Adaptive Observer for Diesel Fault Detection in a Ship Propulsion Benchmark

Prime movers on ships deliver power for both propelling and breaking a vessel and failure of the propulsion unit means loss of the ability to manoeuvre. Propulsion system malfunction must therefore be rapidly detected. The paper analyses detection and isolation of faults in a diesel engine and the associated shaft speed sensor. The dynamics involved is non-linear and one fault is in the system parameters while the other affects a system sensor as an additive fault. The problem is shown to require combined state and parameter estimation. With known structure it is shown to be an advantage to use continuous-time methods, where known functions and parameters can be easily incorporated and the physical parameters can be directly assessed. Application to the non-linear ship propulsion benchmark illustrates the method in detail.