TY  - JOUR
Y1  - 2018///
SN  - 1860949X
PB  - Springer Verlag
UR  - https://www.scopus.com/inward/record.uri?eid=2-s2.0-85039997120&doi=10.1007%2f978-3-319-71871-2_4&partnerID=40&md5=a2720d829481ba6b82c86df095cd6ceb
JF  - Studies in Computational Intelligence
A1  - Lemma, T.A.
VL  - 743
EP  - 97
AV  - none
N2  - The previous chapter has explained the concepts behind NF based model identification and how it relates to other models and the design in the framework of OBFs. It was stated that a good nonlinear model can be developed from plant operation data or a simulated output without knowing the model structure. However, the model alone is not enough for condition monitoring. In fact, the accuracy of a model is dependent on the estimated model parameters. In this regard, we may have one optimum parameter set out of many parameter sets all capable to characterize the system. In fault detector design, the knowledge of the whole set is critical as the fault detection and diagnosis system relies on model thresholds. In Sect. 4.2 of the chapter, the methods in the calculation of model uncertainity for linear in parameter models and nonlinear in parameter models, respectively, are explained. In the linear case, the equations for upper and lower prediction bounds are defined relying on iid and bounded error assumptions. In Sect. 4.3 the fault detection will be discussed while Sect. 4.4 is dedicated to the design of a fault diagnosis system that operates on bianry or fuzzy signals. Section 4.5 outlines summary of the chapter. © 2018, Springer International Publishing AG.
N1  - cited By 1
ID  - scholars10924
TI  - Model Uncertainity, Fault Detection and Diagnostics
SP  - 75
ER  -