Bunch-by-Bunch Prediction of Beam Transverse Position, Phase, and Length in a Storage Ring Using Neural Networks

TL;DR

The paper tackles the challenge of real-time, multi-parameter diagnostics in diffraction-limited storage rings by predicting transverse position, longitudinal phase, and bunch length directly from BPM waveforms. It introduces a hybrid neural network with dedicated branches (MLP for position, 1D-CNN for length, LSTM-Attention for phase) and a Dynamic Feature Fusion module to combine temporal features with Tshift compensation, enabling end-to-end joint prediction. validated on SSRF and HLS-II data, the approach achieves high accuracy (R^2 up to 0.997 for bunch length and up to 0.991 for position) with sub-millisecond latency (0.042 ms per bunch) and demonstrates robustness across steady-state and injection scenarios. This work advances real-time, multi-parameter beam diagnostics and supports active feedback in next-generation light sources, while highlighting how phase drift due to sampling can be mitigated via Tshift inputs and dynamic fusion. The framework offers a scalable, data-driven alternative to traditional methods like HOTCAP, with potential for extension to multi-turn analysis and broader accelerator diagnostics.

Abstract

Real-time, bunch-by-bunch monitoring of transverse position, longitudinal phase, and bunch length is crucial for beam control in diffraction-limited storage rings, where complex collective dynamics pose unprecedented diagnostic challenges. This study presents a neural network framework that simultaneously predicts these parameters directly from beam position monitor waveforms. The hybrid architecture integrates specialized Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and Long Short-Term Memory with Attention (LSTM-Attention) sub-networks, overcoming key limitations of traditional methods such as serial processing chains and batch-mode operation. Validated on experimental data from the Shanghai Synchrotron Radiation Facility and Hefei Light Source, the model achieves high prediction accuracy with a sub-millisecond latency of 0.042 ms per bunch. This performance demonstrates its potential as a core tool for real-time, multi-parameter diagnostics and active feedback in next-generation light sources.

Figures (11)

• Figure 1: Normalized BPM voltage waveforms of the first bunch in the first revolution at HLS-II.
• Figure 2: Architecture of the joint prediction model for bunch-by-bunch beam parameter estimation.
• Figure 3: MLP-based beam transverse position prediction model with two hidden layers.
• Figure 4: 1D-CNN based bunch length prediction model incorporating two convolutional, two pooling, and one fully-connected layer.
• Figure 5: Operating principle of the LSTM stage.