Application of Bayesian Forecasting to Change Detection and Prognosis of Gas Turbine Performance

Abstract

This paper presents a novel technique for automatic change detection of the performance of gas turbines. In addition to change detection the proposed technique has the ability to perform a prognosis of measurement values. The proposed technique is deemed to be new in the field of gas turbine monitoring and forms the basic building block of a patent pending filed by the authors [1]. The technique used is called Bayesian Forecasting and is applied to Dynamic Linear Models (DLMs). The idea of Bayesian Forecasting is based on Bayes’ Theorem, which enables the calculation of conditional probabilities. In combination with DLMs (which break down the chronological sequence of the observed parameter into mathematical components like value, gradient, etc.) Bayesian Forecasting can be used to calculate probability density functions prior to the next observation, so called forecast distributions. The change detection is carried out by comparing the current model with an alternative model which mean value is shifted by a prescribed offset. If the forecast distribution of the alternative model better fits the actual observation, a potential change is detected. To determine whether the respective observation is a single outlier or the first observation of a significant change, a special logic is developed. Studies have shown that a confident change detection is possible for a change height of only 1.5 times the standard deviation of the observed signal. In terms of prognostic abilities the proposed technique not only estimates the point of time of a potential limit exceedance of respective parameters, but also calculates confidence bounds as well as probability density and cumulative distribution functions for the prognosis.

Prediction Reliability of a Statistical Methodology for Gas Turbine Prognostics

“…[11]. In the same area, Lipowsky et al [24] present a novel statistical method called Bayesian Forecasting as an instrument to forecast gas turbine performance, in order to fit historical data. Zaluski et al [25] develop a data mining methodology to build prognostic models from operational and maintenance data to predict the failures of a bearing and main fuel control of CF-18.…”

“…A different approach for predicting the future trend of a given parameter may be the extrapolation of past trends, as made in Refs. [15,24,27,28] -requires the use of statistical tools and operation data, while it does not require a physics-based model of the considered system -requires limited computational resources, which allow its use in real-time applications. In this manner, the prediction of engine availability can be updated by means of the most recent data, so that the tool can be usefully adopted for fault prognostics, in parallel with gas turbine operation This paper reports an extension of the work of the authors presented in Ref.…”

A Bayesian machine learning approach for online detection of railway wheel defects using track-side monitoring

“…Because of the lack of monitoring data from defective wheels in the model training phase, we are unable to elicit an alternative hypothesis representative of defective wheels by using the training data. Instead, the alternative hypothesis H 1 for defective wheel is made here by shifting the expectations m 3 with small values of h. 44 Thus, referring to equation ( 22), the two hypotheses are given by…”

Sensor-Based Degradation Prediction and Prognostics for Remaining Useful Life Estimation: Validation on Experimental Data of Electric Motors

“…RUL vs Time (cycles) (Coble, 2010). A significant portion of the published literature in prognostics and RUL estimation research focuses on solutions to specific problems, such as electronic prognostics (Mishra & Pecht, 2002;Vichare & Pecht, 2006), vibration analysis (Carden & Fanning, 2004), helicopter gearbox monitoring (Kacprzynski et al,2004;Vachtsevanos et al, 1997;Wang & Vachtsevanos, 2001), JSF applications (Ferrell, 1999;Roemer et al, 2005), turbines prognostics (Cavarzere & Venturini, 2012;Li & Nilkitsaranont, 2009;Lipowsky et al, 2010;Venturini & Therkorn, 2013), etc.. The goal in developing generic prognostic algorithms is to develop methods which may be rapidly configured for a new system to allow for effective and efficient deployment of CBM technology on complex systems; because, while a specific approach may result in very good point solutions for specific problems, generic prognostic algorithms which may be more broadly applicable are clearly of higher interest.…”