Fault Isolation Using Extrinsic Curvature of Nonlinear Fault Models

This paper presents an online fault isolation methodology for identifying faulty components in a dynamical system. It is hypothesized that faults in a dynamical system can be suitably represented via nonlinear functions. The isolation scheme, which is implemented online, relies on adaptive nonlinear estimates of these nonlinear fault functions based on the system input output data. The nonlinear fault estimation is achieved using a radial basis function neural network (RBFNN) architecture while the fault isolation is accomplished using extrinsic curvature of the learned RBFNN model. A simple simulation example is presented to illustrate the concept