Many existing control systems for vehicles use an effect of a prediction of a friction coefficient between wheels and rails. It is connected with the impossibility of measuring a friction coefficient at the moment of running vehicle movement. One of the ways to solve this task is to use noise spectrum analysis for friction coefficient detection. At the present time different noise sources are localized, calculated and measured for railway vehicles. Research shows that at a speed up to 300 km/h, the noise of wheel rolling on the surface is predominant. And the noise can be subdivided into three categories: rolling noise, grind and scream on the curve parts of roads and dynamical noise (usually, this noise depends on impact loads). However, the noise for wheel-rail interaction is less researched. The researches also have not done a diagnostic or a monitoring of technical conditions for the wheel-road contact (for example, third body in the contact zone between rail and wheel). Some of these conditions are able to have a negative effect on adhesion characteristics. The form of working contact surfaces has a major influence on the tense distribution, which also has a strong influence on the spectrum of noise. Based on this approach it is possible to develop an adhesion control system of railway vehicles. Results from this study will be presented and the effectiveness of this work will be discussed.