The computer numerically controlled (CNC) system is the main functional component of the CNC machine tool control system, whose essential part is the servo-drive system. The failure cause of the servo-drive system can be challenging to diagnose due to the complex working environment. Therefore, effectively detecting servo-drive system faults ensures the regular operation of CNC machine tools. In this paper, an improved self-organizing mapped neural network method is proposed and applied to fault diagnosis of the servo-drive system. The MDC technology can identify hidden fault features in data and diagnose fault types based on various data collected by the MDC system and working parameter range indicators. The method's core is the self-organizing mapped neural network that employs the competitive learning mechanism of unsupervised learning to perform cluster analysis on data with different characteristics, find the winning neurons, and diagnose the fault data. In addition, feature standardization and principal component analysis are introduced to preprocess the input data set, which can balance the influence of different feature scales, enhance the fault data features, and reduce data dimension. The rationality of this technique in practical application is validated via a series of fault data sets tests. Lastly, the advantages of the proposed technique are verified by comparison with other standard methods.