Semi-intrusive multivariable model invalidation

Abstract

This paper introduces a mechanism for testing multivariable models employed by model-based controllers. Although external excitation is not necessary, the data collection includes a stage where the controller is switched to open-loop operation (manual mode). The main idea is to measure a certain "distance" between the closed-loop and the open-loop signals, and then trigger a flag if this "distance" is larger than a threshold level. Moreover, a provision is made for accommodating model uncertainty. Since no hard bounds are assumed with respect to the noise amplitude, the model invalidation mechanism works in a probabilistic framework.

Diagnosis of Poor Control Loop Performance Due to Model−Plant Mismatch

“…The above methods indicate the presence of MPM, but do not attempt to specify which subset of models is mismatched and requires reidentification. In Kammer et al, 21 the control loop is opened so that a disturbance model may be identified. The loop is reclosed, and a time series model is fit between the white noise driving the disturbance and the output error.…”

“…Certain subsets of the inputs can be held constant to determine which specific inputs are mismatched. 21 A detection method based on a twomodel divergence algorithm was proposed for the validation of dynamic models under closed-loop conditions by Huang. 22 Jiang et al 23 proposed a new scheme to detect and isolate MPM for multivariate dynamic systems.…”

“…This method has the advantage that it does not require any external excitation, but it is sensitive to changing disturbances, and the control loop must be initially opened to identify the disturbance model. Certain subsets of the inputs can be held constant to determine which specific inputs are mismatched . A detection method based on a two-model divergence algorithm was proposed for the validation of dynamic models under closed-loop conditions by Huang .…”

A closed-loop cross-correlation method for detecting model mismatch in MIMO model-based controllers

“…The above methods indicate the presence of MPM, but do not attempt to specify which subset of models is mismatched and requires re-identification. In [9] the control loop is opened so that a disturbance model may be identified. The loop is re-closed, and a time series model is fit between the white noise driving the disturbance and the output error.…”

“…This method has the advantage that it doesn't require any external excitation, but it is sensitive to changing disturbances, and the control loop must be initially opened to identify the disturbance model. Certain subsets of the inputs can be held constant to determine which specific inputs are mismatched [9,10].…”