Rela AIRela AI Docs
Condition Monitoring

Trend Analysis

Trend analysis transforms sensor readings into useful information: whether temperature is rising, how fast it is doing so, and when it could become a problem.

Trend Analysis

A sensor reporting 72°C right now does not say much on its own. But if you know it was at 65°C a week ago, 68°C three days ago, and 71°C yesterday, then you have a clear trend: the equipment is progressively heating up. Trend analysis does exactly that — it converts a series of measurements into an understandable pattern that allows anticipating problems before they occur.

What is it for?

Fixed-limit alarms only detect problems when a value has already crossed a threshold. Trend analysis detects that the value is approaching the problem, even if it has not arrived yet. This allows acting days or weeks in advance.

Trend analysis allows you to:

  • Detect whether a variable is rising or falling in a sustained way
  • Calculate how quickly it is changing (rate of change per hour)
  • Identify point anomalies that deviate from the normal pattern
  • Smooth electrical noise and normal variations to see the real trend
  • Feed the calculation of the asset's Remaining Useful Life (RUL)

How does it work?

The system takes all sensor measurements and processes them in time windows — groups of data from the last hour, last 6 hours, last day, or last week. For each window it calculates statistics (minimum, maximum, average) and then fits a mathematical trend line to determine whether values are rising, falling, or stable.

It also calculates two types of moving averages that smooth noise and highlight the real trend:

Average typeBehaviorWhen it is most useful
SMA (simple moving average)Treats all measurements equallyWhen you want to see the general trend without recent peaks distorting it
EMA (exponential moving average)Gives more weight to recent measurementsWhen you want to detect rapid changes as soon as they occur

It also defines operating bands for each sensor:

BandMeaning
NormalExpected range for everyday operation
Upper / lower limitAlert zone — equipment can operate here but deserves attention
Critical upper / lowerImmediate intervention zone

When a measurement crosses a band, the system generates a violation that is added to the anomaly analysis.

How to use it?

  1. Go to the asset profile and open the Condition Monitoring section.
  2. Select the event source with the sensor you want to analyze.
  3. In the trends view you will see:
    • Line chart with measurement history
    • Operating bands marked on the chart
    • Trend line (regression) overlaid
    • Moving averages as optional dotted lines
    • Anomalies marked in red

Interpret the trend

The system classifies the direction of each trend:

ClassificationWhat it indicatesRecommended action
Strong upward trendVariable rising consistently and quicklyInvestigate — generally problematic (temperature, vibration)
Moderate upward trendVariable rising slowlyMonitor more frequently
StableVariable within normal variationNo action required
Moderate downward trendVariable dropping slowlyMay be normal (cooling) or problematic (pressure dropping)
Strong downward trendVariable dropping consistently and quicklyInvestigate — may indicate loss of capacity

A trend is "strong" when the data fits well to a straight line. If temperature rises 1°C every day with high consistency, the trend is reliable. If values jump up and down randomly, the calculated trend has no predictive value.

WindowBest use
1 hourDetect sudden changes during the shift — temperature that rose abruptly
6 hoursReview behavior during the full shift
24 hoursSee if there is a difference between day and night shift
7 daysEvaluate gradual deterioration week by week

Configure operating bands

Bands are configured per sensor in the event source. To configure them:

  1. Open the asset's event source.
  2. For each metric, define the four limits: low normal, high normal, low critical, high critical.
  3. Save. The system will start detecting band violations automatically.

Example for an air compressor: Normal temperature between 50°C and 75°C. Critical upper limit at 90°C. When temperature reaches 76°C no critical alarm is generated, but a band violation is recorded that the system takes into account when calculating anomalies.

Key benefits

  • Deterioration detection before values reach critical alarm levels
  • Rate of change per hour to know whether the situation is worsening quickly or slowly
  • Operating bands per sensor without needing to program manual alarms
  • Two types of moving average for different sensitivity needs
  • Trend data feeding directly into the remaining useful life calculation
  • Visual chart with all analysis layers overlaid for quick diagnosis

Common use cases

Scenario 1: Detect progressive overheating The night shift ends without incident. Reviewing the trend analysis for the past week, the maintenance engineer notices that compressor C-01's temperature has risen 2°C per day for 5 consecutive days: from 63°C to 73°C. The upward trend has high consistency. Although there is still no alarm, the projection indicates that in 8 days it will reach the 90°C critical limit. A heat exchanger cleaning is scheduled for the weekend.

Scenario 2: Compare behavior between shifts The production manager notices that motor M-08 has more failures on the night shift. They open the trend analysis with a 24-hour window and compare temperature by hour of day. The chart shows that between 11 PM and 3 AM average temperature is 8°C higher than the rest of the day. Investigation reveals that the area's cooling fan is turned off on that shift to reduce noise. The operating procedure is changed.

Scenario 3: Validate that a repair resolved the problem Before maintenance, pump B-04's vibration showed a strong upward trend: 0.3 mm/s per week for 6 weeks. After the bearing replacement, the trend in the following week shows complete stability: vibration dropped to 2.1 mm/s and the trend line is horizontal. Trend analysis confirms the repair was successful.

On this page