Explore a comprehensive analysis of industrial process monitoring and optimization in this BSidesLV 2014 conference talk. Delve into the complexities of data collection, control loops, and temporal alignment challenges in industrial settings. Learn about machine learning applications for tracing variability sources and behavior-based event detection systems. Discover innovative approaches to feature selection, cluster tuning, and threshold optimization. Gain insights into operator response characterization and state of health assessment techniques. Understand how these advanced methodologies can enhance industrial facility operations and security.
Overview
Syllabus
Intro
Bottom Line Up Front
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: Establishing Truth is Difficult!
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
Characterizing Operator Response
Conclusion
