Abstract: The blueviolet radiation(400-440 nm), infrared radiation(1200-1700 nm) and audio sound(20-20 kHz) are used as signals detected by a multi-sensor real-time monitoring system in the process of CO₂ laser butt welding. The signals can be collected and recorded online. The valuable information in the signals was found out by time-domain, frequency domain and time-frequency analysis. According to the signal features, it can be concluded that the high frequency part(higher than 7000 Hz) of the optical and acoustic signals, especially the blueviolet optical radiation signal can reflect the stability of welding pool and keyhole well, and can be used to identify the weld seam defects caused by large gap joint.