Analysis of frequency-robust multivariable dynamical systems

We consider the problem of studying sensitivity ellipsoidal frequency estimates quality multivariable dynamic systems to parameter variations. To solve problem, we use apparatus functions extreme elements singular value decomposition real-valued transfer matrices. The joint usage with method state space allowed us construct models sensitivity. On basis obtained models, for state, output and error linear continuous in form majorant minorant these have been determined. matrices composed parametric has applied calculations. property minimum sufficiency due substantial possibilities This approach made it possible left corresponding values, select output, spaces subspaces characterized each by largest smallest normal variation amplitude-frequency response. Using right identify which produce proposed solved “optimal nominal” — choice nominal vector primary physical parameters control object aggregates that deliver controlled process. Such include: dimensions various parts characteristics their manufacturing accuracy, properties materials as well values determining design. compare course multidimensional processes