Question 2 of 7: The engineer and the artist as designers

Question 2: The engineer and the artist as designers (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 — Differences in goals

The engineer's goal is to satisfy a specification: a set of requirements, each with a metric, a target and a verification method, that has been derived from a stated need and bounded by law, safety, cost and manufacturability. Success is external to the designer and objectively decidable — the chair either passes its static-load proof test or it does not — and it is decided by test rather than by argument. The engineer also carries a duty that the artist does not: under the Engineers and Geoscientists BC Code of Ethics the engineer must hold paramount the safety, health and welfare of the public and the protection of the environment, which converts certain requirements from preferences into obligations that survive commercial pressure.

The artist's goal is expressive: to produce an artefact that carries meaning or provokes a response. Success is judged by reception, is contested and revisable, and is not verifiable in the engineering sense — there is no test that a sculpture passes. The artist is also free to define the problem entirely, and indeed the choice of problem is often the work; the engineer inherits the problem from a stakeholder and is judged on how well it was solved, not on how interesting it was.

Two consequences follow that are worth naming because they drive parts B to D. First, the engineer optimises under constraint while the artist explores without one, so the engineer's design space is bounded and the artist's is not. Second, the engineer's output must be reproducible by other people — a hundred thousand times, by workers who never met the designer — whereas the artist's output is usually the singular artefact itself. That single difference explains most of what follows. It is also why the comparison is not a ranking: industrial design sits between the two and draws on both, and an engineer who cannot recognise when a requirement is really an aesthetic judgement will specify the wrong thing.

Part B — Comparison of the design processes

The engineer follows a structured, convergent, gated process: clarify the task and build a requirements list, generate concepts, select one by an explicit and recorded method, embody it, detail it, verify it against the requirements, and release it under change control. Iteration is present throughout but it is instrumented — each pass is expected to shrink a measurable gap between the current design and the specification, and the process stops when the residual gap is acceptable. The artist's process is exploratory and divergent: the criteria emerge with the work, an iteration may deliberately move away from resolution, and there is no gate that must be passed before proceeding. Both iterate; only one of them is obliged to converge.

Given. A design gap that each review pass closes by a constant fraction, so that the residual after $k$ passes is $g_k = \rho^{\,k}$ with $\rho = 0.62$; a requirement that the residual fall to 3 per cent of the original gap. Find. The number of iterations the engineer must plan for, and what each one costs at different stages.

1. Model convergence as a geometric decay and solve for the pass count. If every review closes the same fraction of what remains, the residual gap is geometric, so the number of passes needed to reach a target residual $g_t$ is $$k = \frac{\ln g_t}{\ln \rho} = \frac{\ln 0.03}{\ln 0.62} = \frac{-3.5066}{-0.4780} = 7.34 \Rightarrow \boxed{k = 8\ \text{passes}}$$
2. Attach a cost to when those passes happen. Empirically the cost of one change rises about tenfold at each stage boundary: roughly CAD 1 200 at concept, CAD 12 000 in detail design, CAD 120 000 once hard tooling is cut, and CAD 1 200 000 for a field recall. Eight passes taken at concept cost about CAD 9 600; the same eight taken after tooling cost about $\boxed{\text{CAD } 960\,000}$, a factor of one hundred for identical engineering work.
3. Read the design-process implication. The engineer therefore front-loads iteration deliberately — sketch models, analysis, simulation and cheap prototypes — because the number of passes is set by the difficulty of the problem and cannot be wished away, while their cost is set entirely by scheduling. The artist has no equivalent forcing function, because no tooling is committed and no third party has to reproduce the result.

Part C — Differences in the end customers

The engineer serves a chain of customers, most of whom never buy anything. The user is only the most visible: the purchaser may be a different person entirely (a facilities manager, a parent, a procurement department), and behind them stand the regulator who must be satisfied, the assembly operator who has to build it, the service technician who has to repair it, the retailer who has to ship and display it, and the recycler who has to take it apart. Each has acceptance criteria that the design must meet, and they conflict: what is quickest to assemble is not always easiest to service, and what displays best is not always cheapest to ship. Further, the user is a population with a distribution, not an individual, so the engineer must state what fraction of that population is accommodated and accept that the remainder is excluded by design.

The artist serves an audience, which is self-selecting and need not be satisfied at all. Nobody is harmed by not liking a painting, no regulator approves it, and there is no obligation to accommodate the fifth percentile of anything. The artist may also legitimately aim at a small audience, or at posterity, or at themselves; an engineer who designs a chair that fits only people like the designer has made an error, not a choice.

Part D — Communicating the design details to manufacturing and to the end customer

This is where the two practices separate most sharply, because the engineer's design is not the artefact but the information required to make the artefact. That information has to survive transfer to people who were not present when the decisions were made, who may work for a different company on a different continent, and who will interpret any ambiguity in the way that is cheapest for them. Ambiguity is therefore not a stylistic weakness in an engineering drawing, it is a defect. The artist faces no such transfer: the work communicates itself, interpretation latitude is a feature rather than a fault, and where fabrication is delegated the artist supervises it personally.

The concrete form the engineer's problem takes is tolerancing. A phrase such as "close fit" cannot be manufactured, inspected or argued about; a dimension with a tolerance and a datum reference frame can. The following illustrates why the choice of tolerancing model is itself a communication decision.

Given. An assembly gap with a nominal value of 1.20 mm built from five components, each toleranced at $\pm 0.10$ mm; the gap must remain within 0.90 to 1.50 mm for the mechanism to function. Find. Whether the design is acceptable on a worst-case and on a statistical basis, and what each answer commits the manufacturer to.

1. Add the tolerances arithmetically for the worst case. If every component is allowed to sit at its extreme simultaneously, the limits add directly: $$T_{WC} = \sum_{i=1}^{5} t_i = 5 \times 0.10 = \boxed{\pm 0.500\ \text{mm}}$$ giving a gap of 0.70 to 1.70 mm, which violates the 0.90-1.50 mm requirement at both ends.
2. Add them statistically. If each component dimension is independent and approximately normal with its tolerance at three standard deviations, the variances add and the tolerances combine as a root sum of squares: $$T_{RSS} = \sqrt{\sum_{i=1}^{5} t_i^{2}} = 0.10\sqrt{5} = \boxed{\pm 0.224\ \text{mm}}$$ giving 0.976 to 1.424 mm, comfortably inside the requirement.
3. Quantify the residual risk of the statistical route. With $t_i = 3\sigma_i$, $\sigma_i = 0.0333$ mm and $\sigma_{\text{stack}} = 0.0333\sqrt{5} = 0.0745$ mm, so the requirement limits sit at $z = 0.30/0.0745 = 4.03$ standard deviations and the expected out-of-tolerance rate is $\boxed{57\ \text{ppm}}$.
4. State what has actually been communicated. The worst-case answer is a promise the drawing alone can keep; the RSS answer is a promise that depends on the manufacturer holding each process centred and capable, which is an agreement about process, not about geometry. Choosing RSS and not saying so transfers a risk to the factory silently — the clearest example there is of a design detail that must be communicated explicitly, and one for which the artist has no analogue.

Communicating to the end customer inverts the problem. Here the engineer must suppress the very detail that manufacturing needs and translate performance into consequence: not "seat height 375-505 mm" but "fits 96 per cent of adults", not "passes the BIFMA static-load proof test" but a stated user weight rating. The artist again has the easier task, since the work addresses its audience directly. The failure mode for the engineer is to publish the specification instead of its meaning, which leaves the customer unable to tell a genuine improvement from a number.