Advanced Neural Network-Based Computational Schemes for Robust Fault Diagnosis

Application properties of methods for fault detection and isolation in the diagnosis of complex large-scale processes

“…The approach based on determination of statistical error bounds has the greatest practical importance [7,[48][49][50][51][52][53]. The identification process is carried out without consideration of its uncertainty, while, in the second step, the model uncertainty is modeled (error model) based on residual signal.…”

“…The robustness is determined by the adaptive threshold signal applied to residual. The methodology of forming the envelope of uncertainty in the time domain in respect to fuzzy and neural models is intensively developed at the University of Zielona Góra [48][49][50][51][52]. Fault signaling takes place after exceeding by the residual value upper or lower envelope of the area of uncertainty -adaptation limit.…”

“…The method of sequential residuals [52,53] is based on the analysis of sequences of occurring symptoms. The method allows for designing the secondary residuals in such a way so as to achieve individual symptom sequences for particular faults.…”

Towards Robust Neural-Network-Based Sensor and Actuator Fault Diagnosis: Application to a Tunnel Furnace

“…Unfortunately, the ANNs have disadvantages, e.g., they are usually not available in the state-space form [11,22,30] frequently used for fault diagnosis. Moreover, only rare approaches ensure the stability [21] and there is a limited number of solutions that can settle the robustness problems regarding neural model uncertainty [15,26].…”

“…Neglecting the model uncertainty and measurements noise [12,15] in the FDI system, may result in the undetected faults or false alarms. To solve such a challenging problem, a methodology of dynamic non-linear system identification on the basis of the state-space Group Method of Data Handling (GMDH) neural network [14] was proposed.…”