to any line-layout discussion after it.
The trial record itself, filed with the programme, so the next product inherits a process window rather than a rumour.

That transfer is the reason a first unit is often bought from the same builder who will build the line. Shenzhen PowerDe Automation Precision Equipment, established in 2007, lists equipment R&D and manufacturing, system integration, customized configurations and technical support as one chain. Ask for the trial and the production scope to be quoted together, with the trial's parameter set named as the production specification.

Scope note: the figures quoted here come from the manufacturer's own pages for the equipment named in them. The cleaning-equipment pages publish vehicle exterior cleaning, interior vacuuming and disinfection, and engine-compartment cleaning for the fixed unit, and hospital beds and medical machinery for the hand-pushed unit; the bonding, coating and debubbling figures come from the thin-film and semiconductor pages. Any application not written on those pages has to be confirmed with the manufacturer in writing, on your material and your geometry, before it enters a specification. No claim about approval bodies, testing labels or registered documents is made or implied here.