Cutting it close: improve your profile length tolerances.

Mill lengths don't cut it anymore--precise cut-to-length tubing and profiles are in growing demand. Here's how to achieve the tight tolerances you'll need.

Today's "lean" manufacturing and just-in-time inventory techniques are changing the marketplace for extruded profiles and tubing. In place of the rough-cut mill lengths buyers once favored, OEM manufacturers increasingly expect finished parts and assemblies from their extrusion processors. But first, the processor must develop the ability to cut the extrudate to finished-part tolerances. This sort of precision cutting presents some tough--but not insurmountable--challenges for those who address the trouble spots with appropriate cut-to-length procedures.

One of these challenges is rooted in plastics' relatively large thermal expansion and contraction--especially relative to the metal parts with which plastics must compete. A plastic part cut at room temperature will measure short if shipped in cold weather, and these length changes can be significant. As an example, a 12-ft length of PVC house siding is expected to grow and shrink a total of 3/4 in. from summer to winter. Fortunately, the actual amount of shrinkage can be accurately calculated for any material over a given length. ASTM D696 is the acceptable test method, and resin suppliers can provide a shrinkage coefficient for use in the calculations. With that information in hand, tolerance specifications for finished parts should then be written to include a statement about measuring dimensions at a specific ambient temperature.

TIME ISN'T ON YOUR SIDE

Of all the remaining factors affecting cut length, it is time--or more to the point, the lack of time--that is the single most difficult to overcome. A universal truth relating to cut-length tolerances is that some period of time exists between the moment a sensor measures the part length and the moment of the finished cut. Let's define this critical period as the "Elusive Little Millisecond" or ELM.

All cut-off devices suffer from ELM and it tends to be inconsistent in its duration. In a flyknife cutter, for example, the rise time of the electromechanical clutch/brake will vary with the amount of d-c power and the condition of the friction surfaces. This effect usually lasts more than a millisecond and prevents achieving acceptable tolerances at higher line speeds. Of course, a new flyknife cutter will hold better tolerances than an older unit that shows wear and tear on the friction plates or has not been properly maintained.

In a boosting saw, the ELM is the time required to clamp tight enough to start boosting. Normal air-pressure variations cause that time to vary. Higher pressure means shorter lengths; and conversely, lower pressure means longer ones.

The actual amount of ELM or other time variables does not matter as much as consistency. A stable cutting-cycle time of 100 millisec, for instance, would not produce length variations. Yet, a cycle time ranging between 99 and 100 millisec would yield a variation of |+ or -~1 millisec. That may not be critical at a line speed of 10 ft/min. But at 100 ft/min it equals |+ or -~0.020 in.--too much for many cut-to-length applications.

MORE TIME PROBLEMS

Two other time-related variables can affect tolerances for the worse. One is slippage occurring during the ELM. Whether that slippage takes place between the extrudate and the puller belts or the belts and the driving pulley, it will subtract length by not advancing the extrudate past the blade.

The other variable comes from line-speed variations. Puller speed regulation usually depends on armature feedback capable of holding a |+ or -~2% line speed. This level of variation may produce acceptable mill lengths, but it won't do for precision cutting because any change in line speed during the ELM will add or subtract length. These speed variations often account for tolerances suddenly drifting from spec for no obvious reason. In fact, sudden friction increases from a sizing device can cause as much as a 2 ft/min line-speed change, resulting in an obvious departure from specs.

Isolating the cutting machine from the rest of the line can solve both slippage and line-speed problems. This approach employs a secondary feeder--such as a set of pinch rolls or caterpillar timing belts--to form a controlled loop of extrudate between the primary line puller and the cutter feeder. Most extrudates, even rigid PVC, will flex enough to create such a loop to separate the cutting from any upstream speed variations. Since the weight of the plastic in the loop is only a few pounds, a fractional-hp motor can be used to drive the secondary feeder.

Heavy, multiple-hollow window shapes or any profile too stiff to bend can accomplish the same sort of isolation with dual boosting tables. The first table locks onto the extrudate at line speed. The second table holds the cutting mechanism and is mounted on the first. A servo drive then moves the blade to the correct length. Here, the second table--rather than the extrudate loop--isolates the cutter from the rest of the line.

STOP BEFORE YOU CUT

The same feeding system that isolates the cutter can also be used to stop the downstream advance of the extrudate, eliminating cutting time from the ELM. Once the machine extrudate is stopped at a given length, the ELM is over for that cycle. So, the time required to cut a part to length is no longer important. Consequently, the method works equally well with flyknife cutters, saws, and punch presses. Traditionally, an electromechanical clutch/brake has been used to stop the feeder in these applications. When maintained regularly, or about every million cuts, a clutch/brake can provide tolerances as good as |+ or -~0.032 in.

Servo motors, however, do a better job of reducing the stopping time--and thus ELM. In addition, servo motors have recently become affordable in stop/go feeder applications. A brushless d-c servo motor can come to a full stop from 3000 rpm in 14 millisec. At typical extrusion line speeds, the rpm is under 300, so the stopping time is something less 1.4 millisec. The inherent digital programmability of a servo drive also allows the cutter to ramp down to a constant speed just before stopping. Going to 30 rpm, for instance, would bring the stopping time down to 0.14 millisec, nearly eliminating the ELM factor and giving actual cutting tolerances of better than |+ or -~0.010 in.

LENGTH-DETERMINING SENSORS

Beyond the cutting and feeding equipment, tolerance problems also result from even the most commonly used length-sensing devices. Rotary pulse encoders with predetermined counters may appear to have excellent resolution on paper--a 1200-pulse/revolution encoder with a 12-in. circumference will divide that 12 in. into 0.01 in./pulse. But this fine resolution disappears in the real world because encoder drive wheels slip--and because certain profile shapes won't drive the encoder wheel properly.

What's more, an encoder wheel riding on the puller or feeder belts is subject to a little-known obstacle called "changing radius factor"--or the change in measured length from variations in the radius of the driving pulley. If the thickness of an extrudate increases by 0.010 in., it deforms the driving pulley, changing its radius by 0.005 in. A 3-in.-radius driving pulley will feed 18.8496 in./revolution. A 3.005-in.-radius driving pulley will feed 18.9124 in./revolution for a gain of 0.0628 in./revolution. These variations may seem small, but they will destroy attempts at precision cutting.

Also, the signal from most predetermined counters and photo-electric eyes ends with relay closures having reaction times of 20-40 millisec. This time lag--ELM again--is too long to hold close tolerances. For the fastest reaction times and an accompanying reduction in ELM, high-speed fiber optics are the sensors of choice for length determination. Fiber optics offer reaction times around 100 microsec (0.1 millisec), and their output signal will drive most electronic circuits directly without a relay, saving the 20-40 millisec inherent to the encoders. To be sure, fiber-optic reaction time is still a variable, but it is a more fleeting one with less potential to throw cutting operations out of spec.

After the fiber optics determine where the cut should be made, machine-vision systems can then be used to measure actual finished cut length. Photo-diode arrays, the "eyes" of a machine-vision system, provide an excellent measuring resolution of 0.0002 in. While photo-diode arrays have been used for years to measure cross-sectional dimensions, only recently has this hardware benefited from the software development needed to measure finished part length.

ULTIMATE AUTOMATION

Together, the equipment solutions discussed here take on greatest significance for improvements to cut-length accuracy. But they also allow increased automation, which can clear the way to on-line fabrication. Automation, in turn, enables production of finished parts at faster rates than secondary manual operations. And in keeping with the principles of JIT, the part never has to be inventoried or manually handled.

As an example of this sort of system, a fiber-optic sensor and machine-vision system can be mounted together on a linear slide. A servo-driven ball screw positions the slide in just the right location to correspond to a desired part length--which can be entered at the cutter from a remote computer terminal or from an extrusion-line process controller. Whenever a batch count is reached, the servo motor moves the assembly to a new position to start another batch at a different cut length. By measuring the actual length of each cut part, the machine-vision information forms the basis for scrap decisions. If parts do drift out of tolerance, a diverter gate swings across the take-off conveyor, tossing the bad part into a scrap container. New software-based statistical techniques control all the sensing and scrap determinations--accurately and without operator intervention.