Integrating Fault Diagnosis with Nonlinear Model Predictive Control

Intelligent state estimation for fault tolerant nonlinear predictive control

“…When NMPC formulation is used for inferential control of some unmeasured quality variables, the biased state estimates can have detrimental effect on the closed loop performance. The accuracy of the state estimates, which is the prime concern in the inferential control formulation, can be maintained only if identical model is used for fault diagnosis and control and the model is corrected at the correct location when a fault or abrupt change occurs [12]. Moreover, the permanent augmentation of state space model cannot systematically deal with the difficulties arising out of sensor biases and sensor/actuator failures.…”

“…If it is desired to achieve tolerance with respect to a broad spectrum of faults (abrupt changes in unmeasured disturbance, parameter drifts, sensor/actuator biases) and sensor/actuator failures in a typical situation where the number of degrees of freedom available for observer design (synonymous with the number of measurements available for observer construction) is limited (i.e. far less than the number faults and failures to be dealt), then it becomes imperative to introduce some degree of intelligence in the state estimation to overcome these limitations [12]. In the present work, an intelligent nonlinear state estimation strategy is proposed, which keeps diagnosing the root cause(s) of the plant model mismatch by isolating the subset of active faults and auto-corrects the model on-line so as to accommodate the isolated faults.…”

Online Sensor/Actuator Failure Isolation and Reconfigurable Control Using the Generalized Likelihood Ratio Method

“…If the number of setpoints specified in the NMPC formulation equals the number of manipulated inputs, then we modify the NMPC objective function by relaxing the setpoint on one of the controlled outputs. This strategy has been described in details in Deshpande et al , …”

“…If additional degrees of freedom are not available for manipulation, then we may have to remove one variable from the list of controlled outputs. For example, if process is controlled using the MPC scheme, then we can modify the MPC objective function by relaxing the setpoint on one of the controlled outputs to a zone control variable . In particular in state estimation and prediction in NMPC(MPC), the failed actuator is treated as constant m j ( k ) = b̂ a j for k ≥ t + N , where b̂ a j is the estimate of stuck actuator signal for j th actuator.…”

“…The proposed work involves development of recurrence relationships for signature matrices for diagnosing sensor and actuator failures, which are amenable for online computations. A nonlinear version of the linear GLR method for isolation of sensor and actuator failures in nonlinear systems is then developed . For nonlinear systems, the signature matrix computations are performed using the extended Kalman filter (EKF), and linearized model matrices are computed at each operating point.…”