FAULT DIAGNOSIS AND ISOLATION FOR TWIN ROTOR SYSTEM USING INTERACTIVE KALMAN STATE ESTIMATORS

Abstract

Dynamic systems remain prone to disturbances that cause system parameters to fluctuate around a desired value. The overall system response is degraded due to the small fluctuations in many system parameters at a time. For sensitive applications ultimate accuracy is required and fluctuations must be diagnosed and isolated accurately in time so that respective correction measures are taken to overcome the disturbance. Moreover practically, in large systems sometimes it becomes difficult or even impossible to diagnose and isolate the hidden run time fault that deteriorates the response of the system. Therefore an accurate fault diagnosis and isolation technique is needed to meet the accuracy and troubleshooting requirements for the dynamic system. The goal of current work is to carry out the Fault Diagnostics and Isolation (FDI) for a dynamic system using the Interacting Multiple Models of Kalman filters (IMMKF). The model based FDI is carried out on the basis that the system model takes different shapes for different faults. Each model represents the presence of respective fault type. The IMMKF performs Model based FDI by running multiple Kalman state observers on multiple models of the system to compute measurement residual and its covariance. The mode probability weights are calculated for each model. The model with highest mode probability weight is declared as the current model in effect and respective system condition is diagnosed. Fault isolation is done by examining the difference between the mode probabilities. Larger the difference between mode probabilities of two models more is the isolation between them. The IMMKFs FDI tool is applied on the Lab Twin Rotor System (TRS). The nonlinear TRS model is first linearized and validated to meet the requirement of Kalman state observer that need linear model. The state estimation and its validation is then carried using Kalman observer. Three models for TRS in different system conditions (normal and faulty) are identified using Subspace System Identification (SSI). The FDI is done in normal and faulty conditions using Interacting Multiple models Kalman filters (IMMKF) that produce accurate results that show that the IMMKFs can be used as a highly precise state estimation and FDI tool for dynamic systems in uncertain environments.