The fork in the road

Four project types, four different things to optimise

The question “which profile should I choose” has no answer until you know how many you are making and whether the design is still allowed to change. A frame that will be built once can tolerate a section that is 20 per cent heavier than necessary. A frame that will be built two hundred times cannot, because that 20 per cent is multiplied by the order quantity every month for as long as the product lives.

Read the third column carefully. In three of the four rows, the thing that a general selection guide tells you to focus on is the thing you should ignore. That is the whole point of this article: the profile-selection question is a different question at each scale.

One or two units

If you are building one, buy the standard die and stop

For a single frame, a custom extrusion is almost always the wrong answer, and the reason is not the die cost. It is the sequence. A new die has to be cut, corrected and sampled before it produces a saleable length. The industry typically quotes two to four weeks for die manufacture and first sample, and that is before you have measured the sample, approved it and released the production run. A frame that could have been built from stock next week is now a six-week project, and it will still need a second iteration because nobody gets a slot geometry right the first time on a drawing alone. Our own section range covers the sizes most one-off frames need, cut to order from 6 m stock.

What you should do instead, for a one-off:

• Choose the section from what is on the shelf, then solve the structural problem with span and bracing rather than with geometry. Adding one upright is a same-day change. A new die is not.
• Spend the optimisation budget on the machining, not the extrusion. Accurate cut lengths, tapped ends, a clean hole pattern and deburred edges are what make a one-off frame go together in an afternoon instead of a week.
• Stay inside one accessory system. One nut part number, one bracket part number, one torque figure. On a one-off build the accessories are more than half the assembly time, and mixing systems means you discover the incompatibility with the frame half built.
• Accept the weight. A section that is 20 per cent heavier than the theoretical optimum costs a few dollars on a single frame. Chasing that 20 per cent costs days.

Five to fifty units

If you are building twenty, the cutting list is the project

Once the section is fixed, the dominant cost in a small batch is not aluminium. It is the time spent turning six-metre bars into the right set of parts, and then finding out that two of them are the wrong length. A 1,200 × 800 × 900 mm workbench contains twelve cut lengths. Twenty of them is two hundred and forty cuts, and at roughly a minute and a half each including measuring, deburring and handling, that is six hours of saw time before anything has been assembled. Measured in loaded labour that is a full day, and it produces a variable amount of scrap.

The alternative is to buy the parts cut, tapped and deburred to a list. It costs a small amount per cut and it removes the measuring, the scrap and the skill requirement from your own floor. The comparison is not “cutting charge against free in-house cutting” — it is “cutting charge against six hours of your own labour plus the cost of the two members you cut 5 mm short”.

What a cutting list has to contain

A list that says “40 × 40, twelve pieces, 1,200 mm” is not a cutting list, it is a wish. A list that can be manufactured without a phone call has one row per distinct part and these columns:

Two of those columns are consistently left off and consistently cause problems. Face features without a datum produce a batch of parts that are correct but mirrored, which is worse than being wrong because they will assemble and then not fit the panel. And no tolerance on the length means the supplier will use whatever is economical, which may be plus or minus 2 mm, and your frame will not be square.

Kit by assembly, not by part number

If you are building twenty identical frames, ask for the parts to be bagged by assembly rather than by part type. Receiving twelve bags that each contain one bench is a completely different operation from receiving twelve piles of one hundred members that have to be sorted on the floor. The supplier cost is nearly identical; the difference on your side is several hours per batch and a large reduction in mis-assembly.

Hundreds of units

If you are building production, the die is the cheap part

A custom extrusion die for a modest hollow section is not the expensive item in a production programme. It is a one-time cost, and it is small compared with the recurring cost of machining a feature onto every length for the life of the product. The die pays back when it removes work, not when it changes appearance.

The arithmetic, before you commit

Custom extrusion minimum order quantities are normally quoted in kilograms, commonly in the range of a few hundred kilograms per section and finish combination, because the press has to be set up and run a minimum billet sequence to be economic. That means the tooling decision and the order quantity decision are the same decision, and it should be made with one line of arithmetic rather than with a feeling.

Run that calculation honestly and two things usually happen. Either the break-even quantity comes out comfortably below your annual volume, in which case the decision is easy, or it comes out above it, in which case you have just avoided a tooling investment that would have been written off.

The sample gate nobody should skip

Before releasing a production run, get a sample and measure it. Not look at it — measure it. A first-article sample should arrive with a dimensional report covering the envelope, the slot mouth, the groove cavity, the wall thicknesses and the straightness, and you should check the report against your own measurements of the sample rather than against the drawing alone. Two things are worth checking that are usually omitted: the anodised sample, because coating builds on the slot lips and changes the nut fit, and the end cut, because a section that measures correctly along its length can still be twisted.

Existing machines

If you are extending something already built, measure it first

Extending an existing machine is the one case where no catalogue can help you, because the frame in front of you is the specification. Two things make this harder than it looks. The first is that a series name does not define a slot, so a “40 series” profile from one source may have a different mouth, cavity and pitch from a “40 series” profile from another. The second is that a profile that has been in service for years is not the profile that was delivered: anodising fades and changes colour between production lots, and a frame that has been in a coolant mist environment has a surface that no new extrusion will match exactly.

Record these eight measurements from the frame you are matching, and photograph the cut end of a member next to a ruler. That photograph is worth more than any drawing you will find.

Then decide which of the three mismatches you can live with. A slot mismatch is not negotiable — if the nuts will not enter, the profile is useless to you. A finish mismatch is usually tolerable on a machine frame and usually not tolerable on a visible enclosure. An alloy or temper mismatch is tolerable structurally in most cases, but it matters if the new members will be welded or hard anodised, because those processes behave differently across alloys. For joining two members end to end on the same line, a flat joining plate is usually the part you are looking for.

Documents

What each stage of a project should produce

Projects that run late usually did not fail at the design stage. They failed because a decision was left implicit and was therefore made twice — once by the designer and once by the supplier. These are the six gates, and the one document that closes each of them.

Questions

Frequently asked questions