Speed sensors in the dual clutch transmission (DCT) powertrain play an important role for a DCT vehicle to make the control actions. Incorrect sensor signal may lead unacceptable control strategy, which may seriously degenerate the performance of the vehicle and threaten the safety of the driver. However, due to the harsh environment and with the increasing of the vehicle serving time, the speed sensors tend to be faulty. It is critical to real-time diagnose the sensor faults as to guarantee the safety and dependability of the powertrain. In this study, unknown input observer-based speed sensor fault diagnosis method is proposed based on the dynamic model of the DCT powertrain. Firstly, the diagnosis-oriented dynamic model of the DCT powertrain is constructed, and the speed sensor fault model is developed as well. Secondly, an unknown input observer is constructed to estimate the speed information of the powertrain. The residual between the speed sensor signal and the estimated speed is calculated as the fault indicator, once the residual exceeds a certain value, it can be indicated that a malfunction has arisen in the speed sensors. Thirdly, a bank of unknown input observers is developed to further isolate which speed sensor is faulty. Simulation results verify that the proposed sensor fault diagnosis strategy can detect and isolate the speed sensor fault timely and accurately.