Extrusion Profile Tolerance: What Holds It and What Loses It

Where the dimension is actually set

A profile leaves the die at the die dimensions, swollen by die swell and still hot. Everything that happens after that, in the first metre, decides the final dimension: the vacuum pulls it against the calibrator, the water removes the heat, and the haul-off sets the wall by controlling how much the profile is drawn down.

That means tolerance is a downstream property. A die that is perfect will produce an out of tolerance profile if the calibrator is worn, the vacuum is low or the cooling is inadequate, and the operator will adjust the die to compensate, which makes the die wrong for the next run.

The order of investigation therefore runs backwards from the finished part: measure at the haul-off, then at the calibrator exit, then at the die. The point at which the dimension changes is the point that caused it.

Most profile tolerances are quoted as a table in a standard, with the tolerance widening as the dimension increases. A 10 mm dimension may be plus or minus 0.1 mm and a 100 mm dimension plus or minus 0.5, and the table is a statement about what extrusion can do rather than about what the customer wants.

Speed exposes a weak setup

The chart on the left plots dimensional deviation against haul-off speed for three sizing setups. With a well sized calibrator the deviation rises from 0.05 mm at 2 metres per minute to 0.11 at 12. With a weak one it rises from 0.15 to 0.42.

The difference is not the slope, it is where the curve crosses the tolerance line. At 0.15 mm tolerance, the well sized line crosses at about 12 metres per minute and the weak one at 2. A line that is limited by its sizing can only run slowly, and the output penalty is the real cost of a worn calibrator.

The mechanism is heat removal. A calibrator that is undersized, worn or short cannot take the heat out before the profile leaves it, so the profile continues to change shape after it is no longer constrained. The change is small at low speed and large at high speed, because the residual heat is larger.

That is why increasing the vacuum does not always help. Vacuum holds the profile against the calibrator, but if the water flow or the contact length is inadequate, the profile is still hot when it leaves. Vacuum and cooling are two separate requirements and both have to be met.

The five causes

Sizing and vacuum at 28 percent is the largest, and it covers a range of problems: a calibrator that is worn, a vacuum slot that is blocked, a water circuit that is scaled, a tank level that is too low. All of them show up as a profile that moves after the tank.

Haul-off fluctuation at 24 percent is the second. A haul-off that surges changes the draw down and therefore the wall thickness, and the variation is periodic, matching the belt or the gearbox. Measuring the wall against time on a chart recorder shows the period and identifies the source.

Melt pulsation at 18 percent comes from the extruder, not from the line. A screw that is not delivering a steady flow produces a wall that varies along the length, and the variation is at the screw frequency. A melt pump removes it completely, which is why pump equipped lines hold tighter tolerances.

Die wear at 16 percent and uneven cooling at 9 percent complete the list. Die wear is progressive and shows as a slow drift in one dimension; uneven cooling shows as a distortion rather than a dimension, and it is often mistaken for a die problem.

Measuring it so the number means something

Measure at a controlled temperature, after the profile has cooled. A profile measured warm is larger than the same profile measured cold, and the difference on a 100 mm dimension can be 0.3 mm. The standard specifies a conditioning time for that reason.

Measure in the same place on the profile every time. A wall that is measured on a corner and then on a flat is two different dimensions, and comparing them is meaningless. Mark the measurement positions on the drawing and use them.

Measure enough of the length. A profile varies along its length at the screw frequency and across its width at the die, so a single cross section tells you almost nothing. Ten sections over ten metres tells you the distribution.

Table 1 — What to measure, and how much of it

Fixing a tolerance problem

Start with the cooling. Check the water flow and temperature, the tank level, and the condition of the calibrator surfaces. A scaled water circuit is the most common cause of a tolerance that drifts over a shift, because the flow falls as the scale builds.

Then check the vacuum. A blocked slot shows as a flat or a dent at that position, and a low vacuum shows as a profile that is generally oversize. Both are visible on the profile and both are cheap to fix.

Then check the haul-off. Measure the wall against time and look for a period. A period that matches the belt pitch is the belt; one that matches the gearbox is the gearbox; one that matches the screw is the extruder.

Only then adjust the die. Adjusting the die to compensate for a downstream problem makes the die wrong for the next run, and the next operator will adjust it back. That is how a die ends up with no known settings.

Table 2 — Options, in the order they should be tried

Specifying a tolerance that can be met

A profile tolerance should be written against the standard table for the material and the dimension, not against what the customer's assembly would like. If the assembly needs better, the answer is a secondary operation or a different process, not a tighter line on the drawing.

The critical dimensions should be identified separately. A profile has many dimensions and only a few of them matter to the fit; specifying a tight tolerance on all of them multiplies the cost of inspection without improving the product.

And the specification should state the measurement condition, because a tolerance without a temperature and a conditioning time is not a specification. Two suppliers measuring the same profile at different temperatures will disagree by more than the tolerance.