FactoryTalk Analytics GuardianAI Overview

FactoryTalk® Analytics™ GuardianAI™ is a machine learning based supervisory application that uses existing plant devices, such as variable-frequency drives, as sensors to monitor the health of assets like pumps, fans, and blowers on a plant floor. It uses device data to establish a baseline signature of each asset’s behavior under normal operating conditions. Then, it monitors the assets for any deviation from the baseline. Once a deviation is detected, a notification is sent to the user identifying the anomaly. If an anomaly is detected but cannot be identified, FactoryTalk Analytics GuardianAI notifies the maintenance engineer that an unidentified anomaly was detected. The engineer can then investigate the issue, determine the cause of the anomaly, and tag the deviation accordingly. The FactoryTalk Analytics GuardianAI machine learning engine then trains to identify new anomalies for future encounters. The following diagram illustrates this process and the variation between a known fault and an unknown deviation.

The FactoryTalk Analytics GuardianAI workflow takes a no-code approach to machine learning. As a result, a data scientist is not required to configure, deploy, or use this AI application. It is designed so that OT personnel, like maintenance engineers, controls engineers, machine operators, and plant managers, can work with FactoryTalk Analytics GuardianAI with minimal training required.

The configuration workflow consists of four steps. First, the user deploys the FactoryTalk Analytics GuardianAI application on a local VM or edge device. Then, the user adds the device, which acts as the sensor. Finally, the user inputs the identifying information about the asset being monitored (pump, fan, blower, or motor). After that, the user is ready to start training the model to establish the baseline.

FactoryTalk Analytics GuardianAI is providing premier integration with PowerFlex® 755, 755TL, 755TR, 755TM, 755TS, and 6000T drives to use as sensors to access three-phase current data for motor current signature analysis. It focuses on anomaly detection and identification for the following asset types: pumps, fans, and blowers. The application is designed to work with single-drive and motor applications. Given its adaptive nature, FactoryTalk Analytics GuardianAI can learn process-centric issues and adapt to asset types beyond those listed above. For this use case, the application comes equipped with generic motor control analytics.

Beyond the classification provided by the maintenance engineers, FactoryTalk Analytics GuardianAI comes equipped with embedded expertise to detect specific anomaly patterns out of the box, as outlined below.