Detecção de falhas em sistemas dinâmicos: abordagens imunoinspiradas.

Abstract

This thesis has proposed, implemented and discussed immune-inspired approaches for dynamic systems fault detection. The fault detection is becoming increasingly challenging due to processes complexity and the agility need to prevent malfunction or even accidents. The challenge lies in diferentiating between normal operation and potential fault. Promising solutions to this problem has emerged through immune-inspired approaches. This thesis has contributed through three approaches: DF-NKC (fault detection inspired on natural killer cell's mechanisms), DF-DM (fault detection inspired on danger model) and DF-Multioperacional (fault detection based on improvements of negative selection algorithms). Firstly, DF-NKC approach has used concepts inspired on natural killer cells biological mechanisms of activation and maturation. Secondly, DF-DM approach has used danger model inspiration and a mathematical model of immune system. Finally, the DF-Multioperational approach was based on existing negative selection algorithms. The DF-NKC and DF-DM were applied in a valve actuator system benchmark provided by DAMADICS (Development and Application of Methods for Actuator Diagnosis in Industrial Control Systems) while DF-Multioperacional was applied to fault detection in a direct current motor system benchmark. The algorithms results show that the approaches are promising for fault detection.