Automated Design Parameter Identification: A New Approach

Abstract

A new method for parameter identification of large dynamic systems is described, and the broad outlines of an automated procedure presented. The dynamic system may consist of an arbitrary assemblage of structural and mechanical elements for which numerical values of certain “design parameters” are to be determined. This “design” problem is formulated in discrete mathematical programming terms as a problem in constrained minimization. The method of solution is indirect, requiring first the time-wise synthesis of element response functions to optimally satisfy the stated design problem. The design parameters subsequently are identified by a function matching procedure in the time domain. For a large class of problems the optimal synthesis phase reduces to a problem in linear programming, while the parameter identification phase is a matter of least squares curve fitting for each design element independently. The great computational advantage over direct methods results from elimination of the need to repetatively solve the system dynamics during the identification process. Thus the computational size of the linear programming problem does not depend on the kinematic degrees of freedom of the dynamic system. An illustrative example involving 32 design parameters is presented.