A single custom machine almost never arrives as a finished production process. It arrives as a station that has to be fed, unloaded, checked, cleaned and scheduled inside a line that already has its own rhythm. That difference decides how the machine is specified, what has to be built around it, and which decisions belong to the buyer rather than to the builder. Non-standard automation equipment is designed around one operation that catalogue machinery does not cover, while the line around it has to keep running at its own pace. The useful question is therefore not what the machine can do on its own, but what has to be true on every side of it for the line to accept the work.
Where custom automation enters an existing process
Lines usually gain custom equipment at one of three points. The first is an operation nobody has standardised, such as a handling step dictated by the shape of the part, a bonding operation with a tolerance the standard machine cannot hold, or a cleaning pass that has to happen before the next station will accept the piece. The second is an operation that exists but cannot keep up, where the constraint is the rate at which work is presented and completed rather than the skill of the operator. The third is a step that has been running outside the line altogether, on a bench or in a batch, and now has to be pulled into the flow so that semi-finished parts stop waiting between processes.
On the site these two sides are held apart on purpose. The non-standard automation equipment page describes automation designed for a specific duty, while the automated production line solutions page describes the line as a whole. A buyer who reads only the first will under-specify the job, because the machine is only one element of what the process has to absorb.
The process has to be written down before the machine is drawn
A standard machine can be bought against a datasheet. A custom machine has to be bought against a description of the process, and that description is the buyer's own work. Before any drawing starts, the sequence needs to be recorded station by station: what arrives at the station, in what condition, what leaves it, what has been proved about it by the time it leaves, and what happens when the station rejects a piece or stops for a changeover.
That written sequence is also the document the builder works from. Where it is vague, the builder fills the gap with their own assumption and charges the buyer later for changing it. Where it names the reject path, the changeover and the idle state explicitly, the same questions get answered in the design instead of in the shutdown.
Interfaces decide whether the station is accepted
Most integration failures are not machine failures. They are mismatches at the boundary, and the boundaries are few enough to list in a meeting.
The physical boundary is how parts arrive and how they are held. A station that expects parts in one orientation, on one carrier, at one height, will spend its life starving if the upstream station presents them differently. Where the part family is mixed, the question is whether the carrier is dedicated to the part or shared across it, because a dedicated carrier buys speed and a shared one buys flexibility.
The timing boundary is the cycle. The station needs a rate, but it also needs to be known what happens when it is faster or slower than the line around it: which side buffers, how much work in progress that buffer holds, and who decides when the line is slowed to match the station rather than the other way round.
The information boundary is the record. A station inside a line is expected to say what it did to each piece, and to know which piece it is holding. Where that record is skipped, the line still produces output, but traceability stops at the custom station and quality work moves downstream where it costs more.
The utility boundary is the least visible and the most expensive to fix late. Power, compressed air, exhaust, water and any process medium the station needs have to be delivered at the point where the machine will stand, and the site has to be able to say whether it can do that today. A machine that is right for the operation and wrong for the services available at that spot will still be delayed.
Flow, buffer and the constraint you actually own
Adding a station changes where the constraint sits. If the custom machine runs faster than the step feeding it, the machine is idle and the improvement is invisible in output. If it runs slower, it becomes the line's rate, and everything upstream queues in front of it. Both states are acceptable if they were chosen deliberately; both are expensive when they are discovered after installation.
Buffers are the honest answer to that mismatch, and they should be sized as a decision rather than as leftover floor space. A buffer absorbs a short stoppage and protects the neighbours from each other. It also hides a problem, because a line with enough work in progress in front of a slow station will keep delivering while the real rate stays where it was. A buyer who wants the line to get faster has to be able to say what the buffer is for and what it is masking.
Changeover is the other half of flow. A custom station designed for one part will be asked, sooner or later, for a second one. What has to be known before ordering is which features are fixed by the part and which can be reset by a fixture, a recipe or a tooling swap, and how long that reset takes when someone who was not present at the design does it.
Cleaning and process hygiene as part of the integration
Custom stations rarely create clean conditions by themselves. Swarf, mist, residue, released process media and dust are produced at the station and then travel to the neighbours, and the line will only run reliably if their removal is designed in rather than added as a weekly task.
This is where equipment selection becomes a line decision instead of a purchasing decision. Where the site's cleaning duty sits on the cleaning pages, see PowerDe cleaning equipment; those pages describe cleaning machines and the scenarios recorded against them, and they are the place to check what a given machine is positioned for before a cleaning task is assigned to it.
For a station inside a line the practical questions are narrower than they look. Whether cleaning happens in place or with the station taken out of service; what has to be isolated, shielded and drained first; where the run-off goes; whether the neighbouring station can be stopped for the same window; and what proves the station is dry and safe before it is handed back. A custom machine that cannot be cleaned without a long stoppage transfers its cost into lost production, and that cost never appears on the machine quotation.
Control, handover and support once the line owns the machine
A custom station is normally started, stopped, alarmed and interlocked as part of the line rather than as a standalone box. The handover therefore has to cover who can clear a fault, who can change a recipe, what the machine does when the line stops around it, and what has to be true before it will restart. Those behaviours are design decisions, and if they are never written down they become permanent workarounds run by whichever operator has been there longest.
The support model follows the same logic. A one-off machine has no catalogue of spares behind it, so the handover should name the parts that can fail, the parts that were made specially for this station, and who is expected to hold them. Engineering support is a separate service from the machine itself, and it is worth agreeing what it covers after installation, since a custom station is usually still being adjusted for some time after it is accepted.
Commissioning: what the buyer signs for
A single successful run proves the machine can work once. Integration has to be proved against the conditions the line will actually impose, and the acceptance list is short enough to write on one page: the part family presented at the range of variation the line really produces, not just the ideal sample; a rate demonstrated while the neighbours are running, with the buffer behaviour recorded; a reject and a recovery, so the fault path is seen rather than assumed; a changeover performed by site staff without the designer present; a cleaning and restart cycle timed as part of the schedule; and a record that survives all of the above.
Commissioning also does not prove one thing, and it is worth stating plainly: it does not prove that the station will still fit the process in two years, when the part family or the volume has moved. That is a specification question, and it belongs in the design conversation, not in the acceptance test.
A short readiness check before ordering
Before a custom station is committed, the buyer should be able to answer the same questions the builder will ask. Which station is being changed, and what the line does without it today. What arrives at it, what leaves it, and what is proved in between. Which of the four boundaries, physical, timing, information and utility, is hardest to satisfy at the intended spot. Where work waits when the station stops, and what the line does when the station is faster than it needs to be. Who cleans the station, how often, and how long the line is prepared to be down for it. Which staff will run and reset it, and what they will have been trained on. And which parts of the design are fixed for the part and which are expected to change.
A buyer who can answer those questions is buying a station that the process can absorb. A buyer who cannot will still get a machine, but the integration cost will be discovered after installation, where the only way out is a change order.
