(19c) Live Column Hydraulics - How Does It Help? | AIChE

(19c) Live Column Hydraulics - How Does It Help?


Overall Manufacturing Asset Effectiveness is significantly affected by availability of an equipment and it’s ability to perform to level of operational expectations. In refinery process, the capability of the overall plant effectiveness is significantly governed by the effective utilization of individual column capacities while ensuring reliable operation i.e. maximize capacity while ensuring on-spec product with optimal product recovery. In addition to this, variable feedstock quality and product pricing adds to the complexity as more than often this governs the overall plant capacity and profitability.

Traditional approaches more than often lead to setting up of conservative limits to ensure reliable operation. With the advancement of big data and solutions focused on AI/ML techniques, the power of analytics can be effectively used to improve column capacity while ensuring reliable operation i.e. ensure on-spec production, maximize product recovery and avoid unplanned downtime à “finding problems before they find you”. Using the power of data analytics ensures that the advantages from process optimization and advanced controls are consistently realized.

More than often the symptomatic indicators reside in the large quantity of a variety of data that are measured on a continuous basis and archived in the plant historian. It would need a rigorous, and time-consuming exercise to review these wide variety of millions of data points - collected on a minute wise or even higher frequencies – to identify such symptoms and this is difficult for an unaided human mind, unless it shows up obviously on a commonly monitored parameter. Several of these parameters measured are non-linearly related to one another, further complicated by different types of errors in the collected data.

Big Data Analytical techniques can be utilized to effectively analyse hundreds of minute wise tag data for a process equipment, in real time. This paper describes how Machine Learning Techniques in combination with first principle models have been applied to Distillation Column operation in an Manufacturing unit to provide useful, non-intuitive insights that are usually hidden in the large volumes of data collected. This paper covers the case studies covering applied machine learning techniques for troubleshooting and improving Distillation Column performance. How Digital Twin model is deployed with real-time analytics options providing augmented intelligence to maintain operations at best possible

Case Study 1: Column – Live Hydraulics: A real time analytics solution was developed for providing live hydraulics, tray loadings at different sections of C2 Splitter Column.

  • Intuitive insights to address capacity constraints by addressing flooding areas with smart changes in operating parameters
  • Ensuring reliable operation, tracking PDI and other critical instrumentation

Case Study 2: Column – Real time Optimizer: A multivariate ML model based on live data provided real time guidance to plant operations to ensure on-spec production without limiting column capacity or overall performance.

These and other case studies would be presented to demonstrate how power of analytics can be used to offer real time decision insight to operations, driving actionable outcomes creating new levels of operational excellence.