CNN-based neural network model for amplified laser pulse temporal shape prediction with dynamic requirement in high-power laser facility

• DOI: 10.1364/oe.461396
• Publication Date: 2022-07-26T07:30:08Z
• Authors: Lu Zou; Yuanchao Geng; Bingguo Liu; Fengdong Chen; Wei Zhou; Zhitao Peng; Dongxia Hu; Qiang Yuan; Guodong Liu; Lanqin Liu
• Keywords: Representation; Physical system

The temporal shape of laser pulses is one the essential performances in inertial confinement fusion (ICF) facility. Due to complexity and instability propagation system, it hard predict pulse shapes precisely by pure analytic methods based on physical model [Frantz-Nodvik (F-N) equation]. Here, we present a data-driven convolutional neural network (CNN) for precise prediction. introduces sixteen parameters neglected F-N equation models expand representation dimension. sensitivity analysis experimental results confirms that these have different degrees influence output cannot be ignored. characterizes whole process with commonality specificity features improve description ability. prediction accuracy evaluated root mean square proposed 7.93%, which better compared three optimized models. This study explores nonanalytic methodology combining prior knowledge map complex numerical models, has strong capability great potential other measurable processes science.