By the end of this course, participants will be able to:

• Understand core concepts of AI and Machine Learning and their relevance to refinery operations
• Identify key refinery challenges that can benefit from AI-based solutions
• Gain hands-on experience with ML modeling techniques for prediction, classification, and optimization
• Evaluate and compare different ML algorithms and deployment strategies
• Understand hybrid modeling approaches that combine first principles with data-driven models
• Explore case studies demonstrating AI applications in CDU, FCCU, blending, energy systems, and more
• Build a roadmap for deploying AI initiatives in their own organizations, considering ROI, change management, and system integration

This course is designed for professionals in the refining and downstream oil & gas sector who are interested in leveraging AI and ML to drive operational and business value. Ideal attendees include:

• Process Engineers seeking to enhance optimization and troubleshooting through data-driven methods
• Operations & Production Engineers aiming to increase reliability, reduce variability, and improve efficiency
• Digital Transformation & Innovation Leads responsible for adopting and scaling AI initiatives
• Data Scientists & Analysts working with refinery or industrial data
• Engineering Managers & Technical Consultants exploring new technologies to solve strategic refinery problems

Day 1: Foundations of AI & Machine Learning in the Refinery Context

Session 1: Introduction to Artificial Intelligence & Refinery 4.0

• What is AI? Definitions & evolution
• The digital transformation journey in oil & gas
• Overview of AI’s role in downstream operations
• Pros & cons of using AI in industrial environments
• Common misconceptions & pitfalls

Session 2: Machine Learning Theory & Concepts

• Supervised, unsupervised, and reinforcement learning
• Regression, classification, and clustering basics
• Overfitting, underfitting, bias-variance tradeoff
• Model evaluation: MAE, RMSE, ROC-AUC, confusion matrix

Session 3: Tools, Frameworks, and Industrial Readiness

• Tools: Python, Scikit-learn, TensorFlow, PyTorch
• Industrial-grade platforms: Aspen AI, SelexMB™, MATLAB
• Data types in refineries: time-series, event, lab data, simulations
• Data preprocessing, cleaning, and contextualization

Day 1 Use Case Spotlight:

AI for Predictive Maintenance of Heat Exchangers (Data wrangling, model building, maintenance planning)

Day 2: AI Methods, Modeling & Deployment in Refinery Units

Session 4: Data-Driven Modeling Techniques

• Feature engineering for refinery datasets
• Time-series forecasting (LSTM, ARIMA, Prophet)
• Clustering process conditions (K-means, DBSCAN)
• Neural networks in chemical processes (ANN, CNN basics)

Session 5: Hybrid Modeling (1st Principles + AI)

• What is hybrid modeling?
• Bridging physics and data: use of SelexMB, HYSYS
• Case for hybrid digital twins
• Handling missing variables, soft sensing

Session 6: AI Model Lifecycle in Operations

• Training, validation, and testing
• Model drift and re-training strategies
• Edge vs. cloud deployment
• Integration with existing DCS/SCADA/PI systems
  
  

Day 2 Use Case Spotlight:

Blending Optimization Using ML + ANN (Blending ratio prediction, quality control, economic optimization)

Day 3: Strategic Application, Business Impact & Practical Challenges

Session 7: Key Refinery Challenges AI Can Solve

• Energy efficiency and emissions reduction
• Catalyst deactivation prediction
• Process optimization (CDU, VDU, HCU, FCCU)
• Real-time anomaly detection and root cause analysis
  
  

Session 8: Business Cases and ROI Evaluation

• Creating a business case for AI in refining
• Value quantification: $/bbl, OPEX reduction, MTBF increase
• Pitfalls in scaling AI solutions
• Skills, culture, and change management
  
  

Session 9: Capstone & Group Activity

• Group project: Design an AI solution for a refinery challenge
• Presentations and feedback
• Q&A with an industry expert panel (optional if live session)
  
  

Day 3 Use Case Spotlight:

Real-time Fouling Prediction in Preheat Trains (Machine learning models on historical OLT and process data)

A CDGA attendance certificate will be issued to all attendees completing a minimum of 75% of the total course duration