Computer numerical control (CNC) machine tool as a core equipment of modern manufacturing, its fault will not only pose a threat to the safety of the staff, but also cause great economic losses to the enterprise. Therefore, it is difficult to implement only by the maintenance of staff. Because of the fault diagnosis technology of CNC machine tools is backward in China, the function and accuracy level are far away from that of the developed countries. Hence, it is urgent to improve the diagnosis technology of domestic CNC machine tools. Based on the systematic summary of the research results at home and abroad, the theory and research progress of fault diagnosis technology of CNC machine tools are analyzed from seven aspects, which combined of Bidirectional Associate Memory (BAM) neural network compensation model, nonlinear dynamic system, neural network and “Transient response characteristics of coolant pressure during shutdown”, and Case Based Reasoning (CBR), Ant Colony Optimization (ACO), integrated Kernel Principal Component Analysis (KPCA) and Particle Swarm Optimization algorithm-Radical Basis Function (PSO-RBF), and etc.