Understanding Prognosis in Predictive Maintenance

Devices

Portable balancer & Vibration analyzer Balanset-1A

€1,975.00 + VAT (if applicable)

Parts

Vibration sensor

€90.00 + VAT (if applicable)

Parts

Optical Sensor (Laser Tachometer)

€124.00 + VAT (if applicable)

Devices

Balanset-4

€6,803.00 + VAT (if applicable)

Parts

Magnetic Stand Insize-60-kgf

€46.00 + VAT (if applicable)

Parts

Reflective tape

€10.00 + VAT (if applicable)

Devices

Dynamic balancer “Balanset-1A” OEM

€1,735.00 + VAT (if applicable)

Prognosis (also called failure prediction, and closely tied to remaining-useful-life estimation) is the process of estimating how much time remains before a detected fault causes functional failure or demands intervention. Prognosis follows fault detection (knowing a problem exists) and diagnosis (knowing what the problem is), and answers the decisive question — “When must we act?” — by analysing vibration progression trends, the characteristics of the fault type, and the machine’s operating conditions.

Accurate prognosis is what makes predictive maintenance genuinely predictive. It allows maintenance to be scheduled at the optimal moment — neither too early, which wastes remaining life, nor too late, which courts failure — and it underpins the procurement of long-lead spares, the allocation of labour and tools, and the coordination of production. Within the broader scheme of condition-based maintenance, prognosis is the forward-looking stage that turns “this machine is sick” into “this machine must be repaired by week eight.”

1. The Three-Stage Chain: Detection, Diagnosis, Prognosis

It helps to see prognosis as the third link in a chain. Detection notices that a parameter has stepped outside its baseline. Diagnosis identifies the mechanism — a bearing defect, misalignment, looseness. Prognosis then projects that mechanism forward in time. Each stage depends on the one before: you cannot forecast the future of a fault you have not yet correctly named, which is why a confident diagnosis is the foundation of any trustworthy prediction. The whole chain is formalised in monitoring standards such as ISO 13374, which lays out detection, diagnosis, and prognosis as distinct data-processing functions.

2. Prognostic Methods

Trend Extrapolation

The most common and practical method, and the natural extension of routine trend analysis:

• Plot the historical vibration data against time.
• Fit a trend line — linear, exponential, or otherwise.
• Extrapolate to find when the alarm or failure threshold will be crossed.
• Update the prediction with every new measurement.
• Accuracy: moderate (it assumes the trend continues).
• Requirements: sufficient trending history — at least six data points.

Physics-Based Models

• Built on an understanding of the failure physics (crack growth, spall propagation).
• The model predicts progression from stress, cycles, and environment.
• Examples: the Paris Law for crack growth, or bearing L10 fatigue-life calculations.
• Accuracy: good, when the model parameters are known.
• Requirements: detailed equipment and operating data.

Experience-Based (Historical Data)

• Built on past failures of similar equipment.
• Uses typical progression rates drawn from history.
• Relies on empirical relationships (vibration level → time to failure).
• Accuracy: fair, and equipment-specific.
• Requirements: a historical failure database.

Statistical / Machine Learning

• Algorithms trained on historical progression data.
• Pattern recognition across many similar cases.
• Produces probabilistic predictions.
• Accuracy: potentially very good, given enough data.
• Requirements: a large dataset and computational resources.

3. Factors Affecting Prognosis Accuracy

Trending Data Quality

• More data points sharpen the trend definition.
• Consistent measurements yield reliable trends.
• Adequate history (months, at minimum) is essential.
• Clean data, with outliers identified, prevents false slopes.

Fault Progression Characteristics

• Predictable progression: easier to forecast — for example gradual bearing wear.
• Accelerating progression: harder — spall growth is roughly exponential.
• Erratic progression: difficult — looseness and intermittent rubs.
• Sudden failures: essentially unpredictable — a shaft fracturing from a crack.

Operating Condition Stability

• Stable conditions support reliable predictions.
• Variable loads and speeds make predictions less certain.
• Process changes can either accelerate or slow the progression.

4. Remaining Useful Life (RUL) Estimation

The headline output of prognosis is the remaining useful life: the time from the current condition to a failure or intervention threshold.

How It Is Expressed

• Stated in operating hours, calendar days, or cycles until intervention is required.
• Updated continuously as new data arrive.
• Reported as an estimate carrying genuine uncertainty.

Confidence Intervals

• RUL is an estimate, not a fact.
• Best expressed as a range — for example “30–90 days with 90% confidence.”
• Uncertainty shrinks as failure approaches and more data accumulate.
• Conservative estimates are appropriate for critical machinery.

Worked Example

• A bearing defect is detected at 2 g envelope amplitude.
• Historical progression: 2 g → 10 g (alarm level) over a typical 60 days.
• Current rate: rising about 0.5 g per week.
• Prediction: alarm level reached in roughly 10 weeks.
• Recommendation: schedule maintenance within 6–8 weeks.

That arithmetic — fitting a slope and projecting to a limit — is exactly what an RUL estimator from a vibration trend automates, and a more formal treatment following ISO 13381 is available in the RUL prognostics calculator.

5. Applications

Maintenance Scheduling

• Plan an outage for when the RUL indicates optimal timing.
• Coordinate with production schedules.
• Group repairs to minimise total downtime.
• Avoid both premature and late interventions.

Parts Management

• Order spares with the right lead time.
• Avoid expediting costs.
• Reduce safety-stock requirements.
• Provision just-in-time, guided by the prognosis.

Resource Allocation

• Prioritise among several degrading machines.
• Direct limited resources to the most urgent needs.
• Plan workforce assignments and tool staging in advance.

6. Challenges and Limitations

Prediction Uncertainty

• Fault progression is never perfectly predictable.
• Operating conditions may shift without warning.
• Unexpected accelerations are always possible.
• Maintain safety margins as a matter of course.

Data Requirements

• Adequate trending history is needed.
• Early in a fault’s development, predictions are less certain.
• They improve as more data are collected.

Multiple Failure Modes

• One mode may be forecast while another causes the failure.
• Comprehensive, multi-parameter monitoring reduces the risk.
• All active degradation mechanisms must be considered together.

7. Improving Prognostic Accuracy

• Increase measurement frequency: more data points sharpen the trend, catch acceleration earlier, and cut uncertainty. Stepping up the periodic-monitoring interval on a suspect machine is often the single most effective move.
• Use multiple parameters: combine vibration with temperature and oil analysis; corroborating indicators raise confidence, and different parameters carry different lead times.
• Update continuously: revise the prognosis with each new measurement rather than trusting a single early prediction, and adapt to the actual progression rate.

In practice the quality of a prognosis is only as good as the data feeding it, so the measurement step matters as much as the maths. A portable two-channel instrument such as the Balanset-1A lets a technician capture repeatable spectra and amplitude readings at each route visit — the consistent trend points from which a credible remaining-life estimate is built. Prognosis is, ultimately, the element that separates true predictive maintenance from mere condition monitoring: by forecasting failure timelines from trend data and an understanding of fault progression, it enables the optimised timing that maximises equipment utilisation while protecting reliability.

---

← Back to Main Index