No abstract

No abstract

Many chemical processes exhibit highly nonlinear dynamic behavior when operated over a wide operating range. The fault diagnosis schemes based on a linear perturbation model often prove to be inadequate, with regard to addressing fault diagnosis problems in such systems. In this work, a novel multiple-operating-regimes-based technique is proposed for performing online fault diagnosis in nonlinear systems. A Bayesian approach is used to identify the combination of linear perturbation models in different operating regimes that best-represents the plant dynamics at the current operating point. Nonlinear versions of the generalized likelihood ratio (GLR) method that use multiple linear models for fault identification are proposed. The proposed multimodel based fault diagnosis approaches are computationally efficient and exploit the linearity of each submodel. To arrest the performance degradation caused by the occurrence of faults, the information provided by the fault diagnosis component is then used for online fault accommodation. The efficacy of the proposed diagnosis schemes is demonstrated by conducting simulation studies on two benchmark continuously stirred tank reactor (CSTR) systems and a high-purity binary distillation column system. Analysis of the simulation results reveals that the proposed multimodel Kalman filter-based fault diagnosis schemes outperform the linear GLR method when a nonlinear process is in a transient state over a wide operating range.

No abstract

No abstract