BPI SEMINAR BY EMILIE THIBAULT
October 21, 2022, 12:00 pm to 1:00 pm
TITLE
Data Processing and Usage: Added-Value, Barriers, and Practical Applications
DATE & TIME
October 21st Friday 12:00 PM - 1:00 PM PDT
LOCATION
CHBE Room 202, 2360 East Mall, Vancouver, BC V6T 1Z3
AGENDA
- Introduction by Dr. Paul Stuart
- Keynote Presentation by Émilie Thibault, Polytechnique Montréal
- Q&A / Discussion
*Sandwiches and coffee will be provided.
ABSTRACT
Today in the process industries, abundant data are typically collected and stored according to a standardized protocol, with the goal of using this information for decision-making. These data are “contaminated” due to various factors, resulting in error because measurements are made by unreliable or unpredictable instrumentation, and represent complex operations. The implementation of data processing (or signal processing) approaches enables the correction of measurement errors to improve data fidelity. For the context of decision-making, what is the “best” data processing framework?
To better appreciate the industrial context for this project, interviews regarding signal processing methods were undertaken with industry practitioners to explore (1) how various methods are used for different decision-making contexts, (2) what the added value of process data can be, (3) how these data are used, and (4) what the major barriers are that prevent operating facilities from using them. These interviews were conducted with mill experts as well as with software developers working in support of the pulp and paper industry. The results of this survey will be summarized. It was found that for a variety of reasons, that common signal processing approaches include Kalman Filtering, Exponential Weighted Moving Average, Short Time Fourier Transform and Wavelet Transform. Their robustness, and performance to remove noise were compared for the cleaning of contaminated signals. However there are thousands of signals stored from an industrial facility, and the tuning of filter parameters to process industrial signals is tedious and time-consuming. To expedite the signal processing procedure, there is a need to systematically, structurally, efficiently, and automatically identify or select the values of parameters. This may be accomplished by assessing the correlation between the optimal filter parameter values using a heuristical method, which will be presented.
ABOUT THE SPEAKER
Cycling enthusiast, advocate for sustainable development, and Chemical Engineering PhD student at Polytechnique Montréal in Paul Stuart’s Product and Process Design Laboratory - Émilie Thibault is passionate and motivated about leveraging data science for problem-solving and industrial decision-making. Émilie’s research includes developing an operating regimes-based data processing framework for process troubleshooting and decision-making and building an Activity-Based Cost (ABC) model for industrial facilities. Her research is strongly focusing on the industrial side as she is working closely with a pulp mill that kindly supplies their contaminated data for her PhD project. Besides that, Émilie is involved in her community, teaches courses at Polytechnique, promotes engineering programs at various events, and is the public relations advisor for the Arbour Foundation which offers scholarships to graduate students in fields that benefit the Quebec economy. Émilie also sits on the selection committee of the Foundation, is an administrator on two Boards, and sits on boards related to the place of women in engineering.
REGISTRATION FOR SEMINAR BY EMILIE THIBAULT
Please fill out the below form to register for the event.