Forging machine is widely used in various fields due to its high production efficiency and high utilization rate of material. However, when the forging machine breaks down, it is very difficult to diagnose the fault source of the machine because of the wide variety of faults and the large amount of fault data. To solve this problem, a new diagnostic method combining a fusion algorithm of fault tree and chaotic particle swarm algorithm was proposed, using which the fault source of forging machine can be diagnosed quickly and accurately. Firstly, the fault mode and probability of the forging machine are obtained by fault tree analysis, and then the learning samples of the fault pattern can be summarized according to the fault pattern and the known fault maintenance experience. Finally, the exact location of forging machine faults can be determined quickly using the probability of faults and the ergodic of the particle swarm optimization algorithm. The fault diagnosis experiments were performed on the servo system of forging machine by using the proposed method, and the experimental results was compared with that obtained both from the genetic algorithm and the particle swarm optimization algorithm. It was found that the proposed method has higher accuracy and faster speed in fault diagnosis of the servo system of forging machine.