Abstract: Insufficient stiffness of CNC machine tools is the root cause of cutting vibration, which in turn affects tool life, machining quality and cutting efficiency. Based on the testing and analysis technology of the dynamic characteristics of the whole machine tool, this paper proposes a set of technologies that can effectively solve cutting vibration of CNC machine tools and diagnose weak structural stiffness. The core of this technology lies in the testing of the whole-machine dynamic characteristics using the Bayesian operational modal analysis method. This method takes the ambient vibration during machine tool operation as excitation, enabling one-time, high-precision acquisition of the whole-machine modal parameters and overcoming the shortcomings of traditional methods. Through the systematic workflow of "modal testing-vibration spectrum analysis-frequency and mode shape comparison", the specific weak-stiffness components causing vibration can be accurately identified. Taking a certain gear grinding machine as a case study, this method successfully diagnosed stiffness defects in its column, base and other components. Practice shows that this technology has been applied to the diagnosis and improvement of more than 40 CNC machine tools of various types. It can provide direct and quantitative support for machine tool structure optimization, and effectively improve machining stability and cutting efficiency.