A Robot Pays Back at Two Cavities, and Costs You Money at One

Half the cycle is the oven, and the robot cannot touch it

A typical vacuum forming cycle on a 1.2 mm sheet runs 45 seconds: 22 seconds heating, 3 forming, 9 cooling, 2 demoulding, 5 trimming and 4 packing. The heating stage is 49 percent of the cycle and it is pure physics. The sheet has to absorb enough energy to reach its forming temperature through its thickness, and no amount of automation changes the rate at which heat moves through polypropylene.

That number sets the ceiling on what automation can deliver. A robot that removes every second of handling from a 45 second cycle saves at most 11 seconds, which is 24 percent. A robot that also reduces the cooling time, by taking the part out earlier and cooling it outside the tool, can save more, but that is a process change rather than a handling change.

So the business case cannot be built on cycle time alone. It has to be built on the parts of the cycle where labour actually is: demould, trim and pack, which together are 11 seconds and which are the three stages a person does by hand.

One cavity is worse with a robot

Handling cost per part behaves differently for a person and a machine. A person's handling time per cycle is roughly fixed regardless of how many parts come out of the tool, so the cost per part falls in proportion to the cavity count. A robot's time per cycle is also roughly fixed, but it has a capital cost that has to be recovered, and at low cavity counts that capital is spread over very few parts.

At one cavity the numbers are 9.5 cents a part for manual handling and 11.0 for a robot. At two cavities they are 9.5 and 6.4. At three, 9.5 and 4.6. At six, 9.5 and 2.7. The crossover is at two cavities, and it is not a small effect at the top end: six cavities with a robot is a third of the manual cost.

The shape of that curve is the whole argument. A robot is a fixed cost per cycle, so it wants as many parts per cycle as possible. A single-cavity tool running a long cycle is the worst possible case for automation, and it is also the most common first project, because it is the cheapest tool to buy.

The take-out is a thermal decision, not a labour one

The strongest argument for a robot in thermoforming is not labour at all. It is that a part has to be removed at a repeatable temperature, and a person cannot do that.

A thermoformed part is dimensionally unstable until it has cooled below its glass transition or its crystallisation range. Take it out hot and it distorts on the table. Take it out cold and the cycle is longer than it needs to be. The window between those two is a few degrees wide, and it moves with the ambient temperature of the shop, which is why the same tool produces different parts on a Monday morning and a Friday afternoon.

A robot removes the part at the same temperature every cycle, because the trigger is a timer and an infrared reading rather than a person's judgement. That does not make the part better in itself. It makes the process capable of holding a tolerance, which is what a customer with an automated assembly line is actually buying.

The second thermal benefit is that the part can be taken out hotter and cooled in a fixture. A cooling fixture holds the part flat while it cools, which removes the distortion that free cooling allows, and it shortens the in-tool cooling time because the tool no longer has to do all of the cooling. That is where the cycle time saving actually comes from in a well-integrated cell.

The cell is only as fast as its slowest station

A forming cell is a chain of stations, and the robot is one link in it. Buying a faster robot for a cell whose trim press is the constraint changes nothing except the queue in front of the press. The first job in any automation project is to find the station that sets the rate, and it is usually not the one the project was written about.

Table 2 — Station times in a 45 second cell, and which ones automation can and cannot shorten

The pattern in that table is that the two stages a robot replaces, take-out and pack, are the two shortest. The stages that dominate, the oven and the trim, are the ones automation does not touch. A project sold on labour saving in take-out is therefore selling a small number, and the honest version of the proposal says so.

Where the payback actually comes from

Four things pay for a cell, and only one of them is the one in the proposal.

Table 1 — The four sources of payback, and the assumption each one rests on

The reject rate line is the one most often overstated. A robot removes handling damage, which is a real and measurable category, but it does not remove trimming defects, material defects or forming defects. A shop with a 4 percent reject rate where 1 percent is handling damage should expect to go to 3 percent, not to zero.

The uptime line is the one most often understated, in the wrong direction. A cell that jams is worse than a person who is slow, because a person clears a jam in ten seconds and a cell sits until somebody walks over. The availability of the cell is usually the difference between the projected payback and the actual one.

What to specify before you buy

Most failed integrations fail on interfaces rather than on robots. The specification that prevents that has to cover the hand-offs.

• The take-out temperature, with a tolerance, and how it is measured.
• The tool interface: locating features, vacuum or gripper access, and clearance for the robot envelope.
• The nest or fixture that receives the part, and its repeatability.
• The reject path: how a bad part is detected and where it goes, because a robot that stacks rejects is worse than no robot.
• The recipe structure, so a tool change is a program change and not a re-teach.

Then run the cell at the slowest speed that works, not the fastest. A cell running at 80 percent of its maximum is a cell that absorbs a variation in sheet gauge or a cold morning without stopping. A cell running at 100 percent is a cell that stops every hour and takes the whole line with it.