📋 Case Study

Refinery FCC Unit Predictive Maintenance Twin

Unplanned shutdowns averaging 17 hrs/unit/yr costing ~$1.2M each

🏗️ Project Overview

Digital twin deployment across 4 fluid catalytic cracking units in Texas Gulf Coast refinery

🎯 Challenge

Unplanned shutdowns averaging 17 hrs/unit/yr costing ~$1.2M each

🔧 Design Approach

Multi-physics twin coupling reactor hydrodynamics, regenerator thermodynamics, and catalyst circulation dynamics; embedded with unscented Kalman filter for state estimation

📐 Design Diagram

AI-generated project design illustration

📐 Key Calculations

Catalyst Circulation Rate Deviation

Δṁ_cat = ṁ_measured − ṁ_twin

Result: +4.2 t/h

Early indicator of cyclone erosion

Regenerator Bed Temp Gradient

∇T = T_top − T_bottom

Result: −82°C

Threshold >−75°C triggers inspection

📊 Results

83% reduction in unplanned outages, 92% accuracy in >48-hr failure prediction, 4.7x ROI in Year 2

💡 Lessons Learned

• •Twin must include mechanical wear models—not just thermofluidic states
• •Edge preprocessing reduces cloud inference latency by 63%

✅ Key Takeaways

• 1Twin must include mechanical wear models—not just thermofluidic states
• 2Edge preprocessing reduces cloud inference latency by 63%