The paper attempts to evaluate the effect of acceleration sensor mounting on the recorded vibration time course. The study used a prepared model of a railroad rail and triaxial acceleration sensors. Three non-invasive methods of mounting the vibration acceleration transducers were selected for analysis: mounting with cyanoacrylate glue, mounting with a magnet, and mounting with wax. The information capacity of the signals was analyzed based on the recorded time waveforms, which totaled more than 90, and their vibration signals. The analysis compared both the basic parameters of the signals (maximum amplitudes and root mean square values) and a comprehensive analysis of the signals using the short-time Fourier transform method, as well as the wavelet transform. The results show significant differences in the recorded signal parameters depending on how the acceleration sensor is mounted, as well as the axis analyzed. The differences can negatively affect the correctness of the measurements made and falsify the picture of the real condition.