Modal analysis testing of a mechanical structure is performed usually by artificial excitation of a structure and measuring input forces and output responses of a mechanical system. The excitation is either transient (impact hammer testing), random, burst-random or sinusoidal (shaker testing). The modern signal processing tools enable to determine properties of a mechanical structure such as resonance frequencies, damping ratios, and mode patterns by measuring the response of the structure without using an artificial excitation. The advantage of this technique is that modal parameters of a structure may be evaluated while the structure is under actual operating conditions. That will allow developing a model within true boundary conditions and actual force and vibration levels. The machine tool structure characteristics that effect productivity and quality have to be evaluated by testing. These characteristics include natural frequencies, modes of vibration, and external sources of high level vibration. Not all modes of machine tool structure effect machine quality. As a result only the modes that are excited during cutting have to be taken in the account. This approach narrowed the frequency range, which has to be considered in test. The machine tool during cutting and/or idling is loaded by a set of external and internal exciting forces. Spectrum, frequency range and application points of these forces are unknown in many cases. Under these exciting forces the vibration between the tool and workpiece, and vibration of machine tool components are sums of many independent vibrations and may be considered as stationary random processes. This assumption allows applying the theory of stationary random processes to machine tool dynamic testing during cutting. Several characteristics of random processes are used to separate harmonic vibration from narrow-band random vibration at natural frequencies. The spectral analysis of machined surface profiles and its correlation with observed vibration allows choosing modes that have to be developed. The analysis of these modes provides a basis for machine tool structure improvement. The proposed experimental approach was verified by experiments at different machine tools. Results of these tests are presented in the paper.