Fault diagnosis and fault tolerant control using reduced order models

The present work considers a model reduction approach for fault diagnosis with generalised likelihood ratio (GLR) method to improve upon its diagnostic performance and computational efficiency in large dimensional applications. Model reduction techniques are widely used in controller design. A similar concept with balanced truncation technique is employed to obtain a reduced order diagnostic model of the process for GLR implementation. The proposed method is incorporated in the fault tolerant control strategy developed by Prakash et al. (2002). Simulation results for a binary distillation column demonstrate the efficacy of the proposed fault detection and identification (FDI) method and the fault tolerant control strategy in comparison to the full-scale implementation for a variety of faults.