A Revised Minimum Information Loss Method for Model Reduction

Aiming at the problem that the reduced-order model generated by the MIL(Minimum Information Loss) method is not unique, we propose the RMIL(Revised Minimum Information Loss) method. By restricting the system to be the output normal model and transforming the observability Grammian to be an identity matrix, the presented RMIL method minimizes the total information loss and preserves the reduced-order model to be unique.