UQ4Pro: A Demonstration Tool for Uncertainty Quantification in Process Industry Applications
Year: 2024
Type: conference-paper
Abstract: We demonstrate an Uncertainty Quantification tool based on conformal prediction for process industry applications.The tool generates uncertainty estimates as prediction-sets for classification tasks such as those used in fault classification, detection, and severity estimation.The conformal prediction framework for uncertainty estimates provides a finite-sample guarantee of accuracy and is model and data-distribution agnostic.The essence of using conformal prediction lies in the choice of a conformal score function.Our tool, UQ4Pro, incorporates multiple such score functions and provisions the user to select a score function based on their specific industrial process context.
Authors: Deepti Maduskar, Divyasheel Sharma, Shivang Modi
Institution: ABB (India)
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4411624628