Analysis of an Actual Distillation Process Stream by Stan Siranovich
Details
In this presentation, we analyze actual process data from a Canadian oil refinery. Our dataset consists of 253 snapshot measurements of 27 variables from a distillation column, measured over 2.5 years. For independent variables, we have 15 temperature columns, 4 flow columns, a pressure column, and 5 calculated columns. For the dependent variable, we have the Reid Vapor Pressure.
Several characteristics make this dataset interesting:
· Seasonality, as the refinery transitions from winter to summer “blends” and back again through 2.5 cycles.
· A reboiler, which takes the process stream after it leaves the column, increases the temperature, and returns it back to the column.
· Numerous outliers, mainly on the high end of the range.
Because in a commercial setting, business objectives should guide our analysis, we conduct this analysis both to predict an outcome variable and to understand the process. Finally, to better demonstrate analytical flow and the process of successive discovery, we conduct most of this session as a live demo.
Bio:
Stan spent most of his career with Bayer as a Senior Research Chemist and Technical Marketing Specialist. While there, he was introduced to JMP version 3. Prior to Bayer, he worked for Cargill and Mobil Oil doing polymer synthesis and product development.
He is currently a JMP Training Partner and performs contract work through Crucial Connection LLC. He is interested in working at the intersection of Business and the natural Sciences.