Ask five automation sales engineers which concept to use and you will get five answers, each one attached to a product they happen to sell. Ask a machine designer who has installed all five and you get a much shorter list of questions. I have built rotary indexers, walking beams, gantries, pick-and-place robots, and one memorable machine that used all of them at once, and I stopped being impressed by the hardware years ago. What separates a good automation concept from a bad one is not the badge on the actuator. It is whether the concept matches three things: the cycle time budget, the part that has to cross the machine, and the flexibility your customer genuinely pays for.
Start with the cycle budget, not the layout
I always begin with a number: the target cycle time and today’s required volume, with a crude growth factor. From the cycle time, everything else follows. If the customer wants 60 parts a minute, your handling concept is decided before you open a CAD file, because few mechanisms will load, index, and unload that fast with real-world settle times. If they want 6 parts a minute, a whole universe opens up and most of it is cheaper than you think.
The budget math is a fraction, and I keep it deliberately crude in concept stage. Take the cycle time, subtract the process time, and what remains is the handling budget. In a 10-second cycle where the process itself takes 7 seconds, you have 3 seconds to get the part in and out, position it, and settle. That is comfortable for a hand-loaded or simple axis machine, and tight for a robot that needs a full choreography of motions and a dwell for its own settle. The concept that cannot protect its share of the budget gets eliminated, usually before any fancy kinematics gets discussed.
Let me make the numbers specific because vague conversations are where automation budgets die. Consider a washing operation taking 8 seconds inside a 12-second cycle; your handling window is 4 seconds. A rotary index table with the process performed in the fixture at a station gives you the whole 4 seconds, because loading happens at one station while the process runs at another. A robot that loads the fixture, steps away, waits 8 seconds, then returns has to cover the process time with its own hours — your robot is effectively idling at 70% of its life. The table looks less sexy and wins the budget. This is why I tell junior engineers: cycle time is the parent of the concept, and everything else is negotiation.
The part is the second boss
The cycle budget says how much time you have; the part says how you may spend it. I look at four part characteristics before any layout work. Mass and stiffness tell you whether the part can be flung, clamped, or must be cradled. Geometry and orientation tell you whether it lands in the fixture naturally or needs a pose change. Sensitivity — optical, coated, soft — decides whether you can touch it hard at all. And the tolerances of the process tell you how much locating error the mechanism is allowed to add, which controls how stiff your positioning has to be.
An example I use constantly: a stamped bracket that weighs 40 grams, has one clean datum face, and absolutely must survive contact without marks. That wants a light gripper and a scoop-and-place, never a magnetic pick tool, and the orientation arriving from the upstream bowl feeder decides everything. Meanwhile a casting at 4 kg with a drilled datum hole wants rigidity and repeatability, and a gripper with a compliant jaw is out of the question. The concept, in both cases, follows the part like a shadow.
The five concepts I actually choose between
When the sales brochures are stripped away, the machine designer’s toolbox is small, and that is a comfort, not a limit. I choose between five architectures, and each has a personality I can describe from experience.
| Concept | Best when | Watch out for | Typical sweet spot |
|---|---|---|---|
| Rotary index table | Process at stations, parallel operation, moderate parts | Index accuracy at high speed, in-process time fixed | 4–15 s cycles |
| Linear transfer / walking beam | High volume, one motion axis, heavy parts | Multiple stations lock step, less flexible | 2–8 s cycles |
| Cartesian gantry | Large work area, high payload, straight paths | Inertia, frame stiffness, cost of the bridge | 10–60 s cycles |
| Pick-and-place robot | Flexibility, changing parts, complex paths | Cycle reports lie; settle eats the budget | 12–60 s cycles |
| Hand-loaded with positioner | Low volume, high flexibility, easy changeover | Operator rhythm, safety guarding cost | 30–180 s cycles |
My heuristics are unglamorous. Under an 8-second handle-with-care cycle, I refuse a robot until proven otherwise, and I let the frame take the vibration rather than the arm. For cycles under 6 seconds with a clean part, a cam-driven indexer or a walking beam will out-suffer any servo system for a decade — the parts count is embarrassing but the uptime is biblical. Above 30 seconds, unless there are dozens of variants, I hold a quiet meeting with the customer about whether the “automation” is actually a clamp-and-rotate positioner with an operator, because the honest cost of full robots on slow lines is guarding, programming, and regret.
The flexibility conversation everyone avoids
The third boss is the hardest to interview, because it is a promise to the future. The flexibility question is not “can this machine handle another part?”, it is “which future changes will you actually pay to handle?”. Tool-less changeover for three variants, in a day, is a real requirement some customers have. A machine that can turn into a different machine if the part changes shape every year is a fantasy everyone budgets for and nobody uses at full cost.
I put the options in front of people in plain money. A dedicated rotary line for a single part might cost 60% of a flexible robot cell and run faster, but the robot cell survives a part revision with a software tweak while the dedicated line needs new tooling and worry. When I put those two price tags side by side and ask the customer how likely the part is to change, the decision usually makes itself. What I refuse to do is quietly build the expensive flexible thing “to be safe” — “to be safe” is how budgets get blown on features that a signed volume forecast never promised.
Flexibility is not a feature; it is an option on the future. Price it honestly, or it will price you unexpectedly.
Can we simulate our way out of the choice?
A recurring question from younger engineers is whether simulation can decide the concept for them. My honest answer: simulation settles the winners of the finalists, but it does not generate the finalists. You still have to propose two or three credible concepts from the heuristics above before any software earns its fee. What I do run at concept stage is a rigid-body cycle simulation of the two finalists, with the actual motor curves and the actual payload, and I watch the settle tails. The settle time is where concepts lie to you: a servo axis quotes a move time of 0.4 seconds and forgets that the load is still ringing for another 0.3 while the position control argues with the beam it is bolted to.
I have a memory of exactly this trap. We were comparing a gantry and a robot for a 20 kg tray transfer in a 15-second cycle. The datasheet settle times suggested the robot squeaked by. The rigid-body sim with the tray mass and the wrist inertia showed the robot’s final axis still settling at 0.6 seconds past the budget, because the tray carried the wrist off its balance point. Within the same simulation, a gantry with a heavier but stiffer frame and a tuned servo settled in 0.2. The robot would have made the machine late on every single cycle, forever. Nobody had meant to mislead; the datasheet simply did not include our tray. That is the entire argument for simulating finalists, and the entire argument for never letting a simulation replace the heuristics that generate the candidates.
A case where the ugly concept won
I want to close with a machine that technical people find embarrassing and customers love. The part was a large lawnmower deck, 12 kg, awkward geometry, one handle-off-the-shelf orientation, and a required cycle of 45 seconds. By my rules, this screamed hand-loaded with a rotating fixture: an operator places the deck, a servo positioner rotates it forty-five degrees through two stations, and the operator repeats. The quoted robot cell was elegant and horrified my project manager; the fixture-and-positioner machine was ugly and cost a third as much. It has run for six years, changed over in minutes to a second deck model, and the operator rhythm is so gentle the cycle is exceeded only when the human stops for coffee.
The lesson I repeat at every design review: elegance is for the brochure, not the budget. The concept you choose should survive the three bosses — the cycle budget, the part, and the flexibility promise — and your ego appears nowhere in that list. When a boring fixture beats a beautiful robot at all three questions, build the boring fixture and sleep well.
A checklist I run before any concept leaves the room
The final gate before a concept is approved is a short list of brutal questions, and I run it out loud so everyone hears the answers. Does the concept protect its cycle-time share, including settle and dwell, on the worst day and not the best one? Can the part enter, locate, and leave without a fight, in every orientation the feeder may deliver? Can maintenance reach every actuator they will have to replace, with the guard door open and the mind already tired at 2 a.m.? Does the safety architecture — guards, light curtains, reset routines — add time to the cycle the budget did not see? And if the part changes six months in, is the pain a tooling change or a machine project? I have never had a concept that answered all five perfectly. I have frequently had to reject a concept that failed one of them loudly and that marketing had already photographed.
Mistakes that keep paying me back
| Mistake | What it costs | The habit I now use |
|---|---|---|
| Choosing the robot for the brochure | Guarding, programming and settle regrets | Cycle budget is the parent of the concept |
| Trusting datasheet settle times | Every cycle late, forever | Rigid-body sim of the two finalists |
| Buying flexibility to be safe | Blown budget, unused features | Price the flexible option next to the part-change probability |
| Ignoring the settle tail on fast axes | Missed cycle, mysterious downtime | Watch the tail in simulation, not the move time |
| Designing for the pretty layout | Operator reach and maintenance misery | Run the 2 a.m. maintenance question out loud |
If this article has a single thesis, it is that concept selection is a filtering job, not a creativity contest. You filter by cycle time so the time-killers leave early. You filter by the part so the mechanisms cannot betray you. You filter by the flexibility promise so the budget survives contact with the customer’s future. What remains is usually two finalists, and then and only then do simulation, cost, and a glance at the maintenance crew’s expression make the final call.
I will add one closing confession, because it is the truest thing I know about this craft. The best automation decisions I have made were the ones where the customer said “but that is so simple”, and I had to resist the urge to apologize for it. A machine that does exactly the boring job, at the promised cycle, with reachable parts and an honest price, is the rarest animal in our trade. Choose for that animal. The sales brochure will forgive you.
