For a long time, analysis solutions for diagnostic purposes in industrial operations required data experts who needed weeks, months or even years to understand and model the system. Rockwell is now pooling this expertise in Project Sherlock, a new module with artificial intelligence (AI).

The data-based analysis algorithm is provided in a module that is integrated directly into the control housing. Once implemented, the AI "learns" the application managed by the controller using physics-based modeling. To determine this, the solution queries control variables. In addition, users can select inputs and outputs themselves via add-on commands to determine what should be modeled. Project Sherlock's AI quickly learns about the data stream running through the controller and creates a model. The entire process can be implemented within minutes. No extensive historical data is required, nor is it necessary to extract the data from the automation level.

Continuous monitoring of the model

Once the model has been created, Project Sherlock continuously monitors the operation for deviations from the basic findings obtained so far. If a problem arises, the alarm is automatically forwarded to an HMI screen or KPI dashboard via a trigger. In future, the module will not only create diagnoses, but also suggest solutions or automatically adjust system parameters to solve the problem without operator intervention.

"Project Sherlock is an easy-to-deploy component that enables comprehensive intelligent analytics for manufacturing companies," explains Ashkan Ashouriha, Solution Architect Integrated Architecture & Connected at Rockwell Automation. "As part of digital transformation, our customers are increasingly using production data to improve their business outcomes. They can't wait for experts to analyze it. Even if there were enough 'data scientists' for the industrial manufacturing sector, not every company has the time or budget to employ one. The AI uses our specially developed machine learning algorithm and creates meaningful analyses based on the automation infrastructure."

The diagnostics generated by Project Sherlock reduce potential false alarms compared to other AI solutions because they are based on physics-based modeling and industrial applications. For example, Project Sherlock AI can recognize whether the temperature fluctuations in a boiler are due to the flow fluctuating within the permissible range or whether it is a deviation for which corrective action needs to be taken.

Templates for process or hybrid applications

The initial version of the AI has ready-to-use templates for boilers, pumps and coolers that are suitable for process or hybrid applications. Manufacturing companies can use the configuration guide to develop further applications. Users can choose to what extent and at what intervals the PLC CPU should receive the data from the module , with the CPU setting the trigger in each case. The CPU load of the controller and the network load are not significantly affected by the module. Pilot projects of the AI have already been in successful use for 18 months. The module will be available from mid-2018. as