Smart Platform combines and updates three Curtiss-Wright
advanced condition monitoring applications into a single,
flexible, anomaly detection solution. One element of the
platform is PdP (predictive pattern recognition), a similarity-based,
machine learning application presently deployed across multiple power
generation fleets. The second component is EAD (Equipment Anomaly Detection),
a neural-network based predictive engine that utilizes a long-short term memory
(LSTM) neural network implementation to include the time series data and multiple
digital states in diagnosing equipment behavior. The final element is StressWave,
a sensor-based, vibro-acoustic monitoring system and associated diagnostic tools
currently detecting precursor conditions prior to equipment damage on diverse assets
such as cruise ships, steel mills, wind turbines, military equipment and more.
A SmartPlatform server provides real time data collection, processing,
and storage of results, as well as supporting model maintenance and
storage of model configuration data. Real time alarms and diagnostic
tools are provided to end users by a web-based graphical user interface (GUI).
PdP and EAD model configuration and training are accomplished on a Windows based
client networked to the Smart Platform Server. Once configured, models run
independently on the server.
While any of the three applications can be applied independently,
leveraging their capabilities can provide superior equipment protection.
Each has its merits for providing early warning utilizing only traditional
condition monitoring sensors, which typically do not register changes until
conditions are primed to produce at least a minimal amount of equipment damage.
In contrast, StressWave monitoring identifies conditions before damage occurs.
Using the power of PdP and EAD to predict the expected range of StressWave indications
during normal operation permits the creation of a dynamic alarm band that provides earlier
warning than fixed alarm thresholds. This means a combined system that highlights
precursor conditions before any damage occurs.