Explore the concept of ground truth in industrial processes through this 38-minute BSidesLV conference talk. Delve into the challenges of establishing accurate baselines in complex industrial environments, including temporal alignment issues and variability factors. Learn about machine learning techniques for tracing sources of variability and implementing behavior-based event detection systems. Discover the advantages of these advanced methods over traditional approaches. Gain insights into toolkit flow, feature selection, cluster tuning algorithms, and state of health assessments. Understand how to optimize thresholds and incorporate operator expertise into the process. Enhance your knowledge of industrial data analysis and process optimization techniques.
Overview
Syllabus
Intro
Motivation
Overview of Example Industrial Process
Resources from plant
Simplified Process Diagram (DCS)
Industrial Database
Understanding the Data
Data Collection Points (Control Loops)
Temporal Alignment Problem
Variability Challenges in Establishing Ground Truth
Machine Learning
Supervised learning for tracing sources of variability
Further Correlation Analysis
Behavior Based Event Detection System for Industrial Facilities
Advantage over Current State of the Art
Overview of Toolkit Flow
Feature Selection
The Cluster Tuning Algorithm
The K-Means Algorithm
Threshold Optimization
GUI FOR Cluster Tuning
Operator in the Loop
State of Health Assessment
Conclusion
Taught by
BSidesLV
