Vibration monitoring has become an important mean for wear state recognition at cutting tools, bearings, gears, engines and other highly stressed machine components , . The majority of mechanical vibration used to identify the wear state is found in the frequency range from several Hertzto 10kHz . At present, vibration measurement systems are usually based on wideband piezoelectric sensors completed with sophisticated analyzing electronics to observe the spectrum. Because of high costs permanent monitoring is only practicable in safety related applications or at extremely expensive machinery. Future developments in the field of vibration measurement equipment are expected to lead to "smart" sensors with fully digital interface, self-test functionality and on-board storage . In this paper we present a frequency selective capacitive sensor for vibration detection with electrically tunable band selectivity fabricated using near-surface silicon micromechanics (SCREAM). The selectivity is based on the mechanical resonance of the structure, the center frequency of which is variable by direct electrostatic stiffness modulation. This represents a capability of resonance frequency tuning by a control voltage to adjust the measurement range to the desired value. Linearity of the sensor characteristics has been achieved by an optimization of the detection and tuning comb capacitors including FEM-analysis and MATLAB optimization algorithms. By grouping sensor structures with stepped base frequencies into an array the frequency range can be largely extended. To be used as a measurement system the sensor array requires a control unit such as a microcontroller. It handles tasks like cell selection, AD-conversion of the conditioned measurement signal and the generation of the tuning voltage. This measurement system will match the idea of a "smart" sensor since it includes calibration, self test and a digital interface to the outer world. Furthermore it can store data and even draw decissions based on the measured data and in- system algorithms for fault detection.