The Rattle That Only Appears When Warm
A team is finishing a battery-powered handheld pump. In room-temperature bench runs the unit is quiet, but in an environmental chamber at 45 °C a sharp rattle appears between 60% and 80% of maximum speed, and disappears above 85%. The rattle is intermittent: three of ten units show it, and only after 20 minutes of running. The team has a launch review in two weeks and a functional prototype build of 30 units. A junior engineer proposes adding a rubber damper to the motor mount and shipping if the rattle goes away. A senior engineer wants to delay launch until the root cause is found. The program manager asks for a decision with evidence, not opinion. Budget allows one more design iteration and one more chamber test campaign.
1. Evaluate the junior engineer's damper proposal against the senior engineer's position. What should the team do, and what evidence would justify each path?
Read the worked answer
The damper is a symptom treatment, and the symptom is narrow: 60–80% speed, hot, 3 of 10 units, after 20 minutes. A fix validated only against that exact signature risks masking a mechanism that will shift with tolerance stack, wear, or a different supplier lot.
The senior position is stronger because the failure is intermittent and temperature-dependent, which points to a clearance or preload change with thermal expansion, not a generic vibration problem. The right move is a short root-cause sprint before committing the iteration.
Evidence to gather: (1) instrument the suspect interfaces with accelerometers and compare spectra cold vs hot to locate the source; (2) measure critical clearances on a rattling and a quiet unit at 45 °C; (3) run a design-of-experiments varying the suspected gap and speed to see if the 60–80% band moves predictably.
Decision rule: if the root cause is identified and the fix is a geometry or preload change, take the iteration. If two weeks is genuinely insufficient, ship with a damper only if it is backed by a demonstrated mechanism and a test that stresses the mechanism, not just the observed signature.
2. The team has time for one chamber campaign. Design the test plan: what units, conditions, and pass criteria would give the most decision value?
Read the worked answer
Maximize information per run rather than repeating the same condition. Build a matrix that spans the suspected mechanism.
Units: include at least 10 production-representative units, deliberately spanning the tolerance extremes of the suspected gap (measure and bin them), plus the three known rattlers as positive controls and two known-quiet units as negative controls.
Conditions: cold (room), 45 °C, and one hotter point (e.g., 55 °C) to see whether the rattle band moves with temperature as a thermal-clearance mechanism predicts. Sweep speed slowly through 50–100% in fine steps, and run long enough (≥45 min) to catch the time-dependent onset.
Measurements: accelerometer spectra, not just audible pass/fail, so the band location and amplitude are quantified.
Pass criteria: no rattle across all units and conditions, and the spectral signature absent at the tolerance extremes. Prediction test: if the mechanism is thermal clearance, the band should shift downward in speed at 55 °C. If it does not, the hypothesis is wrong and the damper is unjustified.
3. The two-week deadline holds regardless. How should the team communicate risk and the go/no-go recommendation to leadership?
Read the worked answer
Separate what is known from what is inferred, and state the decision explicitly with its residual risk.
Structure: (1) Observation - rattle at 60–80% speed, 45 °C, 3/10 units, delayed onset. (2) Leading hypothesis - thermally driven clearance change, with the evidence for and against. (3) What the one campaign will and will not resolve. (4) Recommendation with a confidence level and a fallback.
A defensible recommendation: proceed to the campaign; if root cause is confirmed and the geometry fix validates, launch on schedule. If the campaign is inconclusive, do not silently ship the damper. Instead, either narrow the shipped speed range in firmware (a controlled, testable mitigation that removes the operating band) or delay, and say which.
Quantify residual risk: with 3/10 showing the issue, the true rate could be materially higher; note the confidence interval and that 10 units cannot prove a low defect rate. Give leadership a clear choice between schedule, mitigation scope, and residual field risk rather than a vague "we're investigating."