Turboshaft gas turbine engine is one of the main components in most marine vehicle propulsion systems. The most important part of the turboshaft engines, which has direct impact on the performance of the engine and, as a result, on the performance of the propulsion system and the vehicle, is the engine fuel control system which requires much attention and precise design. The proper design of the fuel control system requires accurate modeling of the fuel system components, such as fuel control unit. Fuel control unit is an electrohydraulic fuel flow control system, which consists of a pump and control valves, which controls the fuel flow to the combustion chamber based on the electronic control unit command. Because of the physical laws governing the hydraulic systems, fuel control unit exhibits purely nonlinear behavior and also for some behavior caused by hysteresis and friction in valves and internal components of the pump, modeling of the fuel control unit is complicated. Therefore, in this article, black box modeling approach based on neural Nonlinear Autoregressive Model with Exogenous Input (NARX) structure is employed to accurately model the fuel control unit. For this, at first a test bench including hydraulic system, sensors, and data acquisition system are designed and constructed to measure and record data from the fuel control unit inputs and outputs. The training as well as validation data were generated using amplitude-modulated pseudorandom binary signal as an excitation signal. Then, the identified model is evaluated with both validation data and different test data. Results show that the obtained model follows the real system with good accuracy and demonstrate the effectiveness of the NARX structure to model the fuel control unit. This model can be used for fuel controller designing or model-in-the-loop/hardware-in-the-loop simulation/test of controller in future works.
