Hysteresis compensation studies focus on developing systems to predict and correct for Magnetic Hysteresis effects in SPS operations. The goal is real-time field prediction and correction to maintain beam stability.
Modeling Approaches
Neural Network Models
- Temporal Fusion Transformer - Primary production model
- PhyLSTM - Physics-informed approach
- PETE - Physics-enhanced transformer encoder
Training Strategies
- Curriculum Learning - Progressive complexity training
- Transfer Learning and Fine-tuning - Adapting models to new data
- Choice of optimizer - Training optimization
Data Sources
Machine Development Sessions
- Machine Development sessions provide validation data
- Dedicated MD sessions for focused hysteresis studies
- Parallel MD sessions for concurrent validation
Operational Data
- B-Train measurements - Field mapping data
- MBI data - Main bending magnet measurements
- Dipole datasets v9 - Production dataset
Real-time Deployment
Infrastructure
- sps-app-hysteresis - Real-time application
- FGC Real-time channel correction - Hardware integration
- Economy modes for hysteresis compensation - Performance optimization
Operational Integration
- Hysteresis Compensation - Main production system
- Compensation event flow - Processing pipeline
Related Concepts
- Magnetic Hysteresis - Underlying physical phenomenon
- SPS Operations - Broader accelerator context
- Real-time Prediction Systems - Technical infrastructure