Edge Analytics with AXON Predict

AXON Predict is an Industrial IoT edge analytics platform for anomaly detection and predictive maintenance. Predict enables you to learn what is happening with your devices now, and predict what will happen in the future.

AXON Predict is comprised of two components

Predict Atom

An extremely compact analytics engine that runs on the device or gateway. It has built-in pattern and anomaly detectors and real-time analytics for vibration, location, and machine performance.

Predict Hub

A set of docker services that provide the command, control, and coordination of aggregated analytics. Predict Hub can run on all major Cloud providers and on-premise.

Predict Atom

Provides simple SDKs for Java, Python, and C++ which let developers easily stream sensor data from the device into Atom in only a few lines of code. Atom performs real-time analytics and detects patterns of abnormal behavior; subsequently, Atom can take immediate corrective action and/or notify external users or systems.

Then, rather than send all device data to the Predict Hub on a Cloud, Atom sends only the information needed to give the user insight into connected device activity and analytical insights captured in real-time.

Predict Hub

Processes real-time device analytics from the Predict Atom and performs second stage, or cross-device analytics blended with historical information to compare behaviors across many devices. Analytics and patterns can uncover unexpected anomalous events or fraud behaviors or changes in performance (both degrading or improving), utilization, and quality.

Prediction AI modeling (based on the TensorFlow framework) enables AXON Predict to “learn” the data and process patterns for your devices. Models are built and trained by ingesting valuable historical data that includes the specified anomalies or values you want to detect. The resulting deep learning models can predict anomalies before they happen.

AXON Predict’s built-in real-time vibration analytics (based on Fast Fourier Transform (FFT)) identifies vibration anomalies over time. Issues such as imbalance, misalignment, and component wear can affect device performance and ultimately cause device failure. The ability to automatically detect vibration outliers can provide important insight into your devices and processes.

Combining vibration analytics with deep learning, AXON Predict will predict when erratic and outlier vibrations will occur and notify and apply corrective actions, ensuring near zero interruptions.

AXON Predict’s built-in location analytics lets you to create patterns and analytics based on things that move. These can be small devices with GPS or vehicles such as connected cars and delivery vans. Know where your devices are at all times, get insight into where they’ve been, learn where they’re going, how far off course they are in their travels, and more. Patterns based on location analytics can run configured actions at the device at any point in its travels and correlate with other sensor data such as temperature and vibration.

AXON Predict provides “AI at the edge” by running the trained deep learning models right at the devices, for instant anomaly detection in future (predicted) data. The ability to correct or neutralize detected or predicted issues immediately helps keep devices running smoothly and performing optimally.

Processes real-time device analytics from the Predict Atom and performs second stage, or cross-device analytics blended with historical information to compare behaviors across many devices. Analytics and patterns can uncover unexpected anomalous events or fraud behaviors or changes in performance (both degrading or improving), utilization, and quality.

Ready?

Now, try it out for yourself!​