New technology allows intelligent sensors by integrating new functionalities: error measure correction, self-calibration, self-diagnosis of measures and sensor status, reconfiguration, digital communication. Industrialists take advantage of more accurate measurements, cost reductions and use facilities. For industrial safety, new dependability issues appear. Some functionalities as self-diagnosis and digital communication seem to be in favour of control systems availability. On the other hand, the high amount of electronics and programmable units implies new failure causes and modes. Availability assessments of intelligent sensors are quite low in literature. Moreover, “intelligent” functionalities are usually not taken into account. In this paper, a discussion about dependability issues and modelling is presented. An approach is proposed. Both functional and structural decompositions of the system are included which allow representing sensor functionalities and types of dependencies. Failures make up the third part of the model. First analyses show the consequences of these failures on sensor functions and the corresponding failure modes during time. An infrared gas sensor is used for example. Some tools for availability assessment of complex systems can take advantage of these results, especially when dysfunctional behaviour is not well known.