Citation Context

...a referential). A central problem can be pointed out: the accuracy of a prognostic system is related to its ability to approximate and predict the degradation of an equipment; the prediction phase INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT Issue 2, Volume 4, 2010 20 is a critical one. A look at prognostic metrics enables to point it out. C. Prognostic metrics There is no general agreement as to an appropriate and acceptable set of metrics that can be employed in prognostic applications, and researchers and maintenance practitioners are still working on this [46], [39], [40], [26]. Various measures emerge however from literature and are presented hereafter. As for any industrial task, prognostic can be evaluated at least in two ways: – the main objective of prognostic is to provide the efficient information that enables the underlying decision process, i.e., the choice of maintenance actions. Thus, a first set of metrics are those that quantify the risks incurred by the monitored system. This kind of metrics can be called the prognostic measures, – assuming that prognostic is in essence an uncertain process, it is useful to be able to judge from its "qualit...

Citation Context

... Several recent publications have suggested metrics for quantifying prognostic uncertainty and evaluating the ability of control modifications to affect meaningful change on fault growth predictions (=-=Saxena et al., 2008-=-; P. Wang et al., 2012; Edwards et al., 2010). This paper describes a process for abstracting uncertain models of environmental and fault growth dynamics into the generalized notation of a non-determi...

Citation Context

...e the PF estimates are shown as blue dots. The large spread of the blue dots is caused by the bias errors and noise in the Hall effect current sensors used. Since this uncertainty has not been expressly modeled, the actual EOD can sometimes lie outside the predicted pdf as shown in Figures 6 – 8. Figure 6. Sample prediction using Model 1 0 500 1000 1500 15 16 17 18 19 20 21 22 secs V o lt s Figure 7. Sample prediction using Model 2 Figure 8. Sample prediction using Model 3 For statistical validation, we ran each model 100 times over the same data to generate the - metric plots as defined in Saxena et al. (2008). This prognostic metric measures whether RUL predictions continue improve in accuracy with time as more run-time data is made available, where t i RUL = t i EOD – tp. It also enforces the notion that the prediction error needs to reduce as the prediction time instant approaches the end of life (EOD in this case) since there is less time to take corrective action. For these experiments, the value is chosen to be 0.1 and is chosen to be 0.5 (i.e. it is desired that the prediction trajectories be within 90% accuracy with 50% battery life left). Figures 9 – 11 show the - plots for Models 1 ...

Citation Context

...ality, predictably can be measured on the basis of forecast error based approach. Various measures have been reported in literature to judge the quality of prediction or selecting a prediction model (=-=Saxena et al., 2008-=-, 2009, 2010; Monnet & Berger, 2010). See Eq. (1) for a set of potential metrics that can be used to assess predictability: MSE = 1 N × ∑ ( N i=1 MAPE = 100 N × ∑ ( N i=1 ∣ RMSE = √ MSE CVRMSE = RMSE/...

Citation Context

... based metrics that focus on how well prognostic algorithms are doing. These metrics can be categorized into accuracybased metrics, precision-based metrics, and robustness-based metrics. Saxena et al =-=[16]-=- provide a comprehensive review of metrics for evaluating performance of prognostic techniques. Accuracy-based metrics describe the closeness of RUL estimates to the actual remaining life. Estimates t...

Citation Context

...y activities of the methodology (see Table 2). Some of these activities have been formalized: transformation from high-level requirements to business case (Saxena et al., 2012); (2) metric selection (=-=Saxena et al., 2008-=-); and (3) validation and verification tests (L. Tang, Orchard, Goebel, & Vachtsevanos, 2011). A key activity that the methodology must integrate is the definition and integration of metrics as a mean...

Citation Context

... support, and cost measures.sPHM characteristics are measures of the PHM system itself and how well it diagnosessand predicts future failures.sWhile up to 18 possible PHM metrics have been identifieds=-=[9]-=-, there are four basic metrics that are fairly representative:sfalse alarm rate, accuracy,s7sprecision, and prediction lead time.sFalse alarm rate is the number of times a PHMssystem indicates there i...

Citation Context

...04). In Section 5, wesdescribe uncertainty quantification in RUL estimation andsreview the density estimation methods.sIn Section 6, wesinclude the discussion of the performance metrics describedsin (=-=Saxena, et al., 2008-=-). In Section 7, we review the datasets(Saxena & Goebel, 2008) and describe the procedures forsthe experiment. In Section 8 and 9 we present results andsfindings then conclude the paper in Section 10....

Citation Context

...ost factors. Statistical values as Mean Time Between Repair (MTBR) are not evaluated in this study, because the results will not have any impact on these parameters. For further information see e.g. (=-=Saxena et al., 2008-=-). 3. MODEL APPLICATION In this section the results of an exemplary simulation model application are summarised. Due to IP reasons a detailed description of the maintenance process logic as well as pa...