One of the areas where Information Technology (IT) and Operational Technology (OT) teams are more closely collaborating is in machine learning applications.
At the recent ARC Industry Forum conference in Orlando, Blair Fraser, with Emerson Impact Partner Lakeside Controls, presented in a session, IT/OT convergence, Machine Learning and IIoT enables predictive performance monitoring of Fermentation Process.
The challenge this application was to solve was to improve batch performance and reliability in a fermentation process. It was a challenging application since the process response was non-linear and variable. Additionally, the application was for an industry requiring regulatory validation and the solution needed to be part of a validated system.
When working to identify a solution for this challenge, the first step was to start with a failure mode and effects analysis (FMEA). Blair and the project team identified 112 failure modes in which predictive maintenance was identified as a control type. Using the SAP enterprise resource planning system, the team sifted through the data to identify common failure types. They also interviewed the operations and quality staff members to identify common issues.
Machine learning (ML) technology was applied even before data was streamed from the batch process. Prior to this application, the operations staff was able to accurately identify 3 phases of fermentation, and only define them after the batch had completed—all with an accuracy of 92% for “good” batches. After apply ML, they were able to identify 6 phases with 98% accuracy for all batch quality. And, they were able to predict these batch phases while the batch was in progress.
After conducting the FMEA, Emerson IIoT sensors were added and the existing data was reviewed to see if it could detect degradation of the specific failure mode.
The simplified version of IT/OT convergence is the incorporation of data from facilities management systems (FMS) and wireless IIoT sensors into a Quartic IIoT Edge Gateway. The information is processed in the Quartic enterprise FOG node and sent to Plantweb Optics and Quartic visualization software where the data has been transformed in to information ready for decisions and action.
The vision for this solution was to enhance real-time batch information, provide a batch real-time performance index and component health index, identify real-time variables predicting poor batch performance and component health degradation, predict batch quality, and provide key data real-time trends.
The benefits realized from this approach included automatic batch classification and comparison, multivariate process tracking, real-time batch troubleshooting, quicker response to deviations, comparisons against “golden batch” performance, and faster yield prediction—without having to wait for lab results.
Learn more about ways to improve overall operational performance in the Operational Certainty section on Emerson.com. You can also connect and interact with other IIoT and digital transformation experts in the IIoT & Digital Transformation group in the Emerson Exchange 365 community.