Advanced Data Analytics for Utilities
The objectives of any Asset Monitoring & Diagnostics (M&D) program include minimizing the opportunity for catastrophic failure, maximizing machine and cycle efficiency, and eliminating time-based preventive maintenance tasks. Benefits of each of these three objectives rely on understanding the principles behind applications found in M&D programs as well as developing their own applications to meet specific M&D program needs.
Advanced Data Analytics for Utilities (DA201: Data Analytics - Advanced) will help you understand and execute advanced data modeling techniques beyond traditional Advanced Pattern Recognition with real-life examples of equipment failures and how each could have been detected using various statistical methods. We’ll examine rule-, model-, and case-based predictive analysis, focusing on when and how each should be used and the strengths and limitations of each. Next, students will learn alert management techniques and best practices, utilizing previous data to complete alert management tasks.
To help you retain and develop long-term recall of the course material, over 50% of class time is spent on hands-on exercises using visual association tools to break down complex topics. This course prepares you to identify and execute proper analytical techniques to help you develop these in-demand skills.
You Will Learn
-
Comparison of basic analytics to advanced analytics model results
-
Introduction to rule-based, model-based, and case-based diagnostics methods
-
Weighting method exercise for determining faults
-
Bayesian Belief Network (BBN) exercise for determining faults
-
Forecasting using Mean Time To Action (MTTA)
-
Introduction to Prognostic Methods
-
Statistical analysis of alert settings using box-and-whisker plots
-
Alert prioritization best practices
-
Complete alert philosophy, metrics, and continuous improvement strategies
-
Review and reinforce concepts from Data Analysis 101
DA201 Section Descriptions
DA201.1: Diagnostics and Prognostics Methods
Continue to explore multiple technologies and when each should and should not be used, including rule-based, model-based, and case-based technologies, neural networks, and pure database query, including the pros and cons of each. Learn forecasting and mean-time-to-action techniques when determining action limits, determining reliability, stressors, symptoms, and effects. Topics include rule-based analytics, case-based analysis, and Bayesian Belief Networks for more complex diagnostics. Exercises include weighting for simple diagnostics, Bayesian Belief Networks for more complex diagnostics, and Time-to-Action forecasting.
DA201.2: Advanced Data Analytics Modeling
Build upon the models that were developed in DA101 to cover additional operational modes (transients and steady-state) and use advanced capabilities of modern machine learning techniques. Additionally, simple models from previous exercises are compared to more advanced models to determine which perform optimally. Comparison and contrasts are discussed to learn trade offs of simple vs. advanced analytical models. Topics include steady-state and transient analysis, introduction to machine learning, and moving beyond Advanced Pattern Recognition. Exercises include regression modeling, Advanced Pattern Recognition modeling, and residual analysis for alert setting.
DA201.3: Alert Management Practices 1
Learn basic alert-setting philosophy and the importance of actionable alerts. Using box-and-whisker plots, learn how to set residual alerts in Advanced Pattern Recognition tools to maximize efficiency and minimize false or non-actionable alerts. Topics include actionable vs. non-actionable alerts, box-and-whisker plots, and proper persistence window adjustments. Exercises include continuations of proper alarm setting and alarm setting iterations.
DA201.4: Issue Management - Process and Communications
Learn the link between model management, alert management, and case management, all from frequent points of view, including the data analyst, unit operations, and a manager or director. Also, learn how to determine response actions when faced with limited data, how to interact with work management systems, and Monitoring & Diagnostics center operations. Topics include keeping the big picture in mind, how to manage you or your customer’s program, and how to continually improve. Exercises include properly reacting to critical-to-generation turbine bearing oil temperatures with limited data.
DA201.5: Alert Management Practices 2
Learn how to create a strategy for prioritizing alerts, choosing persistence windows for modeled components that drive actions. Also, learn how to develop a remediation strategy for previous or poor practices that led to numerous, persistent alerts with no value. Topics include prioritization, consistency, organization, and establishing a routine. Exercises include establishing a prioritization strategy and establishing a remediation strategy for numerous non-actionable alerts.
Who Should Attend
-
Data Analysts
-
Asset Monitoring & Diagnostics Program Managers
-
Directors or VPs of Data Analytics for manufacturing and utilities
-
Information Technology professionals
-
Systems Administrators who are responsible for M&D program implementation