An Exploratory Optimization Plus Kalman Filtering Based Method for Parameter Estimation in Model Based Diagnostics

Physics based models can be used to detect and isolate system faults. As a machine degrades, system outputs deviate from desired outputs, generating residuals defined by the error between sensor measurements and corresponding model simulated signals. Residuals contain valuable information to interpret system states and parameters. A framework for parameter estimation and system identification of non-linear dynamical systems is presented with focus on DC motors and 3-phase induction motors. Tuning combines artificial intelligence techniques like Quasi-Monte Carlo sampling (Hammersley sequencing) and Genetic Algorithm (NSGA II) with an Extended Kalman filter (EKF) that utilizes the system dynamics information via physical models. A tentative Graphical User Interface (GUI) simplified interactions between machine operator and module. Implementation details and results comparing healthy and faulty systems are included.