With the development of Natural Gas Pipeline in China, gas turbine & compressor unit has been widely used, so the fault diagnosis of its equipment is important particularly. In this paper, the method based on SDG (Signed Directed Graph) and Bayes theory is applied to fault diagnosis of the equipment. According to SDG model and Bayes theory, this method finds the consistent path and gets the optimizing model of the diagnosis. Then the optimal combination is calculated by implicit enumeration method. Finally, this method is applied to the lubrication system of gas turbine & compressor unit. The results show that this method can complete the multi-sources fault diagnosis quantitatively and improve the diagnosis resolution effectively.Index Terms -SDG, Bayes, multi-sources fault diagnosis, the gas turbine & compressor unit.