On-board sensor failure detection of an active suspension system using the generalized likelihood ratio approach

The authors extend the application of the generalized likelihood ratio (GLR) approach for on-board sensor failure detection and identification (FDI) of an active suspension system using a half-car (bicycle) model. It is assumed that the control design allows for one accelerometer and two linear variable differential transformers. The failures considered are bias and increased sensor noise. The advantages and limitations of the GLR approach in locating the failed sensor and in detecting different types of failures are assessed. It is concluded that the application of this approach is feasible when the failure can be modeled as a deterministic additive term. In other situations the computational requirements make it less practical