Share with a co-worker
All Resources
Use case
The Challenge
A leading energy company operates a cloud-based predictive maintenance solution to identify potential failures in critical equipment, such as valves and gas turbine air filters. Operational data is uploaded daily from the plant historian, analysed by the predictive maintenance platform, and the results are returned to the historian. When a potential issue is identified, an alert is automatically generated for the maintenance support team, who validate the finding and investigate the underlying condition.
A key gap was the startup context. Generator startups are high stress operating phases, and their behavior depends strongly on how long the unit has been offline. Without automatically identifying and classifying startups, engineers could not consistently compare events, quantify differences, or link alerts to the correct operating context.
- Startup periods are critical and stress intensive
- Behavior changes depending on downtime duration
- No consistent way to detect and label startups automatically
- Hard to compare startups and validate alerts at scale
The Approach
The team built a custom calculation in TrendMiner that automatically detects each startup window and assigns a standardized startup type, enabling consistent analysis and reporting.
- Startup window detection, a startup is defined as the period between power off, less than 1 MW, and steady production, 360 MW plus for at least 1 hour
- Startup type classification, the calculation assigns a discrete tag, 1 hot, 2 warm, 3 cold, 0 no startup
- Value based searches, two searches detect power off and steady production periods, the time between them is labeled as the startup
- Downtime based categorization, the startup type is determined by how long the unit was off before the startup, under 24 hours hot, 24 to 96 hours warm, over 96 hours cold
- Event analytics readiness, once tagged, startups can be searched, filtered, and compared using event based statistics and visual analytics
Key Insight
When startup context becomes a tag, every alert and every event can be interpreted against the correct operating regime, turning isolated signals into actionable reliability insight.
The Results
The Takeaway
By standardizing generator startups into a single searchable classification tag, the team enabled faster and more consistent failure analysis, improved validation of predictive maintenance alerts, and created a scalable foundation for comparing startup behavior, reducing equipment stress, and strengthening reliability across power generation assets.
The Challenge
A leading energy company operates a cloud-based predictive maintenance solution to identify potential failures in critical equipment, such as valves and gas turbine air filters. Operational data is uploaded daily from the plant historian, analysed by the predictive maintenance platform, and the results are returned to the historian. When a potential issue is identified, an alert is automatically generated for the maintenance support team, who validate the finding and investigate the underlying condition.
A key gap was the startup context. Generator startups are high stress operating phases, and their behavior depends strongly on how long the unit has been offline. Without automatically identifying and classifying startups, engineers could not consistently compare events, quantify differences, or link alerts to the correct operating context.
- Startup periods are critical and stress intensive
- Behavior changes depending on downtime duration
- No consistent way to detect and label startups automatically
- Hard to compare startups and validate alerts at scale
The Approach
The team built a custom calculation in TrendMiner that automatically detects each startup window and assigns a standardized startup type, enabling consistent analysis and reporting.
- Startup window detection, a startup is defined as the period between power off, less than 1 MW, and steady production, 360 MW plus for at least 1 hour
- Startup type classification, the calculation assigns a discrete tag, 1 hot, 2 warm, 3 cold, 0 no startup
- Value based searches, two searches detect power off and steady production periods, the time between them is labeled as the startup
- Downtime based categorization, the startup type is determined by how long the unit was off before the startup, under 24 hours hot, 24 to 96 hours warm, over 96 hours cold
- Event analytics readiness, once tagged, startups can be searched, filtered, and compared using event based statistics and visual analytics
Key Insight
When startup context becomes a tag, every alert and every event can be interpreted against the correct operating regime, turning isolated signals into actionable reliability insight.
The Results
The Takeaway
By standardizing generator startups into a single searchable classification tag, the team enabled faster and more consistent failure analysis, improved validation of predictive maintenance alerts, and created a scalable foundation for comparing startup behavior, reducing equipment stress, and strengthening reliability across power generation assets.
Access now
Share with a co-worker
The Challenge
A leading energy company operates a cloud-based predictive maintenance solution to identify potential failures in critical equipment, such as valves and gas turbine air filters. Operational data is uploaded daily from the plant historian, analysed by the predictive maintenance platform, and the results are returned to the historian. When a potential issue is identified, an alert is automatically generated for the maintenance support team, who validate the finding and investigate the underlying condition.
A key gap was the startup context. Generator startups are high stress operating phases, and their behavior depends strongly on how long the unit has been offline. Without automatically identifying and classifying startups, engineers could not consistently compare events, quantify differences, or link alerts to the correct operating context.
- Startup periods are critical and stress intensive
- Behavior changes depending on downtime duration
- No consistent way to detect and label startups automatically
- Hard to compare startups and validate alerts at scale
The Approach
The team built a custom calculation in TrendMiner that automatically detects each startup window and assigns a standardized startup type, enabling consistent analysis and reporting.
- Startup window detection, a startup is defined as the period between power off, less than 1 MW, and steady production, 360 MW plus for at least 1 hour
- Startup type classification, the calculation assigns a discrete tag, 1 hot, 2 warm, 3 cold, 0 no startup
- Value based searches, two searches detect power off and steady production periods, the time between them is labeled as the startup
- Downtime based categorization, the startup type is determined by how long the unit was off before the startup, under 24 hours hot, 24 to 96 hours warm, over 96 hours cold
- Event analytics readiness, once tagged, startups can be searched, filtered, and compared using event based statistics and visual analytics
Key Insight
When startup context becomes a tag, every alert and every event can be interpreted against the correct operating regime, turning isolated signals into actionable reliability insight.
The Results
The Takeaway
By standardizing generator startups into a single searchable classification tag, the team enabled faster and more consistent failure analysis, improved validation of predictive maintenance alerts, and created a scalable foundation for comparing startup behavior, reducing equipment stress, and strengthening reliability across power generation assets.
Access now
Subscribe to our newsletter
Stay up to date with our latest news and updates.
Webinars on Demand
Press Play on Operational Improvement
.jpg)
.jpg)
.jpg)
Other Resources
Explore Our Newest Content to Maximize Your Operational Efficiency
