Failures in pitch systems may cause fatal damage to industrial wind turbines. One of the main reasons for failures in pitch systems is gas leakages of hydraulic accumulators. Due to the limited accessibility of offshore turbines, automated fault detection algorithms potentially increase turbine availability. The gas leakage is detected without downtime by using a model-based approach together with a bank and extended Kalman filters (EKF’s). The residual is analyzed using multi-model adaptive estimation (MMAE). The applied accumulator model relies on a thermal time constant describing the heat flux from the gas to the surroundings. The thermal time constant has been empirically derived from a prepressure of 50 to 172 bar. The fault detection algorithm is tested experimentally in a laboratory on a 25 liters piston accumulator using a load scenario obtained from real turbine data and a prepressure range of 50–140 bar. The Bank of EKF’s can classify the prepressure within a range and thereby detect if a gas leakage has occurred before it results in failure.