Question 7 of 7: Resolving the Conflict Between Design for Manufacturing and Design for Assembly

Question 7: Resolving the Conflict Between Design for Manufacturing and Design for Assembly (15 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part A — The conflicting requirements

Design for manufacturing seeks to minimise the cost of producing each individual part. Its preferences are simple part geometry, generous draft, uniform and modest wall thickness, no undercuts requiring side actions or collapsing cores, tolerances no tighter than the process naturally holds, short cycle times, high-yield processes and cheap tooling. Design for assembly seeks to minimise the cost of putting the parts together. Its preferences are the fewest possible parts, self-locating and self-fastening features, symmetry or clear asymmetry to prevent misorientation, straight-line vertical insertion, no fasteners, no adjustments and no reorientation of the workpiece.

The conflict is structural rather than accidental, and it has one dominant axis. Design for assembly says integrate: combine six parts into one moulding with snap features. Design for manufacturing objects that the integrated part is exactly the part that is hardest to make — it needs undercuts, side actions, thick sections and tight tolerances across a longer dimension, so its tool costs more, its cycle is longer and its scrap rate is higher. Design for manufacturing says simplify and split, which adds parts, joints and assembly operations. Two further axes matter: tolerance, where design for assembly wants tight tolerances so that parts self-locate and design for manufacturing wants loose ones; and material, where design for assembly wants one material that can carry integral features while design for manufacturing wants each part in the material its own process prefers. Underneath all three lies the real conflict — the two disciplines optimise different terms of the same sum, and neither is authorised to look at the other term.

Part B — An improvement that satisfies manufacturing but not assembly

Take the PC case chassis of Question 1. A single deep-drawn pan carrying an integral motherboard tray requires a redraw operation, an intermediate anneal and a lubricant application, and it forms close to the material's limiting draw ratio, so the process runs with a scrap rate around 6 per cent and is sensitive to coil variation. Splitting it into two shallow-drawn halves removes the redraw, the anneal and the lubricant stage entirely, drops scrap to about 1.5 per cent, and allows both halves to run on smaller, cheaper presses.

That is an unambiguous manufacturing improvement and an assembly regression. The part count rises by one, a joint appears that must be located, clamped and fastened or welded, a tolerance stack is introduced across the new joint where a single formed part had none, and the assembly gains an operation and a fixture. The same pattern recurs whenever draft is added or an undercut is eliminated by splitting a moulding into two straight-pull halves: the tool gets simpler and cheaper, and the bench gets a part and a joint it did not have.

Part C — An improvement that satisfies assembly but not manufacturing

The complementary case is the front bezel of the same case. Integrating the bezel, the light pipe, the front I/O surround and six snap retention features into one moulding is an excellent assembly improvement: four parts and an ultrasonic welding operation collapse into one part that snaps into place in 8.5 seconds and cannot be fitted the wrong way round.

It is a poor manufacturing part. The integrated geometry requires three side actions in the tool, which roughly doubles its cost and adds two more mechanisms that can fail in service. The snap bosses need a 4.4 mm section against a 2.2 mm nominal wall, and since cooling time scales with the square of the wall thickness, that local section takes $(4.4/2.2)^{2}=4$ times as long to cool — driving the cycle from 34 s to 62 s, an 82 per cent increase — while also producing a visible sink mark on a cosmetic surface. Scrap rises from about 2 per cent to about 4.5 per cent. Every one of those is a direct consequence of the integration that the assembly index rewarded.

Part D — A process for resolving the conflict

The resolution is to stop treating the two as competing philosophies and recognise that both are proxies for one quantity: the total delivered cost of the product at the volume the programme will actually build. The process has six steps.

Given. The two bezel architectures of parts B and C. Option A, integrated: tooling CAD 185,000; one moulding of 96 g ABS at CAD 2.85/kg, 62 s cycle in a 4-cavity tool at a machine-and-operator rate of CAD 78/h, 4.5 per cent scrap, 8.5 s assembly at CAD 28/h. Option B, split into three straight-pull mouldings plus a separate light pipe and gasket: tooling CAD 96,000; 88 g ABS and 6 g PMMA, cycles of 34, 26 and 22 s in 4-, 8- and 8-cavity tools plus an 18 s light-pipe cycle in 8 cavities, 2.0 per cent scrap, a CAD 0.135 ultrasonic welding operation, and 22.0 s assembly.

Find. Which architecture is cheaper, and over what range of programme volume the answer holds.

1. Step 1 — build a should-cost model instead of an argument. Total unit cost is the sum of a volume-independent tooling charge and a volume-independent piece-plus-assembly cost, $$c(n)=\frac{T}{n}+u$$ Accumulating the elements above gives, for the integrated option, material CAD 0.2736, moulding $(62/4)\times(78/3600)=\text{CAD }0.3358$, scrap CAD 0.0274 and assembly CAD 0.0661, so $u_A=\text{CAD }0.7030$. For the split option, material CAD 0.2784, moulding CAD 0.3630, scrap CAD 0.0128, welding CAD 0.1350 and assembly CAD 0.1711, so $u_B=\text{CAD }0.9603$. Note what this reveals: the integrated part is dearer to mould and cheaper to own, and neither index alone would have said so.
2. Step 2 — find the break-even volume. Setting $c_A(n)=c_B(n)$, $$n^{*}=\frac{T_A-T_B}{u_B-u_A}=\frac{185{,}000-96{,}000}{0.9603-0.7030} =\boxed{345{,}920\ \text{units}}$$ Below this the split, manufacturing-friendly bezel is cheaper; above it the integrated, assembly-friendly bezel is. The conflict was never a disagreement about principles; it was a disagreement about volume that neither party had stated.
3. Step 3 — test the volume forecast against the break-even. At a one-year commitment of 240,000 units, $c_A=1.4738$ and $c_B=1.3603$, so B wins by CAD 0.1135 per unit, CAD 27,252 in total. At a three-year programme of 720,000 units, $c_A=0.9599$ and $c_B=1.0936$, so A wins by CAD 0.1337 per unit, CAD 96,245. The forecast straddles $n^{*}$, so this is a risk decision, not a cost calculation.
4. Step 4 — make the risk explicit. Choosing A and building only 240,000 units costs CAD 27,252 more than the best alternative; choosing B and building 720,000 costs CAD 96,245 more. On minimax regret the integrated option is chosen, and the reasoning is recorded so that the decision can be revisited if the volume forecast moves rather than re-argued from scratch.

Step 5 — apply the gates that neither index can see. Cost is not the only criterion, and the check is cheap. Does either option fail a regulatory or safety requirement, as the tool-less panel seam did in Question 1? Can the part be serviced, and can the failed light pipe be replaced without scrapping the whole bezel — an argument that favours the split option and does not appear in either cost model? Is the cosmetic sink mark on the integrated part acceptable on a visible surface? Is the supply base capable of a three-side-action tool, and what is the consequence of that tool being down? A cost advantage of CAD 0.13 per unit does not survive a warranty claim or a certification failure.

Step 6 — look for the option that dissolves the conflict rather than trading it. This is the step that separates a good answer from a competent one. The conflict here arises from three side actions and a thick boss, and both are attackable. Replacing the side actions with two internal lifters and one pass-through core removes most of the tooling premium; coring out the snap bosses to a uniform 2.2 mm wall removes the sink mark and most of the cycle-time penalty at the same time. A third option that integrates the parts and still moulds in a near-straight-pull tool would sit below both curves in the figure at every volume, and it exists only because someone asked what was actually driving the conflict instead of scoring the two candidates against each other.