Machine learning aided inverse design for flattop beam fiber

• Published in: Optics communications 2022-12, Vol.524, p.128814, Article 128814
• Main Authors: Guo, Yinghao, Cheng, Yudan, Jiang, Youchao, Cao, Min, Tang, Min, Ren, Wenhua, Ren, Guobin
• Format: Article
• Language: English
• Subjects: Artificial neural network; Flattop beam; Inverse design; M-type fiber; Machine learning
• ISSN: 0030-4018; 1873-0310
• DOI: 10.1016/j.optcom.2022.128814

The flattop (FT) beam, one important laser beam, is often applied to the high-power fiber laser. It is preferred to generate the FT beam by M-type fiber. However, the design of optical fiber structure is complex and time-consuming. In this work, based on the M-type fiber, a machine learning method using artificial neural network (ANN) is proposed to inversely design the FT beam fiber. By using this trained ANN, the inverse design of the FT beam fiber is realized, according to the performances of FT beam, the structural parameters of M-type fiber are determined. In addition, the influence of structural parameters on the performances of FT beam, including the flatness, the power confining factor and the effective area, are discussed in detail. The proposed ANN-based machine learning method provides an efficient, accurate prediction for FT beam fiber with excellent performances. •An ANN-based machine learning method for the purpose of inversely designing FT beam fiber is proposed.•The proposed ANN technique performs faster and accurate prediction of the M-type fiber compared with conventional forward parameters sweeping methods.•The proposed ANN-based machine learning method provides an efficient, accurate prediction for FT beam fiber with excellent performances.•For any given special optimized aims, the optical fiber structure can be inversely designed theoretically based on this method.