Persistent thermal monitoring generates a continuous stream of temperature data from every point in the field of view of every installed camera. For a facility with multiple cameras monitoring dozens of pieces of electrical equipment, this data stream can involve millions of temperature measurements per hour.
Raw thermal data at this volume is not useful in itself. The challenge is extracting signal from noise: distinguishing the normal thermal variation that accompanies changing load conditions, ambient temperature shifts, and equipment cycling from the genuine anomalous temperature patterns that indicate developing electrical faults.
This is the problem that the Sigma Delta Tau algorithm was developed to solve.
Why Simple Threshold Alerts Fail
The most intuitive approach to thermal monitoring alerts is a temperature threshold: if a monitored point exceeds a defined temperature, trigger an alert. This approach has the appeal of simplicity, and it is how many basic thermal monitoring systems operate.
It also generates enormous numbers of false positives in real-world operation.
Electrical equipment temperature varies significantly with load. A transformer running at peak load during a summer heat wave will be substantially hotter than the same transformer at light load in the winter. Bus connections, switchgear, and power cables all exhibit load-correlated thermal variation as a normal part of operation. A temperature threshold calibrated to avoid false positives at peak load will miss genuine anomalies at moderate loads. A threshold calibrated to catch moderate-load anomalies will generate constant false alarms at peak load.
The practical result of a high false-positive rate is alert fatigue: maintenance teams learn to ignore the alerts because most of them are not actionable, and genuine fault conditions get missed in the noise.
The Three Components of SDT
The Sigma Delta Tau algorithm addresses this problem by analyzing the characteristics of temperature change over time rather than monitoring absolute temperature values. The name encodes the three mathematical components of the analysis:
Sigma (Σ): Standard Deviation of the Temperature Signal
Sigma measures the statistical variability of the temperature signal over a defined time window. Normal electrical equipment operating under varying but consistent load conditions produces temperature variation with characteristic statistical properties. Developing fault conditions change the statistical character of the temperature signal in detectable ways.
A developing loose connection, for example, produces intermittent micro-arcing that appears as high-frequency temperature variation superimposed on the load-correlated baseline. The sigma component of the SDT analysis can detect this increased variability as an anomaly signature even when the absolute temperature values remain within normal range.
Delta (Δ): Magnitude of Temperature Change
Delta measures the absolute and relative magnitude of temperature changes over defined time windows. This component captures the rate at which a developing condition is progressing and the severity of temperature anomalies relative to comparable equipment or reference points.
Delta analysis is particularly valuable for distinguishing load-correlated temperature changes from fault-driven ones. A temperature increase that tracks proportionally with load increase is likely normal operation. A temperature increase that is disproportionate to the load change, or that occurs without a corresponding load change, is anomalous.
Tau (τ): Rate of Change Over Time
Tau measures the time-derivative of the temperature signal: how fast temperature is changing. This component is particularly valuable for early detection of fault conditions that are accelerating.
A developing fault that has been progressing slowly for months may produce only modest absolute temperatures. But as the fault approaches a threshold condition, the rate of temperature increase accelerates. The tau component of SDT detects this acceleration as an early warning of approaching criticality, providing actionable lead time even when absolute temperatures have not yet reached alarming levels.
"The insight behind SDT is that faults have signatures, not just temperatures. The rate of change, the variability, and the character of how a temperature anomaly develops over time contain far more diagnostic information than the absolute temperature value at any single point."
How SDT Works in Practice
SDT analysis operates on continuous radiometric temperature data streams from all monitored points. The algorithm establishes baselines for each monitored point by observing normal thermal behavior across varying operating conditions during an initial learning period.
Once baselines are established, the algorithm continuously analyzes incoming data for deviations from baseline behavior across the three components: sigma, delta, and tau. Anomaly detection is weighted based on the consistency and severity of deviations across multiple components. A single brief temperature spike may not trigger an alert; a persistent change in sigma combined with an increasing tau value and an anomalous delta reading will.
The multi-component nature of the analysis is what enables SDT to distinguish genuine developing faults from the many sources of thermal noise in a real operating environment. Ambient temperature changes affect all equipment similarly. Load changes produce correlated thermal responses. Equipment that is genuinely developing a fault condition shows a distinctive pattern that is different in character from these background variations.
The Early Detection Advantage
The practical benefit of SDT-based analysis over threshold monitoring is substantial lead time. Industry data on electrical fault development consistently shows that the period from initial thermal anomaly to failure can span months. The earlier a developing fault is detected in this period, the more options and lead time exist for corrective action.
SDT detection typically identifies developing conditions at an early stage of their thermal signature, often before they would be visible to a periodic thermographic inspection and well before they would trigger even a well-calibrated absolute temperature threshold. This early detection translates directly into maintenance planning flexibility: the ability to schedule corrective work during planned outage windows rather than responding to emergency failures.
DOD Origins and Field Validation
The Sigma Delta Tau algorithm was developed from foundational research originating in U.S. Department of Defense programs, where the operational cost of equipment failures in deployed military applications created compelling requirements for advanced predictive maintenance capabilities.
Power Intelligence has deployed and validated SDT-based monitoring across a range of operating environments, including electric utility substations, airport electrical infrastructure, sports and entertainment venues, and industrial facilities. The algorithm's performance across these diverse applications reflects its fundamental design: it analyzes the physics of fault development rather than the specifics of any particular equipment type or installation context.
The result is a monitoring capability that brings genuinely advanced predictive analytics to the electrical infrastructure challenge, distinguishing developing fault conditions from normal operational variation with a reliability that simple threshold monitoring cannot match.