Incorporating Fault Diagnosis into Prognostics: A Fuion Prognostics Strategy for Stochastic Distribution Control Systems

A novel fusion method is proposed for fault diagnosis and prognostics in nonlinear, stochastic distribution control systems. An indirect fault diagnosis algorithm based on fault increment is proposed, which can achieve state estimation and fault diagnosis only through the system output, even when the system state is unobservable. By merging data-based methods with model-based methods, multi-step ahead prediction of the system output is achieved through a double-layer long short term memory network, restoring the resampling functionality of the particle filter and enabling the prediction of the fault magnitude. By integrating fault diagnosis with fault prognostics, the proposed fault prognostics method can effectively reduce the uncertainty of remaining useful life prediction under the condition of unknown prior fault knowledge. Finally, the effectiveness of the proposed algorithm is validated using the wet end control system of a paper machine.