444018 Integration of First-Principles Based Unit Modeling with Data Analytics Techniques

Tuesday, April 12, 2016: 8:30 AM
335B (Hilton Americas - Houston)
Apurva P. Samudra, Alexander B. Smith, Kadir Liano and Bijan Sayyar-Rodsari, Rockwell Automation, Austin, TX

This paper details the infrastructure built to integrate big-data technologies in cloud with first-principle models for distributed diagnostic, predictive, and prescriptive analytics. In many industrial applications, manufactures produce process components which are geographically distributed. Oil well pumps, electric motor drives, and heavy construction equipment are examples of such products. These units require advanced and robust analytics capabilities to maximize their efficient utilization and minimize costs associated with unplanned downtime and uncertainty. While embedded control solutions manage the safe operation of the asset, analytics technologies perform fault-diagnostics, operational predictions, and performance modeling for a battery of assets. We present smart-analytics technologies to model and monitor the equipment using cloud facilities such as high throughput, multi-source data ingestion, stream processing of data, and elastic web applications. These engines include but are not limited to:
  • A Proprietary model identification system which identifies the basic first-principles behaviors such as mass and energy balances, along with time-depended performance metrics such as fouling and physical performance degradation. The modeling engine is asset agnostic and provides an auto-configurable, continuously-updated unit model with high fidelity. The model retraining is automatic and is done based on the error measure produced by the engine.
  • A Data clustering and prediction engine which monitors the individual unit and its trajectory using PCA and LDA analysis. This information is then used to classify the condition of the unit and provide trajectory of these conditions to the end-user. The engine is completely data-driven and uses small set of pre-labeled data to generate the model of complex operation of unit.

These engines are deployed and provisioned as cloud services using web application frameworks. Thus, these are extremely scalable and can be accessed at very high frequency. The data is ingested by cloud services and presented to the end-user via Microsoft Power-BI dash boarding service. These front-ends include standard offerings from the power bi interface as well as custom built components from Rockwell.

Extended Abstract: File Not Uploaded
See more of this Session: Big Data Analytics and Fundamental Modeling I
See more of this Group/Topical: Topical A: 2nd Big Data Analytics