This paper introduces a fuzzy logic system that is able to recognize common disturbances during automatic gas metal arc welding (GMAW) using measured welding voltage and current signals. A statistical method was employed to process the captured transient raw data, and the probability density distributions (PDDs) and the class frequency distributions (CFDs) were obtained. Based on the processed data (PDD values of welding voltage and current and CFD values of the short-circuiting time), the system automatically generates fuzzy rules and membership functions of linguistic variables, conducts inference and defuzzification, and completes the evaluation process without further expert knowledge.