Real-time monitoring of optical component damage points in long path laser system via a neural network approach

To tackle the current challenges of inadequate speed and accuracy in real-time monitoring and diagnosing damaged optical components within laser systems, neural network-based methods are presented to identify the defective points of optical components in long-path laser systems. When a collimated beam illuminates a damaged optical component at various coaxial positions, the spatial distribution of the near-field spot’s diffraction ring at a fixed receiving location changes. By leveraging deep learning technology, we establish the relationship between the diffraction ring and the axial position of the damaged optical component, which enables precise axial positioning of the defective optical component. Our method successfully achieves axial localization of damaged optical elements within a 25-m-long laser system, offering an axial positioning accuracy of 1 m. It can simultaneously detect 23 optical components while maintaining a mean average precision of 98.8% at an intersection over union threshold of 50%. In addition, the system operates at 214 f/s, meeting the requirements for real-time detection. This research provides an efficient and accurate technical approach for the maintenance and fault prediction of the optical path in laser systems.