Serwis Informacyjny "Mechatronika"

Publikacje

tytuł: Using bayes networks in pre-posteriori analysis
kategoria: materiały konferencyjne
dział: inne
opis:

Diagnostic inference and the associated evaluation of usefulness of diagnostic information in the decision-making process is one of the essential elements of the diagnostic process. For example, while starting operations based on the actual technical condition determined with the use of extensive diagnostic systems, apart from the selection and adaptation of diagnostic methods we should also solve numerous additional issues of organizational and economic nature. Let us note that the economic effects, whose evaluation serves as the basis for pre-posteriori analysis, constitute the decisive criterion telling us whether to use a diagnostic system.
Following the preliminary reliability analysis of an object, it often turns out that the result obtained is uncertain – the obtained reliability characteristics have extensive trust range. This is connected with the problems related to making operational decisions, which in turn can be reflected in the object’s maintenance costs which on the one hand are generated by too frequent check-ups and repairs, while on the other hand they result from occurrence of  failures which are accompanied by serious financial and non-financial consequences. In connection with the above it is desirable to reduce the uncertainty of the results of the analysis by taking additional information into account. Obtaining of new information on the object means costs. Performing the pre-posteriori analysis offers the possibility of choosing the relevant diagnostic method, in line with the economic criterion.
The article will present the statistical basis of pre-posteriori analysis as well as the method of building the utility function. We will evaluate the rationale for measuring the average square speed of vibration, which is recommended by PN-ISO 10816-1 norm for evaluating the machine’s vibration. Finally, we will present an example of utilizing the obtained diagnostic information in the conducted analysis. The results obtained will be confirmed with the use of Bayes’ networks which allow the probabilities of occurrence of various values of vibration speeds, the conditional probabilities defining the uncertainty of the measurement method used as well as the incurred cost function to be included in the analysis.

rok wydania: 2008
wydawnictwo: VII SDST
autorzy: Robert Gumiński, Stanisław Radkowski,


 

Copyright © 2008