Application of intelligent signal processing to dynamic measurement systems

Abstract

A new method for dynamic measurement is presented. A feature extractor and two-layer artificial neural network is used to predict the final value of a sensor's response while it is still in oscillation. The method permits arbitrary inputs and initial conditions and does not make any assumptions about the model of the sensor. It also copes with non-linearity defects in primary sensors. Introducing a pre-processor as a feature extraction block before the neural network decreases the effect of noise and dramatically reduces the required number of neurones. This, in turn, reduces the complexity of computation and speeds up the real-time measurement. One important advantage of the proposed method is that it can be used in situations where the input function is an impulse, i.e. the transducer senses the measurand for only a very short time interval. This method also allows the possibility of using some features of the sensor signal, such as frequency, that are rarely used in other methods, despite them having a unique relation with the steady state value of the signal. Amplitude noise also has less effect on these characteristics. In addition dynamic neural networks are used in a novel way to cancel the interference signals. The proposed methods are established by theoretical analysis and justified by means of both simulation and measurements on real data.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata