Methods of robust virtual sensor design

Abstract

The problem of technical system efficiency increasing by application of virtual sensors is considered. The problem arises when physical sensors are deficient for reducing diagnostic system complexity or solving fault isolation problem. The use of physical sensors to achieve the necessary results may be expensive; besides such sensors as a rule are of non high reliability. The problem of robust virtual sensors design in technical systems described by nonlinear models subjected to the unmatched disturbances is studied. To design such sensors, two different canonical forms are used: identification and Jordan canonical forms. The relations allowing to design virtual sensor of minimal dimension estimating prescribed component of the state vector of the system and insensitive or having minimal sensitivity to the disturbances are obtained. Minimal sensitivity is achieved by using singular value decomposition of the matrices describing the disturbances and the original system. The virtual sensors can be used in addition to existing physical sensors or for replacing the faulty sensor. Theoretical results are illustrated by practical example. The obtained results can be used to solve the problem of fault tolerant system design.