Key Points
- Remaining useful life (RUL) is the estimated operating time left before an asset fails or needs intervention, measured in hours, cycles, or days.
- The formula is RUL = total useful life - current age. The work is predicting the failure time using a health indicator, a failure threshold, and a degradation model.
- Three proven methods calculate it: physics-based models, data-driven models (statistical and machine learning), and hybrid models that combine both.
Every maintenance team wants the same thing: to fix the machine before it breaks, not after. Remaining useful life is the number that makes that possible. Get it right and you schedule the repair on your terms, order the part in time, and keep the line running. Get it wrong and you are either replacing healthy components too early or explaining unplanned downtime to your plant manager.
This guide walks through what remaining useful life is, the formula behind it, the exact steps to calculate it, and the methods that make each estimate more or less reliable. It also answers the question most teams are actually weighing at this point: whether to calculate remaining useful life by hand, asset by asset, or let a platform do it continuously across your whole plant.
Why Calculating Remaining Useful Life Matters
Calculating remaining useful life matters because it converts uncertainty into a schedule. When you know how much life a bearing, motor, or gearbox has left, you can make four decisions that directly protect uptime and budget:
- Time the repair. Schedule maintenance during planned downtime instead of reacting to a breakdown at 2 a.m.
- Order parts on time. Stock the right spare before you need it, without tying up cash in inventory you do not.
- Decide repair versus replace. Compare the remaining life of a component against the cost of a rebuild versus a new unit.
- Protect people. Catch failures in critical equipment before they become safety events.
The human payoff is real. Reliability that you can predict gives your team its weekends back, because the failures that used to happen at the worst possible moment now show up on a maintenance calendar weeks in advance.
The Remaining Useful Life Formula
At its simplest, the remaining useful life formula is:
RUL = Total Useful Life − Current Age
In other words, take how long the asset can run in total and subtract how long it has already run. What is left is the remaining useful life.
The catch is the words "total useful life." You rarely know the full lifespan up front, because it depends on how the asset is actually wearing right now. So in predictive maintenance, the same formula is written with the failure point as a prediction that updates over time:
RUL(t) = T_f − t
Where:
- t is the current time (or current age in cycles or runtime hours).
- T_f is the predicted time of failure, the point at which total useful life is reached.
- RUL(t) is the remaining useful life at the present moment.
The formula itself is trivial. All of the engineering effort goes into estimating T_f accurately, because that value is not a fixed spec sheet number. It is a live prediction pulled from the asset's real condition. That is what separates a rough guess from a number you can bet a production schedule on, and it is the entire reason the methods below exist.
For assets modeled with reliability statistics, teams often work from a Weibull reliability function to describe the probability an asset survives to a given time:
R(t) = e^−(t/η)^β
Where η is the characteristic life and β is the shape parameter that describes how failure risk changes over time. From this curve you can derive the expected remaining life for a population of similar assets. This population-level view is useful for planning, but for a single machine on your floor, condition data will always give you a sharper answer.
How to Calculate Remaining Useful Life: Step by Step
Follow these six steps to calculate the remaining useful life of a specific asset.
- Collect condition monitoring data. Gather the signals that reflect the asset's health: vibration, temperature, ultrasound, oil analysis, or motor current. Continuous data is far more useful than occasional manual readings because degradation shows up as a trend, not a single point.
- Define a health indicator (HI). Reduce your raw data to one value or a small set of values that reliably move as the asset wears. For a bearing, this might be a vibration amplitude at a specific fault frequency. A good health indicator rises (or falls) consistently as damage progresses.
- Set a failure threshold. Decide the health indicator value at which the asset is considered failed or at unacceptable risk. This can come from a standard such as ISO 20816 for vibration severity, from the manufacturer, or from your own failure history on similar machines.
- Model the degradation trend. Fit a curve to how the health indicator has changed over time. This can be a straight line, an exponential curve, or a machine learning model, depending on how the asset degrades.
- Project to the threshold. Extend the degradation model forward to find the time when the health indicator will hit the failure threshold. That time is your predicted failure time, T_f.
- Calculate and update RUL. Apply the formula: RUL = T_f − current time. Then repeat the whole process as new data arrives. Every fresh reading sharpens the estimate.
The last step is the one teams skip and regret. Remaining useful life is only as good as its most recent update. A number calculated last quarter is a historical fact, not a prediction.
Methods for Calculating Remaining Useful Life
There are three established approaches. The right one depends on how well you understand the failure physics and how much data you have.
Physics-Based Models
Physics-based models use the known science of how a component fails: crack growth, fatigue, wear, or corrosion. A classic example is using Paris' law to model how a crack propagates under repeated stress until it reaches a critical length.
These models are highly accurate when you truly understand the failure mechanism and can measure the inputs it needs. The drawback is that most real industrial systems are too complex to model this way, and building a physics model for every asset on a plant floor is rarely practical.
Data-Driven Models
Data-driven models learn the degradation pattern from historical and real-time data rather than from first-principles physics. They split into two families:
- Statistical and reliability models use failure history to estimate probabilities. Weibull and exponential distributions, along with MTBF analysis, fall here. They are strong for fleet-level planning and weaker for pinpointing a single machine.
- Machine learning models train on sensor data to predict RUL directly. Regression models, random forests, and neural networks such as LSTMs can learn subtle degradation signatures across many variables at once. They shine when you have large volumes of good condition data and complex assets where the physics are unclear.
Hybrid Models
Hybrid models combine physics-based understanding with data-driven learning. You might use a physics model to describe the general degradation shape and a machine learning layer to correct it based on real sensor behavior. Hybrid approaches often deliver the best accuracy, at the cost of more effort to build and maintain.
Method Comparison
| Method | Best for | Data needed | Main limitation |
|---|---|---|---|
| Physics-based | Well-understood single failure modes | Physical parameters, load data | Hard to scale across a plant |
| Statistical / reliability | Fleet-level planning | Historical failure records | Weak for individual assets |
| Machine learning | Complex assets, lots of sensor data | Large, labeled condition datasets | Needs quality data and tuning |
| Hybrid | Maximum accuracy on critical assets | Physics knowledge plus sensor data | Most complex to build |
A Simple Calculation Example
Here is a worked example using a straightforward degradation trend.
A pump bearing is monitored by vibration. Its health indicator is the velocity amplitude in mm/s. Readings over recent weeks look like this:
- Week 0: 2.0 mm/s
- Week 4: 3.0 mm/s
- Week 8: 4.0 mm/s
The health indicator is rising by 1.0 mm/s every four weeks, a linear trend of 0.25 mm/s per week. The team has set the failure threshold at 7.1 mm/s based on ISO 20816 severity guidance for this machine class.
At week 8, the current value is 4.0 mm/s. The gap to the threshold is 7.1 − 4.0 = 3.1 mm/s. At 0.25 mm/s per week, that gap closes in 3.1 ÷ 0.25 = 12.4 weeks.
Remaining useful life ≈ 12 weeks.
That gives the team roughly three months to order the bearing, schedule the swap during planned downtime, and avoid an unplanned pump failure. Notice that the estimate depends entirely on the trend holding steady, which is exactly why continuous monitoring and regular recalculation matter. If the degradation accelerates, the next update will catch it.
What Data Do You Need to Calculate RUL?
The most useful inputs for calculating remaining useful life on rotating and industrial equipment are:
- Vibration for imbalance, misalignment, looseness, and bearing wear.
- Temperature for friction, lubrication problems, and electrical faults.
- Ultrasound for early-stage bearing defects and lubrication condition.
- Oil analysis for wear particles and contamination in lubricated systems.
- Motor current for electrical and mechanical faults in motors and drives.
The quality of these inputs decides the quality of the RUL estimate. Sparse or noisy data produces a shaky degradation model, and a shaky model produces a number nobody trusts.
Common Challenges in Calculating Remaining Useful Life
Even with the right method, several issues degrade accuracy:
- Noisy signals. Real sensor data is messy, and a bad health indicator can hide the true degradation trend.
- Setting thresholds. Deciding exactly when an asset counts as failed is often judgment as much as science.
- Variable operating conditions. An asset that runs at different loads and speeds degrades at different rates, which complicates any single model.
- Not enough failure data. Machine learning models need examples of run-to-failure behavior, and reliable assets, by definition, do not fail often.
The through-line is that remaining useful life calculation is a data problem before it is a math problem. Solve the data quality and consistency first, and the modeling gets far easier.
Should You Calculate Remaining Useful Life Manually or Use a Platform?
You can run every step above by hand: collect readings on a route, build a health indicator, set thresholds, fit a curve, and recalculate after each new reading. On one or two critical assets, that works. Across a plant of hundreds of machines, it stalls. Routes get skipped, spreadsheets go stale, the analysis waits on one specialist, and the estimate is only as fresh as the last manual update.
That is the real decision at this stage. Not whether remaining useful life is worth calculating, but whether your team should keep doing it by hand or let a system do it continuously and at scale.
How Tractian Calculates Remaining Useful Life for You
Tractian removes the manual work and closes the gaps that trip up a hand-run program.
The Smart Trac sensor captures vibration, ultrasound, temperature, and RPM continuously from your rotating equipment and streams it to the platform over 4G/LTE. That solves the first and hardest step: clean, continuous condition data on every monitored asset, without a technician walking a route.
From there, Tractian's AutoDiagnosis engine, trained on billions of collected samples, does the modeling you would otherwise build yourself. Instead of handing you a raw vibration spectrum to interpret, it identifies the specific failure mode and its severity stage, then ranks it by asset criticality. A gearbox anomaly does not arrive as a generic alarm. It arrives as a named diagnosis with the evidence attached.
Because the diagnosis is that specific, it drops cleanly into the tools you already run. Tractian integrates with your existing CMMS, so a detected fault can automatically generate a work order with a recommended procedure. You go from signal, to diagnosis, to prioritized action, in one connected chain.
That is remaining useful life put to work: not a number on a dashboard, but a repair on the calendar before the failure happens.
Frequently Asked Questions
What is remaining useful life in maintenance? Remaining useful life is the estimated operating time left before a piece of equipment fails or needs maintenance. It is a core output of predictive and condition-based maintenance programs.
What is the formula for remaining useful life? The simplest form is RUL = Total Useful Life − Current Age. In predictive maintenance, the failure point is a prediction rather than a known value, so it is written RUL(t) = predicted failure time − current time. The challenge is predicting that failure time, which requires condition data, a health indicator, a failure threshold, and a degradation model.
How do you measure remaining useful life? You measure it by tracking a health indicator (such as vibration amplitude) over time, modeling how it degrades, and projecting when it will cross a defined failure threshold. The time until that point is the remaining useful life.
What is the difference between RUL and MTBF? MTBF (mean time between failures) is an average across a population of assets and is useful for planning. RUL is a forward-looking estimate for a specific asset based on its actual current condition, which makes it far more precise for scheduling individual repairs.
Can remaining useful life be calculated automatically? Yes. Platforms like Tractian pair continuous condition monitoring sensors with AI diagnosis to estimate failure modes and severity automatically, removing the need for manual spectrum analysis or a dedicated data science team.
See it on your own equipment. Book a 20-minute demo, and we will show you how Tractian estimates remaining useful life across your assets and turns those estimates into scheduled work before the failure happens. That is what fewer unplanned breakdowns, and more predictable weekends, actually look like.
Director
Director
Alex Vedan, Marketing Director at Tractian, develops impactful strategies that empower industrial clients across North America and LATAM to achieve operational excellence. By aligning innovation with customer needs, he ensures Tractian solutions drive meaningful improvements in efficiency and reliability.
Director
Alex Vedan, Marketing Director at Tractian, develops impactful strategies that empower industrial clients across North America and LATAM to achieve operational excellence. By aligning innovation with customer needs, he ensures Tractian solutions drive meaningful improvements in efficiency and reliability.
See more
Share
