Rockwell Automation’s Project Sherlock AI Module Models, Monitors Industrial Operations
Rockwell Automation unveils the new Project Sherlock artificial intelligence (AI) module. The data-driven analytics algorithm is delivered inside a module that fits directly into the controller chassis. Once installed, Project Sherlock AI leverages novel physics-based modeling to “learn” the application that controller manages. The solution scours controller tags to identify the application or allows users to choose what they would like modeled by selecting inputs and outputs via an add-on-instruction (AOI). Project Sherlock AI will then quickly learn from the stream of data passing through the controller to build a model, according to the company. The process can reportedly be accomplished in a matter of minutes. Vast quantities of historical data are not required, nor must the data ever leave the automation layer.
Once the model is built, the Project Sherlock solution continuously watches the operation looking for anomalies against its derived, principled understanding. If it spots a problem, it can trigger an alarm on an HMI screen or dashboard. Future iterations will go beyond diagnostics to direct users on how to remedy the issue or to automatically adjust system parameters to fix the problem without human intervention.
Project Sherlock diagnostics offer drastically reduced false-positive alarms as compared to other artificial intelligence solutions due to its physics-based modeling and foundation in industrial applications, according to Rockwell. For example, Project Sherlock AI can tell if a boiler temperature shift is related to a benign change in upstream operations or an abnormality that requires correction.
The initial version of Project Sherlock AI will include ready-to-use templates for boiler, pump and chiller operations, suitable for process or hybrid applications. Users can model additional applications with guided configuration. Communications with the module are prioritized by the controller, so users can select how much data is sent and intervals of communication. The module does not add to controller CPU-load nor add to network traffic. Customers will be able to purchase the module in mid-2018.