Laser beam welding manufacturing (LBW), being a promising joining technology with superior capabilities of high-precision, good-flexibility and deep penetration, has attracted considerable attention over the academic and industry circles. To date, the lack of repeatability and stability are still regarded as the critical technological barrier that hinders its broader applications especially for high-value products with demanding requirements. One significant approach to overcome this formidable challenge is in-situ monitoring combined with artificial intelligence (AI) techniques, which has been explored by great research efforts. The main goal of monitoring is to gather essential information on the process and to improve the understanding of the occurring complicated weld phenomena. This review firstly describes ongoing work on the in-situ optical sensing, behavior characterization and process modeling during dynamic LBW process. Then, much emphasis has been placed on the optical radiation techniques, such as multi-spectral photodiode, spectrometer, pyrometer and high-speed camera for observing the laser physical phenomenon including melt pool, keyhole and vapor plume. In particular, the advanced image/signal processing techniques and machine-learning models are addressed, in order to identify the correlations between process parameters, process signatures and product qualities. Finally, the major challenges and potential solutions are discussed to provide an insight on what still needs to be achieved in the field of process monitoring for metal-based LBW processes. This comprehensive review is intended to provide a reference of the state-of-the-art for those seeking to introduce intelligent welding capabilities as they improve and control the welding quality.

激光束焊接制造(LBW)作为一种具有高精度、高柔韧性和深熔深等优越性能的连接技术，受到了学术界和工业界的广泛关注。迄今为止，缺乏可重复性和稳定性仍被视为阻碍其更广泛应用的关键技术障碍，尤其是对于要求苛刻的高价值产品。克服这一艰巨挑战的一种重要方法是原位监测与人工智能 (AI) 技术相结合，这已通过大量研究工作进行了探索。监控的主要目标是收集有关过程的基本信息，并提高对正在发生的复杂焊接现象的理解。本综述首先描述了在动态 LBW 过程中正在进行的原位光学传感、行为表征和过程建模工作。然后，重点研究了光辐射技术，例如多光谱光电二极管、光谱仪、高温计和高速相机，用于观察熔池、小孔和蒸汽羽流等激光物理现象。特别是，解决了先进的图像/信号处理技术和机器学习模型，以识别工艺参数、工艺特征和产品质量之间的相关性。最后，讨论了主要挑战和潜在的解决方案，以深入了解在基于金属的 LBW 工艺的工艺监控领域仍需要实现的目标。