Utilizing uncertainty information in remaining useful life estimation via Bayesian neural networks and Hamiltonian Monte Carlo

Benker, Maximilian; Furtner, Lukas; Semm, Thomas; Zaeh, Michael F.

doi:10.1016/j.jmsy.2020.11.005

Cited by 53 publications

(18 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Typically, Bayesian inference methods are adopted for quantifying the model prediction uncertainties (Atamuradov et al, 2017;Ramuhalli et al, 2020) because they naturally incorporate information about the target SSC with prior knowledge (e.g., past analysis results, expert opinion) (Coble et al, 2012). Several Bayesian uncertainty quantification approaches have been used in the literature, including BNs (see "Data-Driven Methods" Section), filtering algorithms (such as Kalman filter and particle filter, see "Statistical-Based Prognostics" Section), relevance vector machines (Saha and Goebel, 2008), Bayesian neural networks (Benker et al, 2020), and their variants. Some non-Bayesian approaches have also been proposed for use in the problem of uncertainty estimation, each of which is well suited to specific algorithms or applications.…”

Section: Uncertainty Quantification and Propagationmentioning

confidence: 99%

Prognostics and Health Management in Nuclear Power Plants: An Updated Method-Centric Review With Special Focus on Data-Driven Methods

et al. 2021

View full text Add to dashboard Cite

In a carbon-constrained world, future uses of nuclear power technologies can contribute to climate change mitigation as the installed electricity generating capacity and range of applications could be much greater and more diverse than with the current plants. To preserve the nuclear industry competitiveness in the global energy market, prognostics and health management (PHM) of plant assets is expected to be important for supporting and sustaining improvements in the economics associated with operating nuclear power plants (NPPs) while maintaining their high availability. Of interest are long-term operation of the legacy fleet to 80 years through subsequent license renewals and economic operation of new builds of either light water reactors or advanced reactor designs. Recent advances in data-driven analysis methods—largely represented by those in artificial intelligence and machine learning—have enhanced applications ranging from robust anomaly detection to automated control and autonomous operation of complex systems. The NPP equipment PHM is one area where the application of these algorithmic advances can significantly improve the ability to perform asset management. This paper provides an updated method-centric review of the full PHM suite in NPPs focusing on data-driven methods and advances since the last major survey article was published in 2015. The main approaches and the state of practice are described, including those for the tasks of data acquisition, condition monitoring, diagnostics, prognostics, and planning and decision-making. Research advances in non-nuclear power applications are also included to assess findings that may be applicable to the nuclear industry, along with the opportunities and challenges when adapting these developments to NPPs. Finally, this paper identifies key research needs in regard to data availability and quality, verification and validation, and uncertainty quantification.

show abstract

Section: Uncertainty Quantification and Propagationmentioning

confidence: 99%

Prognostics and Health Management in Nuclear Power Plants: An Updated Method-Centric Review With Special Focus on Data-Driven Methods

et al. 2021

View full text Add to dashboard Cite

show abstract

“…For instance, ensemble approaches were applied for UQ in prognostics in [42] where, rather than simply training independent models, Bayesian model-averaging was also applied to each model in order to obtain multiple predictions for elements in the ensemble. UQ based on Bayesian neural networks and variational inference were only recently investigated in [43], [44] with relatively good results in terms of UQ. The goal of this work is to introduce a new class of methods, DGP models, that tries to integrate the benefits of DNNs into the well-understood Bayesian framework of GP regression.…”

Section: Uq In Data-driven Prognosticsmentioning

confidence: 99%

Uncertainty-Aware Prognosis via Deep Gaussian Process

et al. 2021

View full text Add to dashboard Cite

The problem of Uncertainty Quantification (UQ) is of paramount importance when Machine Learning (ML) and Deep Learning (DL) models are deployed in the real world. In the context of safetycritical applications, such as the prediction of the Remaining Useful-Life (RUL) of infrastructure and industrial assets, the relevance of effective UQ approaches is even higher, given the potentially catastrophic consequences or substantial costs associated with maintenance decisions that are performed either too late or too early. Particularly in such safety critical application contexts, transparency and reliability are essential requirements and any ML-based solution not providing meaningful uncertainty estimates would not fully satisfy such desiderata. However, most ML and DL techniques used for RUL estimation are often not designed to perform UQ, thus limiting their effective applicability in real-world scenarios. To address this limitation, in this paper we investigate the performance of a recently proposed class of algorithms, Deep Gaussian Process (DGPs), for predicting the remaining useful lifetime (RUL). DGPs provide uncertainty estimates associated with their predictions, yet retaining the expressive power and learning capabilities of modern DL methods. DGPs are able to scale to very large datasets, which was a limitation of some of the previous algorithms but has increasingly become an essential requirement for many industrial applications and, in particular, for RUL prediction tasks for which UQ is of key importance. The main contribution of this paper is a thorough evaluation and comparison of several variants of DGPs applied to the problem of RUL prediction. The performance of the DGP algorithms is evaluated on the N-CMAPSS (New Commercial Modular Aero-Propulsion System Simulation) dataset from NASA for aircraft engines. The results demonstrate that DGPs are able to provide very accurate RUL predictions along with sensible uncertainty estimates, providing more reliable solutions for (safety-critical) real-life industrial applications.

show abstract

“…Chinomona et al, for example, applied long short-term memory neural networks to the problem of battery RUL estimation [16], Sun et al applied auto-encoder neural networks to predict the RUL of cutting tools [17] and Yang et al applied convolutional neural networks to the task of bearing RUL prediction [18]. Recent approaches tackle the issue of uncertainty quantification and utilization in DL applications to RUL prediction by applying Bayesian neural networks [9]. Although all the mentioned works demonstrated high accuracies in RUL prediction, they all have in common, that they need large, representative training data sets, which are often not available in real industrial applications [7].…”

Section: Related Workmentioning

confidence: 99%

“…After training, the unobserved sequences from the test data set served as input for the trained GPC model, which predicted the probability of class membership for each single observation within the test sequence up to the latest measurement, resulting in a sequence of HI values over time. which is in accordance with[10] and[9]. In order to account for statistical fluctuations in the results, the experiments were conducted ten times, each time selecting two different random engines for training the GPC model.…”

mentioning

confidence: 93%

A Gaussian Process Based Method for Data- Efficient Remaining Useful Life Estimation

Benker¹,

Bliznyuk

Zaeh

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

The task of remaining useful life (RUL) estimation is a major challenge within the field of prognostics and health management (PHM). The quality of the RUL estimates determines the economical feasibility of the application of predictive maintenance strategies, that rely on accurate predictions. Hence, many effective methods for RUL estimation have been developed in the recent years. Especially deep learning methods have been among the best performing ones setting new record accuracies on bench mark data sets. However, those approaches often rely on numerous and representative run-to-failure sequences of the components under investigation. In real-world use cases, this kind of data (i.e. run-to-failure sequences and RUL labels) is hardly ever present. Therefore, this paper proposes a new, data-efficient method, which is based on Gaussian process classification to derive abstract health indicator (HI) values in a first step, and warped, monotonic Gaussian process regression for indirect RUL estimation in a second step. The proposed approach does neither rely on entire run-to-failure sequences nor on any RUL labels and was tested on the benchmark C-MAPSS turbo fan and FEMTO bearing data sets, achieving comparable results to the state-ofthe art whilst using only a small fraction of the available training data. Hence, the proposed approach allows RUL estimation in use cases, in which gathering enough failure data for the application of deep learning models is infeasible.

show abstract

Utilizing uncertainty information in remaining useful life estimation via Bayesian neural networks and Hamiltonian Monte Carlo

Cited by 53 publications

References 27 publications

Prognostics and Health Management in Nuclear Power Plants: An Updated Method-Centric Review With Special Focus on Data-Driven Methods

Prognostics and Health Management in Nuclear Power Plants: An Updated Method-Centric Review With Special Focus on Data-Driven Methods

Uncertainty-Aware Prognosis via Deep Gaussian Process

A Gaussian Process Based Method for Data- Efficient Remaining Useful Life Estimation

Contact Info

Product

Resources

About