A prognostics approach to nuclear component degradation modeling based on Gaussian Process Regression

Abstract

Abstract-Advanced diagnostics and prognostics tools are expected to play an important role in ensuring safe and long term operation in nuclear power plants. In this context, we use Gaussian Process Regression (GPR) to build a stochastic model of the equipment degradation evolution and apply it for prognostics.GPR is a probabilistic technique for non-linear nonparametric regression that estimates the distribution of the future equipment degradation states by constraining a prior distribution to fit the available training data, based on Bayesian inference. Training data are taken from sequences of degradation measures collected from a set of similar historical equipments which have undergone a similar degradation process. Given new degradation measures from a currently degrading equipment (test trajectory), the distribution of the Remaining Useful Life (RUL) before failure is estimated by comparing with a failure criterion the distribution of the future degradation states predicted by GPR.Applications are shown on simulated data concerning the evolution of creep damage in ferritic steel exposed to high stress and on real data concerning the clogging of sea water filters placed upstream the heat exchangers of a BWR condenser.Index Terms-Remaining Useful Life, Prediction, Prognostics, Bayesian Inference, Gaussian Process Regression, Degradation, Ccreep, Ffilter Cclogging I. INTRODUCTIONUCLEAR INDUSTRY is considering the development and use of advanced diagnostic and prognostic tools to enable longer term safe operation of nuclear structures, systems and components (SSCs) [1][2][3][4][5][6][7]. This interest is mainly originated by the following three reasons:  many nuclear SSCs are becoming old due to the extension of the life of the existing reactors beyond the initial 30 or 40 years. This aging may cause degradation and even failures if SSC maintenance and replacement is not properly planned [8];  many utilities are considering performing power up-rates in their plants. This typically requires increased coolant flow which may cause faster degradation of the SSCs;  there is a strong economic interest towards having longer fuel cycles and decreased outage times. This may pose constraints on the frequency and extent of in-service inspections which may become not able to detect in a timely manner all SSC degradation modes. In this context, in the present paper, we consider the development of prognostic methods for estimating the remaining useful life of nuclear components, structures and systems. Data-driven and model-based methods can be used for predicting the Remaining Useful Life (RUL) of degrading equipment [9-10], i.e., the remaining time during which the equipment can continue performing its function in a safe and efficient way. This allows the implementation of predictive maintenance strategies which have the potential of increasing safety and lowering costs [11]. Model-based methods assume that a mathematical model of the degradation process is available. In practice, the detailed knowledge necessary for b...

Residual life prediction for highly reliable products with prior accelerated degradation data

“…The prediction method for 2 RL was detailed demonstrated in appendix. Figure 4 shows the changing curves of 1 RL ,…”

“…The residual life of a product is defined as the length from the current time to the failure time, and precisely predicting the residual life is important to carry out condition based maintenance (CBM), and prognostics and health management [1,22]. For an individual with high reliability, the degradation data observed under normal stress levels cannot show a distinct degradation trend, therefore it is difficult to precisely predict the residual life.…”

“…Karandikar et al [10] developed a Bayesian method, which applied Monte Carlo simulations to evaluate the posterior means of random parameters, to predict the residual life for an aircraft fuselage panel. Baraldi et al [1] also developed a Bayesian method, which combines the degradation data of historical equipment and the real-time observing degradation data, to predict the residual life for nuclear power plants. Jin et al [9] presented a Bayesian framework based on a Wiener process with random parameters.…”

Residual Useful Life Estimation by a Data-Driven Similarity-Based Approach

“…and Artificial Intelligence (AI) techniques (neural networks, fuzzy systems, etc.) [5][6][7][8][9][10] .…”

A belief function theory based approach to combining different representation of uncertainty in prognostics

“…With respect to the GPR approach, the priors of the mean and covariance function of the GP used to model the clogging process are set, according to [8], as: The GPR approach provides in general narrower prediction intervals than the similarity-based approach, which tends to provide a lower bound of the RUL prediction interval often equal to zero. As pointed out by [3], this does not mean that the evidence of very early failure is high (as demonstrated by the fact that the predicted RUL can be far from zero), but only that the evidence drawn from the reference trajectories is not sufficient to assert with the desired belief 9 .…”