Question 2 of 7: Design for Manufacture and Assembly

Question 2: Design for Manufacture and 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 main goal of DFMA (3 marks)

The main goal of Design for Manufacture and Assembly is to minimise the total cost of the product by simplifying it, while holding or improving its quality and its functional performance. The two halves of the acronym attack that goal from different directions. Design for Manufacture asks whether each part can be made more cheaply — the right process for the volume, geometry that respects that process’s rules (draft, uniform wall, generous radii, avoidable undercuts), tolerances no tighter than the function requires, and materials the process can actually work. Design for Assembly asks the more radical question of whether the part needs to exist at all, and Boothroyd, Dewhurst and Knight give it a formal test: a part is a theoretical necessity only if, relative to the parts already assembled, it must move, it must be of a different material, or it must be separable for assembly or service. Every part failing that test is a candidate for elimination or combination.

The reason the goal is stated as total cost is that the two objectives conflict, and DFMA is fundamentally the discipline of resolving that conflict at the product level. Consolidating six sheet-metal parts and their fasteners into one moulding raises the piece cost of that moulding and its tooling cost, but removes five handling and insertion operations, five parts to purchase and store, five entries in the bill of materials, and the quality risk of five interfaces. Optimising the parts one at a time reliably produces a product that is more expensive than the one that questioned the part count first — which is why DFA is normally applied before DFM. The secondary benefits follow from the same simplification: fewer parts means fewer things to go wrong, so reliability improves, assembly time and therefore lead time fall, and the inventory and supplier-management burden shrinks.

Part B — A joining strategy consistent with DFMA: the integral cantilever snap-fit (6 marks)

The clearest example is the integral cantilever snap-fit, in which a moulded hook on one part deflects over a mating catch on the other and springs back to lock it. Compare it with the four screws it typically replaces.

Features that enhance manufacturability. The snap is formed in the same injection-moulding shot as the part it belongs to, so its marginal manufacturing cost is close to zero: no separate operation, no additional part number, no fastener inventory, no supplier. Because it is moulded rather than assembled, its position is held by the tool and therefore repeats to the accuracy of the mould rather than to the accuracy of a hole pattern in two parts. Designed properly it is a bump-off feature that releases on the normal opening stroke of the tool, so it needs no side-action or lifter, and the mould stays simple — the single largest determinant of tooling cost and lead time. Its design rules are well established: constant or tapered beam section to keep the bending stress uniform, a generous root radius to avoid the stress concentration that causes snap fracture, a return angle chosen for the intended service (about 45° for a serviceable joint, near 90° for a permanent one), and a deflection limited to the allowable strain of the polymer, which is the discipline that keeps the joint from creeping or breaking on the second assembly.

Features that enhance assembleability. It is a single top-down insertion along one axis, which is the ideal assembly motion for both a human and a robot, and it requires no tool, no torque control and no consumable. The lead-in chamfer makes the joint self-locating and self-aligning, so the operator does not have to achieve the alignment — the feature does it — and that also widens the tolerance band the joint will accept. It gives an audible and tactile click on seating, which is error-proofing in the poka-yoke sense: a joint that is not fully home announces itself. Made symmetric, or given an obvious asymmetry, it eliminates the orientation decision. In Boothroyd and Dewhurst timings a snap-fit insertion runs to roughly two or three seconds against ten or more for a screw that must be picked, oriented, started, driven and torque-checked, and it removes a part that could be dropped, cross-threaded or omitted.

The honest limitations, which a complete answer should state: a snap-fit is not readily re-workable and repeated cycling degrades it; it does not develop a controlled clamp load, so it is unsuitable where a gasket must be compressed to a specification; it is difficult to service without a release feature; and it is a stress-concentrating feature in a polymer, so it is sensitive to creep, to temperature and to chemical attack. Where those matter, other DFMA-consistent strategies apply the same logic — a single captive screw instead of four loose ones, an integral living hinge, heat-staking or ultrasonic welding for a permanent joint, or a press-fit — and the choice is made against the service and sealing requirements rather than by reflex.

Part C — When DFMA should be applied, and when it is too late (2 marks)

DFMA belongs in concept development and system-level design — that is, at the point where the architecture and the part count are being decided, and before detail design begins. This is a direct consequence of the committed-cost curve: by the end of concept development something like 70 to 80 per cent of the product’s lifetime cost is already committed by the decisions taken, while only a few per cent has actually been spent. Once the architecture is fixed, DFMA can only tidy the details, because the questions it is best at — does this part need to exist, can these three be one, is this the right process for this volume — are architecture questions that have already been answered. In practice this means a manufacturing engineer sits on the design team from the first concept review, not as a reviewer afterwards; this is the core of concurrent engineering.

It is too late once production tooling has been committed. Cutting a mould or a progressive die converts a design decision into a capital asset with a lead time, and after that point a part-count change means scrapping or reworking the tool, requalifying the part, and re-running the validation. The practical no-return points, in order, are the design freeze at the end of detail design, the release of tooling to the toolmaker, and finally production part approval and the start of the ramp; a change after the first is expensive, after the second is very expensive, and after the third is usually refused. A DFMA study delivered as a cost-reduction exercise on a product already in production is not DFMA at all — it is value engineering on an inherited architecture, and it recovers a small fraction of what the same effort would have found before the concept was frozen.

Part D — Manual versus automated assembly considerations (4 marks)

The two routes reward different part characteristics, and a design optimised for one can be actively hostile to the other, so the route must be chosen before the parts are detailed.

Manual assembly. The human operator is extraordinarily adaptable — able to see, to feel a part seat, to cope with a flexible cable or a part that arrives in an unexpected orientation, and to detect a defect while working. The design considerations are therefore mostly about the human body and human attention. Parts must be graspable and must not tangle, nest or stick together in the bin, must not be so small or so slippery that they need tweezers, and must not have sharp edges. The assembly must be reachable without obstructed vision or awkward posture, and assembly forces and part masses must sit inside ergonomic limits or the line will generate injuries. Because the failure mode is human error, the design should be error-proofed: asymmetric where orientation matters so that the wrong way will not fit, symmetric where it does not so that no decision is needed, with a colour or shape code on parts that could be confused, and audible or tactile confirmation of seating. Manual assembly carries a low fixed cost and a high, roughly constant unit cost, so it suits low and moderate volumes, high product variety and frequent design change.

Automated assembly. Automation removes the adaptability, and everything the human used to absorb must now be designed in. The dominant new requirement is feedability: a part must be able to be delivered oriented, which for a vibratory bowl feeder means it must not tangle or nest, must have a feature that passive tooling can use to orient it, and must not have a near-symmetry that makes the correct orientation ambiguous — a part that is almost symmetric is far worse to automate than one that is plainly asymmetric. Parts must be rigid and dimensionally consistent, because a robot cannot feel a flexible part into place; incoming variation that a human would absorb becomes a jam. Insertion should be along a single vertical axis with generous chamfers, because self-alignment substitutes for the human hand. Sensing and inspection must be added explicitly, since nothing is watching. And the economics invert: automation carries a high fixed cost in the cell and its tooling and a low unit cost, so it pays only above a break-even volume, and it is inflexible in the face of product change.

Given. A manual station assembles the product at 2.40 CAD per unit with negligible fixed cost. An automated cell costs 120,000 CAD to install and tool and assembles the same product at 0.30 CAD per unit. Find. The annual volume at which the two routes cost the same.

1. Write both routes as a cost per unit and equate them. With a fixed investment $T$ amortised over $n$ units and a variable cost $u$, the unit cost of either route is $c(n) = T/n + u$. Equating the two routes and solving for the break-even volume gives $n^* = (T_a - T_m)/(u_m - u_a)$, and substituting, $$n^* = \frac{120{,}000 - 0}{2.40 - 0.30} = \frac{120{,}000}{2.10} = \boxed{57{,}143\ \text{units}}$$ so below roughly 57,000 units a year the manual route is cheaper and above it the automated cell is. The number is deliberately arranged to be reachable without a calculator: 2.10 into 120,000 is 120 divided by 2.1 thousand.
2. Read what the break-even does and does not settle. At twice the break-even volume the automated route costs $120{,}000/114{,}286 + 0.30 = 1.35$ CAD per unit against 2.40 CAD manual, a saving worth having; but the comparison assumes the design is feedable, and if it is not, the true automated fixed cost includes the redesign and possibly a more elaborate feeding solution, which can move $n^*$ by a factor of two. This is the reason the route must be chosen before the parts are detailed: the break-even is a property of the design as much as of the volume.