industrial//reliability//bathtub curve

The bathtub curve is the shape the failure rate of a population of units takes over their life (high at the start, low and flat in the middle, rising at the end), and it is used to decide where to spend effort: screening and process control against the first wall, spares for the floor, replacement schedules against the last wall. It is a picture of the hazard rate described in reliability.

The bathtub curve is the shape the failure rate of a population of units takes over their life (high at the start, low and flat in the middle, rising at the end), and it is used to decide where to spend effort: screening and process control against the first wall, spares for the floor, replacement schedules against the last wall. It is a picture of the hazard rate described in reliability.

Each part of the curve has its own cause. Early failures come from defects of manufacture and assembly that kill weak units quickly, so the rate falls as they are weeded out; the flat bottom is random failure from causes unrelated to age; the late rise is wear-out, materials degrading with time and use. The three overlap, and the curve is their sum.

Telling the phases apart changes the remedy. A rate that falls with time is fixed before shipping, by design, screening and burn-in; a rising rate is managed in the field, by replacing units before they reach it. Replacing units early during an early-failure problem only brings in fresh units with the same defects.

Not every product shows all three walls. Software and many electronic parts show little wear-out, and a slow process such as insulation yielding to years of warm salt water stays invisible until units have run that long, which is why the late wall of a long-lived device is so expensive to measure.

The curve belongs to the population. One unit's life says little about it, and estimating the late wall needs units that have survived long enough to reach it.