The bathtub curve is a reliability engineering model that plots an asset's failure rate over its entire lifetime, and it shows three distinct phases: a high but falling failure rate early on (infant mortality), a low and roughly constant rate during the useful life, and a rising rate at the end (wear-out). Plotted together, these phases trace a shape like the cross-section of a bathtub, with steep sides and a long flat bottom. The curve is a foundational way to reason about when and why machines fail, and it directly shapes how a plant should schedule inspections, replacements, and maintenance strategy.
The bathtub curve combines three separate failure behaviors into one lifetime picture:
Not every component follows all three phases cleanly. Electronics often show strong infant mortality and a very long flat region, while mechanical parts like bearings and seals show a pronounced wear-out region.
The early steep section reflects units that were flawed from the start. Common causes include a defective weld, a mis-torqued fastener, contaminated lubricant, a wiring error, or damage during transport and installation. Because these defects surface quickly under load, manufacturers use burn-in: running equipment under controlled stress before shipment or before full production so latent defects fail in the factory rather than on the plant floor.
On the maintenance side, this phase is why commissioning checks, alignment verification, and early inspections matter so much on new or rebuilt assets. Tracking early failures with a structured method such as FMEA helps you separate one-off defects from systematic problems that will keep recurring.
The flat bottom is where a well-run asset spends most of its life. Here failures are random and largely independent of age, so replacing a part simply because it has been in service for a fixed number of hours does not reduce risk. This is the single most important insight the bathtub curve gives maintenance planners.
During this region the failure rate connects directly to reliability metrics. A constant failure rate produces a stable MTBF and MTTR profile, which makes downtime predictable enough to plan around. For non-repairable components, the analogous measure is MTTF. Understanding the mix of causes behind stoppages here overlaps heavily with diagnosing unplanned downtime.
Because failures in the useful-life region are random rather than age-driven, calendar-based or hours-based replacement wastes healthy component life and can even introduce fresh infant-mortality risk every time you open a machine. The better fit is condition-based maintenance, where you act on the actual measured state of the asset (vibration, temperature, current draw, output quality) instead of a fixed schedule.
This is the core logic behind modern proactive maintenance and the discipline of total productive maintenance: intervene when the evidence says the asset is drifting, not when the calendar says so. Predictive maintenance, an industry concept that forecasts failures from trended data, extends this idea further, though it depends entirely on having clean, continuous machine data to work from.
Failure rate is usually written as the Greek letter lambda and expressed in failures per unit time. Suppose a fleet of 200 identical pumps runs for 5,000 hours during their useful-life phase, and 8 pumps fail in that window.
Now compare the three phases. If the same fleet showed 20 failures in the first 500 hours (infant mortality) but only 8 failures across the later 5,000-hour window, the early failure rate is roughly 20 divided by 100,000 pump-hours, or 0.0002 per hour, which is 25 times higher than the useful-life rate. Later, if wear-out drove 30 failures in a final 2,000-hour window, that rate climbs to 30 divided by 400,000 pump-hours, or 0.000075 per hour, nearly 10 times the flat-region rate. Those three numbers (high, low, rising) are the bathtub curve expressed as arithmetic.
The curve implies a different tactic for each phase:
To act on any of this you need reliable failure data, and that is where a CMMS and real-time monitoring come in. Fabrico is not a predictive-maintenance oracle or a digital-twin simulator; it is the data foundation that makes the bathtub curve usable. Its CMMS product logs work orders, assets, and spare parts so you can compute real failure rates per machine, while its OEE product and camera-based monitoring capture stoppage and quality data even on machines with no PLC. Pairing that with OEE tracking and Pareto analysis tells you which phase each asset is actually in.
No. The bathtub curve is a general model, not a universal law. Many electronic components show a long flat region with almost no wear-out, while some mechanical assets show wear-out almost immediately with little useful-life plateau. Reliability studies have shown that only a minority of failure modes are genuinely age-related, which is exactly why condition monitoring often beats fixed-interval replacement.
MTBF describes average time between failures during the flat, constant-rate useful-life region, where it equals the reciprocal of the failure rate. It is most meaningful in that middle phase. During infant mortality and wear-out the failure rate is changing, so a single MTBF figure can be misleading and must be read alongside which phase the asset is in.
Burn-in is running equipment under controlled stress before deployment so early-life defects fail during testing rather than in production, effectively climbing down the steep infant-mortality slope in the factory. It helps most where infant mortality is significant, such as electronics and complex assemblies, and adds little value for components whose failures are dominated by random or wear-out mechanisms.
Turn the bathtub curve from theory into per-machine numbers: capture real failure, downtime, and quality data across every asset with Fabrico's real-time OEE and CMMS platform, including camera monitoring for machines without a PLC. Book a Fabrico demo to see your own reliability curve take shape.
Вход
Блог