A consumer product prototype has to survive a demo, a photo shoot, and a few dozen curious users. A machine that goes onto a production line has to survive a shift, then the next shift, five days a week, in a plant where an hour of downtime costs more than the machine did. That is not a difference of degree. It changes what you build first and what the word prototype even means. Founders carrying a consumer playbook into industrial equipment usually spend the first budget on a handsome enclosure, a slick touchscreen, and a mechanism that has never run eight hours unattended.
Duty cycle is the requirement that reshapes the whole design
Ask an industrial buyer what they need and they answer in throughput: parts per hour, bags per minute, feet per second. Underneath that number sits duty cycle, the fraction of time each component is actually working. A linear actuator rated for 25 percent duty at room temperature will overheat and stall if your cycle keeps it moving 80 percent of the time. A stepper that runs cool on an open bench cooks inside a sealed cabinet sitting at 113 °F (45 °C) ambient next to a drive that is dumping its own heat into the same box.
Run the arithmetic before you buy anything. A machine that indexes every four seconds across two shifts completes roughly 14,000 cycles a day and more than three million a year. At that count, bearing L10 life, belt flex cycles, and cylinder seal life become sizing calculations rather than afterthoughts, and an honest factor of safety on the loaded parts earns its keep: industrial buyers forgive a machine that is heavier than it needed to be and never forgive one that fails in month five.
A short demo therefore proves almost nothing. Budget for an endurance run, and treat everything that loosens, heats, drifts, or wears over an unattended weekend as a design finding.
Safety is a design input, not a sticker you add at the end
Consumer products get certified. Industrial machines get designed around a risk assessment, and the paperwork is the output of that process rather than a substitute for it. The US frame is OSHA 29 CFR 1910 Subpart O for machine guarding plus the General Duty Clause, backed by the consensus standards OSHA and insurers lean on: the ANSI B11 series, ANSI B11.0 for the risk assessment method, ISO 12100, ISO 13849-1 for safety-related control performance levels, and NFPA 79 for the electrical build.
Start with a written risk assessment before geometry is frozen: each hazard, who is exposed, how often, and what the design does about it. The hierarchy matters. Eliminate the hazard, then guard it, then interlock it, and only then warn about it with a label. Guarding ripples straight back into the mechanism, because a fixed guard that must be unbolted for every changeover gets removed by the operator on day three.
Emergency stop behavior is a design choice too, and the vocabulary is worth learning before your first customer conversation.
| Stop category | What happens | Typical use |
|---|---|---|
| Category 0 | Power to the actuators is removed immediately; the machine coasts to an uncontrolled stop | Simple mechanisms and low-inertia motion that must stop instantly |
| Category 1 | Controlled deceleration with power available, then power is removed | High-inertia spindles and conveyors where a free coast is more dangerous than a braked stop |
| Category 2 | Controlled stop with power maintained to the drives | Servo axes that must hold position or a vertical load that would drop |
Even a rough prototype should carry the real safety architecture: a genuine E-stop circuit, interlock switches on the access doors, and a light curtain or two-hand control wherever the production machine will need one. Plywood panels can wait; the circuit topology cannot, because retrofitting a safety relay usually means rewiring the whole cabinet.
Three-phase power, and the bench supply that lies to you
Consumer hardware plugs into a wall adapter. Industrial equipment lands on 208, 240, or 480 volt three-phase service, and that single fact drags in variable frequency drives, motor starters, a control transformer stepping down to 120 VAC and 24 VDC for logic, a lockable disconnect, and a short circuit current rating for the panel. US panel builds are commonly listed to UL 508A, and a plant electrician looks for that label before your machine is allowed to energize.
A prototype wired to single-phase bench power will mislead you about torque at speed, inrush at start, and the electrical noise a VFD sprays into every sensor cable in the cabinet. Build the real control panel early, even if the mechanism around it is still a weldment on casters. The discipline that applies to any product working at line voltage applies here with far more current behind it, which is a large part of why machine development runs on a spec-to-commissioning arc rather than a design-then-launch one.
PLC or embedded control
This decision quietly determines who can service your machine. A PLC from a mainstream vendor gives you ladder logic any maintenance electrician can read, safety relays with published performance levels, and spares available overnight from a distributor. A custom embedded board is cheaper at volume and smaller, but nobody at the customer site can fix it at two in the morning.
The pragmatic answer for most first machines is a hybrid: a PLC handling motion sequencing and the safety chain, with a small computer alongside it for vision, logging, or a richer operator screen. The tradeoffs behind that second box are the usual ones between a microcontroller and an embedded Linux platform. Go fully custom only once the sequence is proven and volume justifies carrying spares yourself.
Serviceability and spare parts are part of the product
Industrial buyers weigh mean time to repair as seriously as mean time between failures. Design so a worn belt, a failed proximity sensor, or a leaking cylinder changes in minutes with common tools, from the front, without pulling the machine out of the line. Prefer catalog components with distributor availability, and publish a wear-item list with manufacturer part numbers as a deliverable.
Acceptance testing happens on the customer's line, not in your shop
Capital equipment is bought against a factory acceptance test at your facility followed by a site acceptance test at the plant. Both run against numeric criteria written into the purchase agreement before you build: throughput at a stated quality level, changeover time, uptime over a defined run, and scrap rate. Language of the operates reliably variety guarantees a dispute.
Expect the site test to be harder. The customer's cardboard comes from a different converter, their product arrives warmer, their compressed air is dirtier, and their operators are faster and less careful than you are. Bring their real material into your shop months earlier, and treat the whole acceptance and handover sequence as a designed phase with its own budget.
Why the right first prototype is ugly
For industrial equipment, the correct first build is a functional one: a welded frame on casters, oversized components, exposed cabling in a wire duct, extra sensors you may later delete, and adjustment slots everywhere so geometry can be tuned without re-machining. It looks like a lab rig because it is one, and the works-like build always precedes the looks-like build here.
Buyers of production machinery judge payback period, uptime, and support. Sheet metal styling comes later, when a proven machine becomes a repeatable product line, and designing equipment stays a different discipline from designing consumer goods even then. Mixing the two priorities is the most common way a first machine ends up beautiful and unsellable.
Projects House develops industrial equipment from spec through functional prototype, safety architecture, control design, and acceptance testing with US clients. If you have a process that needs a machine built around it, describe the throughput, the material, and the plant environment through our contact form and we will tell you what the first prototype should actually be.