Direct Modeling Parameter Signature Analysis and Failure Mode Prediction of Physical Systems Using Hybrid Computer Optimization

摘要

High speed automated identification and design of dynamic systems, both linear and nonlinear, are discussed. Special emphasis is placed on developing hardware and techniques which are applicable to practical problems. The basic modeling experiment and new results are described. Using the improvements developed successful identification of several systems, including a physical example as well as simulated systems, was obtained. The advantages of parameter signature analysis over signal signature analysis in go-no go testing of operational systems were demonstrated. The feasibility of using these ideas in failure mode prediction in operating systems was also investigated. An improved digital controlled nonlinear function generator was developed, de-bugged, and completely documented. (Author)