The fault prediction and diagnosis of the new generation weather radar have always been a technical challenge for radar operation and maintenance. The delayed response to some severe faults will lead to aggravated damage to the corresponding components of the radar. In this paper, based on the combination of artificial intelligence technology and radar data characteristics, a SA radar transmitter system fault warning and determination method based on the adaptive threshold is proposed. An algorithm framework for transmitter fault online monitoring based on a reconstruction model is established. The algorithm framework can identify anomalies and warn early when radar transmitter failures occur. The actual operation records are used for calculation and analysis, verifying the method’s effectiveness.