Buying a machine that is bolted together from catalogue parts is one decision. Commissioning a cell that nobody has built before is another. In the second case the drawing is not finished when you sign, and the questions you ask before the first meeting decide whether the project ships or stalls. The checks below are the ones that separate a workable non-standard automation project from an expensive prototype.
Start with the parts you make, not the machine you want
A machine builder cannot design a fixture until they know what has to be held. Send three drawings with the critical dimensions and tolerances marked, plus a photo of the worst variant that comes off your press or mould. State the material, the surface coating, and whether the part is flexible, sticky, or oily when it reaches the cell. Those details change the gripper, the vision approach, and sometimes the whole layout.
Ask the same question in reverse: what happens when the part varies. Incoming dimension spread, burrs, flashes, and a coating that changes with the supplier are all normal on a real production floor. Describe them honestly. A cell specified against a perfect sample part will reject good product, and re-tooling a gripper after acceptance costs more than the conversation that would have found it.
Quantify takt time and separate it from changeover
"We need it fast" is not a specification. Give the cycle you must hold, the units per shift, and the number of shifts per week. Then state how often the line changes products and how long each changeover may take. A builder sizing a non-standard automation equipment project will reach a different design for 900 parts a shift run on one SKU than for 300 parts a shift across twelve SKUs, because the second design spends most of its engineering effort on quick changeover and recipe handling.
Write down what happens during a stoppage as well. Which station may hold work in place, which must be cleared, and how long the line may be down before the stop becomes a quality problem. This determines whether the cell needs an intermediate buffer, a pass-through position, or a purge routine in the control software.
Describe the room the equipment will live in
Floor space, door width, ceiling height, floor loading, and the position of existing utilities end the argument that the machine should simply be made bigger. Measure the footprint you can give up, including the swing of doors and the space an operator needs to load and maintain. Note the power supply available at that spot, compressed air pressure and flow, and whether water or extraction is within reach.
The environment changes the bill of materials more than buyers expect. Coolant mist, metal dust, food residue, or a washdown area force different motors, enclosures, and frame components. A clean assembly area and a wet, oily maintenance zone need different protection on the same functional design, so raise this before the quotation rather than during it.
Define what "working" means before the contract
Custom machinery is accepted on terms agreed early or not at all. Agree a run test in writing: which parts, which batch, how long the cell runs, and what rate and quality result releases the final payment. Decide now who provides the sample material for that test, since a trial with your own parts is the only way to see how the machine behaves on your tolerances.
- One person or two, and what they do between cycles
- Which languages the display must use, and who is allowed to change recipes
- Whether your existing line controls must see a signal from the new cell, and over which interface
- Which records you have to keep, and whether the machine has to produce them
Operator skill level belongs in the same list. A cell designed for a technician who can recalibrate a vision system behaves badly in a plant where the shift leader has ten minutes and a manual. Ask the builder to demonstrate the fault messages and the recovery steps you would actually use at two in the morning.
Ask who will still be there in year three
A one-off machine is only useful while spare parts and drawings exist. Ask which components are standard catalogue items and which are proprietary, and get a price and lead time for the wear parts before you order. Request the electrical drawings, the program backup, and the parts list in a form your own maintenance team can read.
The engineering work behind such a cell usually starts long before fabrication. A supplier offering product R&D accelerator support can build and test a proof of the difficult station, such as a feed or a bond, before the full machine is designed, which moves the risky part of the project into a small, cheap experiment. Where the machine itself is a product you intend to sell, the same team can cover development and hardware testing rather than leaving verification to the first production unit.
Finally, compare quotations on scope, not on price. Two builders may quote the same cycle with different assumptions about guarding, loading, and acceptance. Put both proposals against your takt time, your floor space, and your run test, and the cheaper option often reveals what it left out.
