Rela AIRela AI Docs
Condition Monitoring

Prognostics and Recommendations

The system estimates how many days the equipment has before needing maintenance and generates specific recommendations based on health history, trends, and recent alarms.

Prognostics and Recommendations

Knowing that a compressor has an AHI of 62 today is useful. But knowing that same compressor has been losing 2 AHI points per week for 3 weeks — and that at this rate it will reach the critical level in 21 days — is what allows you to plan maintenance before the failure. Prognostics make that calculation automatically. Recommendations turn that calculation into a concrete action.

What is it for?

In reactive maintenance, equipment is repaired after it fails. In preventive maintenance, work is done on fixed dates. In condition-based maintenance, intervention happens when the data says it is necessary — not before, not after.

Prognostics enable condition-based maintenance by:

  • Estimating when the asset will reach a critical state if no action is taken
  • Classifying current risk (low, medium, high, critical) by remaining time
  • Automatically creating an urgent task when risk reaches critical
  • Generating specific AI recommendations explaining what to do and why
  • Triggering alerts when an individual metric exceeds 150% of its normal maximum

How does it work?

The system analyzes the history of the Asset Health Index (AHI) from the past weeks or months and calculates how many points it loses per day on average. Using that degradation rate and the current AHI, it calculates how many days remain before reaching the critical threshold of 20 points.

Example: Compressor C-03 has an AHI of 65 today. It has lost 0.5 points per day on average over the last month. There are 45 points remaining before reaching the critical threshold of 20. At that rate, in 90 days the equipment will be in a critical state.

The accuracy of the prognosis improves over time: with 30 days of history an estimate can be made with medium confidence; with 90 days or more, the estimate is reliable.

Risk levels by remaining time

Risk levelEstimated time to critical levelRecommended action
LowMore than 90 daysStandard monitoring
Medium30 to 90 daysPlan preventive maintenance
High7 to 30 daysSchedule intervention with priority
CriticalLess than 7 daysUrgent intervention — the system creates a task automatically

Degradation rates and their meaning

Degradation rateMeaningSignal of
Less than 0.1 pts/dayStable assetNormal operation
0.1 to 0.5 pts/daySlow degradationAccelerated normal wear
0.5 to 1.0 pts/dayModerate degradationProblem that needs attention
More than 1.0 pt/dayRapid degradationImminent failure, act today

How to use it?

View an asset's prognosis

In the asset profile, the Health section shows:

  • The current AHI with its grade
  • The current degradation rate (points per day)
  • The estimated date of reaching the critical level
  • The risk level with its color
  • The confidence level of the estimate (high, medium, low)

Interpret prognosis confidence

Confidence levelWhat it meansHow to use it
HighMore than 90 days of history with a clear trendMake decisions directly based on this prognosis
Medium30 to 90 days of historyUse as a guide, validate with visual inspection
LowLess than 30 days of historyIndicative only — the equipment needs more time to establish its pattern

When the prognosis has low confidence, the AI recommendation states this explicitly and suggests waiting for more data before making major decisions. For newly installed equipment or equipment with a recently relearned baseline, confidence always starts low and improves week by week.

Individual metric alerts

The system also monitors each sensor individually. If a metric exceeds 150% of the maximum value recorded in its normal baseline, a condition-based maintenance (CBM) alert is triggered, regardless of the overall AHI.

Example: Normal vibration for compressor C-01 has a historical maximum of 4.5 mm/s. If vibration reaches 6.8 mm/s (more than 150% of 4.5), the CBM alert activates even if the overall AHI is at Grade B. This captures localized deterioration in a specific component that the overall AHI does not yet reflect.

View and manage AI recommendations

The system generates recommendations in plain language explaining what was found and what to do. Each recommendation has a lifecycle:

StatusMeaning
OpenNew, pending review by the technician or supervisor
AcknowledgedThe responsible party has read and noted it
In progressThe recommended action is being executed
ResolvedThe action was completed and the issue was addressed
DismissedThe technician decided not to follow it (with recorded justification)

To review recommendations:

  1. Go to the asset profile and open the Recommendations section.
  2. You will see active recommendations ordered by priority.
  3. Click each one to see the full detail with the AI's justification.
  4. Change the status as you progress.

Key benefits

  • Estimated failure date in days — not "the equipment is deteriorating" but "in 23 days it will reach critical state"
  • Automatic urgent task creation when risk reaches critical — no one needs to remember
  • Plain-language recommendations that explain the problem and suggest concrete actions
  • CBM alerts per individual sensor that capture localized problems before they affect the AHI
  • Explicit confidence level so the technician knows how much to trust each estimate
  • Recommendation lifecycle to manage and document follow-up

Common use cases

Scenario 1: Plan a scheduled shutdown based on data The plant plans a 48-hour maintenance shutdown in 6 weeks. The reliability engineer reviews the prognoses of all critical equipment. Compressor C-02 has medium risk with 45 estimated days to critical level — fitting right within the scheduled shutdown. Turbine T-01 has low risk with 120 days. Motor M-15 has high risk with 18 days — too urgent to wait for the shutdown. A preventive intervention on M-15 is scheduled for the following week without waiting.

Scenario 2: Automatic task prevents unexpected failure During a long holiday weekend, the AHI of compressor C-04 drops rapidly due to accumulated alarms and rising temperature trends. Monday morning the system detects that the degradation rate exceeds 1.2 points/day and risk reaches critical level. It automatically creates an urgent task: "Critical RUL — Compressor C-04: accelerated deterioration detected, intervention required in less than 7 days." The coordinator assigns it to the shift technician before the equipment fails.

Scenario 3: Specific recommendation based on the most deteriorated sensor The AI analyzes pump B-07 and generates the recommendation: "The condition sub-index is at 28/100 due to vibration at 2.4 sigma above baseline for 9 consecutive days. Temperature remains normal, which rules out overheating. Pattern consistent with bearing wear. Recommendation: bearing inspection and shaft alignment check. Confidence: high (87 days of history)." The technician executes exactly that inspection and finds the rear bearing damaged.

On this page