Statistical MIMO controller performance monitoring. Part I: Data-driven covariance benchmark

摘要: In this paper, a data-based covariance benchmark is proposed for control performance monitoring. Within the monitoring scheme, generalized eigenvalue analysis used to extract directions with degraded or improved against benchmark. It shown that eigenvalues and covariance-based index are invariant scaling of data. A statistical inference method further developed corresponding confidence intervals derived from asymptotic statistics. This procedure can be determine subspaces significantly worse better versus The indices within isolated then assess degradation improvement. Two simulated examples, multiloop multivariable MPC system, provided illustrate utility approach. Then an industrial wood waste burning power boiler unit demonstrate effectiveness method.