The paper presents a model-based approach to detect and compensate for errors of PTC sensors used for the measurement of air flows and air velocities in HVAC (Heating Ventilation and Air Conditioning) systems. Like for commonly used anemometers with mechanically moved parts, the main error sources during long-term operation are soiling by dust transported with the air and failures of electric components due to ageing by recrystallization, ohmic bridges or shortcuts. To overcome these failures, four anemometers consisting of a single PTC sensor element robust procedures for self-control and self-calibration have been developed. Before using an exemplar, the calibration procedure starts with the calibration of all model parameters by evaluation of the current-voltage (I,U) characteristic at /spl nu/ = 0 including the resistance measurement for two air velocities. During operation, all parameters changeable in time as well as the electrical components are self-monitored by repeated measurements of the I-U-characteristic and, for /spl nu/ = 0, re-calibrated accidentally on behalf of this model. This re-calibration is then also used to correct the model parameters for /spl nu/ /spl ne/ 0. Hence, during normal operation, it can be checked whether the soiling or aging of the device becomes critical, such that the sensor have to be cleaned or replaced. All necessary information about its status is inherent in the system itself: From the valve's "close "-position ", the calibration point (air velocity /spl nu/ = 0) is defined; the temperature of the flowing air is measured with an existing temperature sensor. The necessary signal processing is implemented without any extra hardware on the micro-controller already existing in the HVAC system to control the ventilators.