Bayesian Fault Analysis

“The effects of making the wrong fault call, which in practice could also equate to making no call when a fault was actually present, are modelled though the loss functions in (1), i.e. LА and LB. Recall that LА is the loss associated with making the wrong the decision, d(X) = θB , when the fault was actually θA. Loss values can be defined on any scale, whether monetary or in criticality metrics; however, as with the a priori’s, the work effort involved in derivation of the loss functions could be substantial, as expert subjective analysis is typically required. Bayesian Fault Classification implies “the PNN’s function is to classify or assign an input symptom pattern to a fault category. For most practical pattern classification problems, the Bayes optimal decision surfaces in this multi-dimensional “pattern space” are non-linear, making classification assignment difficult. For the two category example involving the decision rule from (1) above, the decision boundary between the two fault categories, θA and θB, is given by: PА(X) = C PB(X) (2) where C = P'B LB/P'А LА The decision surface represented by (2) can be highly nonlinear, since there are few restrictions on the density functions, PА(X) and PB(X). In practice, the decision surfaces for diagnostic problems involve a large number of fault categories.”, (Logan, Walker, 2009, pp5-6). The authors stresses that “the key to using the PNN for diagnostics is estimating the class probability density functions (PDFs) so that the Bayes decision rules can be implemented” and also “the PNN is readily trained from the results of the FMEA on the machinery plant. There are non-parametric methods of PDF estimation with good accuracy and PDF estimation which are embedded in the softw... ... middle of paper ... ...Integral process ship maintenance implies all processes from diagnostics faults and symptoms of components, equipment and systems to Automatic Generation of Maintenance Work Orders and Operations-Focused Maintenance. Works Cited [1] Military&Aerospace, Shipboard control system, mart 2001, www.milaero.com . [2] Edward J. Walsh, Navy pushing bridge-combat systems integration for new surface warship, jul 2000, www.milaero.com . [3] Wilson, Military & Aerospace Electronics, 10469079, Jul97, Vol. 8, Issue 7. [4] dr Andrija Lompar, "Održavanje i ekonomika brodskih pogona", Kotor 1997. [5] M. K. Pradhan & J. P. Bhol "Trends and Perspectives in Industrial Maintenance Management", M. K. Pradhan & J. P. Bhol, MKP&JPB, 2009. [6] Kevin P. Logan and Capt. John G. Walker, "Diagnostic Technology Insertions for LSD Mid-Life Upgrade“, ASNE Intelligent Ships Symposium VIII, 2009.