480037 The Canopy Platform for Process Monitoring and Optimization

Wednesday, March 29, 2017: 3:30 PM
211 (Henry B. Gonzalez Convention Center)
Eric Jones and Tim Diller, Enthought, Austin, TX

Enthought provides product and consulting solutions for the chemical, petrochemical, pharmaceutical, and semiconductor industries. At the core of these solutions is the Canopy Platform, which brings together a host of powerful components for data analysis, visualization, local and remote data access, and data exploration. Canopy Platform solutions allow engineers to easily browse and visualize huge amounts of data stored remotely, do interactive MATLAB-style analysis on it, and develop analysis or detection algorithms and deploy them at scale over large datasets using cluster or cloud compute resources. Detection and prediction algorithms developed in this fashion can be applied to monitor live incoming data and provide feedback to operators, used to provide input to maintenance scheduling, or used to debug production issues. These solutions place a strong emphasis on both convenient desktop user interface (useful for experts to develop their algorithms) and web dashboards for monitoring or data consumption by a broader audience. Web service components makes detection or prediction algorithms written on the platform accessible programmatically, so that the development effort can be reused by other systems and software throughout the business.

In this talk, we show how we leverage the Canopy Platform to build modern, customized software for analyzing plant process systems. We'll demonstrate the ad hoc analysis capabilities and workflow-oriented user interface that simultaneously simplifies open-ended and routine tasks. These capabilities make use of the extensive library of analysis routines that are available to the Canopy Platform and the data management component that enables uniform access to raw files, databases, and remote data stores. We'll walk through a cluster analysis workflow that can make use of a company's existing compute resources to scale to thousands of controllers and years worth of batch or continuous data. We will also describe the larger vision for this software as an automated plant advisor that monitors the status of processes, warns when parameters are out of expected bounds, forecasts failures, and offers solutions to those problems.

Extended Abstract: File Not Uploaded