Setup plan: start with the last setup

Where does a setup plan start?

The first question a new drawing gets is how you will hold the part, well before anyone names a price. An experienced manufacturing engineer answers it back to front and pins down the last setup first. That is where the critical features are finished: the bores with a fit, a sealing face with a flatness requirement, the faces that carry a form or position tolerance.

In that last position the part rests on faces that have already been machined, and the workholding touches no functional face. A jaw mark on a sealing face is scrap, not a blemish: the material and the machine time are gone. Once the last setup is on paper, you know exactly which faces the earlier setups have to deliver.

How do you work back to the raw block?

Each earlier setup has one job: to deliver the clamping faces and datums the next one needs. Setup two machines the faces that setup three clamps on, and leaves the reference that setup three's work origin is probed from. The chain runs back like that to the first setup.

That first one clamps on raw stock: the block as it comes off the saw, with a clamping strip of 5 to 10 mm in the vice or a foot for the zero-point clamping system. How much material you leave for it is settled when sizing the blank. This position removes most of the material while keeping the part stiff for as long as it can. Thin a wall too early and the part goes soft and starts to chatter.

What you mill away in setup one, you cannot hold in setup three. So material stays on until the position where it really has to come off.

A tolerance graph shows what stays together

You read it from the tolerances between features, not from where they sit on the model. Draw the features as nodes and the tolerances as edges, with the value and the datum on each. The result is a tolerance graph: a map of what has to stay together. What pulls features apart is the direction the tool reaches them from: on a 3-axis machine a bore in a side face needs a different position from a bore in the top.

Take two bores with a 0.02 mm position tolerance to each other. Split them across two setups and a stack-up comes between them: the repeatability of the workholding plus the error in probing the work origin. Research on fixture design shows that deviations in the locators and in the datum faces carry through into where a drilled hole ends up. That stack-up has to stay below the share of the tolerance your shop's rule set reserves for it.

A split can still work. In our research on why the answer depends on the factory (opens in a new tab) we described two routes that hold the same tolerance: both bores in one setup on a 5-axis machine, or two setups on a 3-axis machine with a jig-bored fixture and CMM discipline. The drawing says what must stay together; the manufacturing route that keeps it together comes from the shop. What a tight tolerance does to the rest of the route is the subject of a separate piece.

A work origin and a check point per setup

Each setup gets a work origin of its own: a face, an axis or a hole you probe, with the orientation of the workpiece coordinate system and the register in the control. It also gets a check point: one dimension you measure after setting up, before the program runs. If the check point is off, look at the repeatability of the workholding first.

How the part sits in that position is already on the drawing. The primary datum face seats on at least three points, the secondary on two and the tertiary on one, which fixes three mutually perpendicular planes. The feature control frame lists the datums in order, primary first. That order says which face meets the workholding first, and so which face the previous setup has to deliver with the most care.

The reference chain between those origins must not loop. The last setup clamps on faces machined in an earlier position that are a datum themselves, or tied directly to one. A face that only appears in the last setup cannot be the reference for an earlier one. That sounds obvious, and it goes wrong the moment someone changes the order halfway through.

Where do the burrs fall?

The order inside a setup follows the same logic. You sequence the cuts so that a burr falls on a side that is still to be machined; the next cut takes it off and nobody reaches for a file. Every burr left standing on a finished face is hand work.

For each edge you record whether deburring happens in the machine, at the bench or outside, and in which setup. That deburring plan belongs with the setup plan, because the right choice for an edge depends on what happens to it afterwards. Edge work that nobody wrote down falls out of the price and comes back later as a question from the customer.

How do you keep the number of setups down?

Choose the smallest number of setups that meets every requirement. The shop's own rule set draws the limits: a maximum ratio of overhang to tool diameter, a minimum clamping strip against mass and cutting force, and the rest material you allow.

The machine drives much of the count. With 5-axis milling you reach five sides of a part in one setup. With 3-axis milling every side asks for a position of its own, and every re-clamp is a chance of error, as the Dutch trade paper MetaalNieuws put it. Every setup also costs setup time, which falls once per batch, not per part. How to count setups and carry them into the cost is covered in counting setups.

An extra setup is the right call when cost, lead time or risk improve with it. A re-clamp instead of a long overhang costs setup time; a rejected part costs the material and the machine time.

A 6082 housing on two machines

Take a housing in 6082 with two bores in a position tolerance to each other and a sealing face with a flatness requirement. The tolerance graph says the two bores and the sealing face stay together, in the last setup.

On a 5-axis machine that makes two setups. The first clamps the raw block, mills the underside and the clamping faces and takes off the bulk. The second stands the part on that machined underside, clamped on a strip with no function, and cuts the sides, the bores and the sealing face in one position.

With 3-axis milling the critical cluster still stays together in the last setup, but every other side needs a position of its own: the underside on raw stock, three side faces with a bore or a pocket, and the top last. That makes five. With an example value of 30 minutes of setting up per setup, the table shows the difference per batch. Over a batch of 10 that is 6 against 15 minutes of setting up per part; a one-off carries the whole amount.

This is where the usual estimate goes wrong. Setup time as a flat surcharge per order, or a setup plan that exists only in the manufacturing engineer's head, cannot see the difference between two setups and five. On this housing that is 90 minutes of setting up per batch: on a one-off the whole amount belongs in the price; leave it out and it comes out of your margin.