Abstract:The high power laser system at Extreme Light Infrastructure—Nuclear Physics has demonstrated 10 PW power shot capability. It can also deliver beams with powers of 1 PW and 100 TW in several different experimental areas that carry out dedicated sets of experiments. An array of diagnostics is deployed to characterize the laser beam spatial profiles and to monitor their evolution during the amplification stages. Some of the essential near-field and far-field profiles acquired with CCD cameras are monitored constantly on a large screen television for visual observation and for decision making concerning the control and tuning of the laser beams. Here, we present results on the beam profile classification obtained from datasets with over 14 600 near-field and far-field images acquired during two days of laser operation at 1 PW and 100 TW. We utilize supervised and unsupervised machine learning models based on trained neural networks and an autoencoder. These results constitute an early demonstration of machine learning being used as a tool in the laser system data classification.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve The paper aims to address the issue of laser beam profile classification in high-power laser systems. Specifically, the research team utilizes machine learning methods to classify the near-field (NF) and far-field (FF) profiles of laser beams. These profile images are captured by CCD cameras during the operation of the laser system to monitor the spatial characteristics of the laser beam and its evolution during the amplification stages. Through this method, the researchers hope to improve the operational efficiency of the laser system, reduce human intervention, and ensure the smooth conduct of experiments. ### Background and Motivation High-power pulsed lasers have pushed the frontiers of science in fields such as high-energy density physics, particle acceleration, and radiation generation. However, complex laser systems require a large number of optical components, such as lenses, gratings, crystals, and filters, to shape the laser beam pulses to the desired parameters. The alignment, calibration, and tuning of these systems are very time-consuming and require significant human resources. Additionally, fluctuations in temperature and humidity in the experimental area or laser room can also lead to changes in the laser beam characteristics, affecting experimental results. ### Solution The research team employed both supervised and unsupervised machine learning models, including trained neural networks and autoencoders, to classify laser beam profile images. The specific methods are as follows: 1. **Supervised Learning**: - Using a simple ANN (Artificial Neural Network) for binary classification to distinguish between near-field and far-field images. - Using pre-trained ResNet-18 and GoogLeNet Convolutional Neural Networks (CNN) for multi-class classification, including near-field, far-field, multi-spot images, and noise images. 2. **Unsupervised Learning**: - Using an autoencoder for unsupervised learning to classify images into near-field, far-field, and multi-spot images. ### Results and Discussion - **Binary Classifier Performance**: - The binary ANN classifier in supervised learning performed well, quickly distinguishing between near-field and far-field images. The accuracy on the test set reached over 90%. - **Autoencoder Performance**: - The autoencoder in unsupervised learning achieved an accuracy of 84.9% in the three-class classification task. The main misclassifications were concentrated on multi-spot images, which were incorrectly classified as near-field or far-field. - **Pre-trained CNN Performance**: - The pre-trained ResNet-18 and GoogLeNet performed excellently in the four-class classification task, with accuracy exceeding 99.8%, completing training and testing within a few minutes. ### Future Prospects This study demonstrates the initial application of machine learning algorithms in the operation of high-power laser systems. Future work will include further development of these technologies to achieve more complex operations, such as correlating laser beam profiles with other beam parameters (e.g., position, pulse duration, and focal spot size). Additionally, researchers plan to apply these models to real-time monitoring of laser systems, providing early warnings to help laser engineers detect and correct errors or misalignments in optical subsystems promptly. Through these efforts, AI-based systems are expected to shorten system preparation and tuning times and potentially initiate a cycle of exponential improvement in experimental results.

Classification of laser beam profiles using machine learning at the ELI-NP high power laser system

Harnessing Machine Learning for Single-Shot Measurement of Free Electron Laser Pulse Power

Automated laser beam characterization using artificial intelligence (AI) for the predictive maintenance of lasers

Machine learning-based analysis of experimental electron beams and gamma energy distributions

Applying Machine Learning Methods to Laser Acceleration of Protons: Lessons Learned from Synthetic Data

Ultra-intense laser pulses and the High Power Laser System at Extreme Light Infrastructure -- Nuclear Physics

MIR laser CEP estimation using machine learning concepts in bulk high harmonic generation

Prediction of laser beam spatial profiles in a high-energy laser facility by use of deep learning

Mixed Diagnostics for Longitudinal Properties of Electron Bunches in a Free-Electron Laser

Beam data modeling of linear accelerators (linacs) through machine learning and its potential applications in fast and robust linac commissioning and quality assurance

Diagnosis of ultrafast ultraintense laser pulse characteristics by machine-learning-assisted electron spin

Toward machine-learning-assisted PW-class high-repetition-rate experiments with solid targets

Investigation of Random Laser in the Machine Learning Approach

Near-field analysis of the high-power laser facility using calculated methods and a residual convolutional neural network with attention mechanism

Beam Detection Based on Machine Learning Algorithms

Virtual Pulse Reconstruction Diagnostic for Single-Shot Measurement of Free Electron Laser Radiation Power

Application of deep learning methods for beam size control during user operation at the Advanced Light Source

Machine Learning Classification of Self-Organized Surface Structures in Ultrashort-Pulse Laser Processing Based on Light Microscopic Images

Neutron dose assessment in laser-generated ultra-short pulsed fields

Machine learning methods for nanolaser characterization

Efficiency improvement of pulse waveform shaping on high power laser facility using deep learning