Forecasting remaining useful life: Interpretable deep learning approach via variational Bayesian inferences

Mathias Kraus; Stefan Feuerriegel

arxiv: 1907.05146 · v2 · pith:F4M4EJRYnew · submitted 2019-07-11 · 💻 cs.LG · stat.ML

Forecasting remaining useful life: Interpretable deep learning approach via variational Bayesian inferences

Mathias Kraus , Stefan Feuerriegel This is my paper

Pith reviewed 2026-05-24 23:10 UTC · model grok-4.3

classification 💻 cs.LG stat.ML

keywords remaining useful lifestructured-effect neural networkvariational Bayesian inferenceinterpretabilitypredictive maintenanceaircraft enginesdeep learning

0 comments

The pith

A structured-effect neural network predicts remaining useful life with both deep learning flexibility and statistical interpretability.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes a structured-effect neural network to forecast the remaining useful life of equipment such as aircraft engines. This method aims to merge the accuracy of machine learning with the explanatory power of traditional probability density functions fitted to lifetimes. By using variational Bayesian inferences for parameter estimation, the approach seeks to maintain accountability in predictions while incorporating sensor data and environmental variables. A sympathetic reader would care because it promises better preemptive maintenance decisions that reduce costs without sacrificing the ability to understand and trust the forecasts. The comparison on real time-to-failure data demonstrates its performance advantages.

Core claim

The structured-effect neural network combines favorable properties of both statistical baselines and machine learning by offering high accountability and the flexibility of deep learning for predicting remaining useful life. Parameters are estimated via variational Bayesian inferences. When compared to baselines on actual time-to-failure data for aircraft engines, it shows strong performance alongside superior interpretability, with implications for decision support in maintenance.

What carries the argument

The structured-effect neural network, which embeds structured effects to retain interpretability while allowing deep learning components.

If this is right

Facilitates preemptive maintenance decisions by providing anticipated time-to-failure.
Promises to reduce costs associated with failures.
Provides forecasts that incorporate deterioration processes and environmental variables through sensor data.
Demonstrates performance and superior interpretability on aircraft engine data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The method could apply to predictive maintenance in other industries like manufacturing or infrastructure.
It may inspire hybrid interpretable models in other time-to-event forecasting tasks.
The variational Bayesian approach might allow uncertainty quantification to inform risk-based maintenance decisions.

Load-bearing premise

Embedding structured effects into the neural network preserves interpretability without variational Bayesian estimation introducing hidden dependencies that undermine accountability.

What would settle it

If the predictions on the aircraft engine data do not show higher interpretability than black-box machine learning methods while maintaining accuracy.

Figures

Figures reproduced from arXiv: 1907.05146 by Mathias Kraus, Stefan Feuerriegel.

**Figure 2.** Figure 2: This plot visualizes the RUL predictions made by the structured-effect LSTMs based on a log-normal and [PITH_FULL_IMAGE:figures/full_fig_p021_2.png] view at source ↗

**Figure 3.** Figure 3: These histograms illustrate the posterior distribution of the estimated parameters (i. e., shape, mean and [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗

read the original abstract

Predicting the remaining useful life of machinery, infrastructure, or other equipment can facilitate preemptive maintenance decisions, whereby a failure is prevented through timely repair or replacement. This allows for a better decision support by considering the anticipated time-to-failure and thus promises to reduce costs. Here a common baseline may be derived by fitting a probability density function to past lifetimes and then utilizing the (conditional) expected remaining useful life as a prognostic. This approach finds widespread use in practice because of its high explanatory power. A more accurate alternative is promised by machine learning, where forecasts incorporate deterioration processes and environmental variables through sensor data. However, machine learning largely functions as a black-box method and its forecasts thus forfeit most of the desired interpretability. As our primary contribution, we propose a structured-effect neural network for predicting the remaining useful life which combines the favorable properties of both approaches: its key innovation is that it offers both a high accountability and the flexibility of deep learning. The parameters are estimated via variational Bayesian inferences. The different approaches are compared based on the actual time-to-failure for aircraft engines. This demonstrates the performance and superior interpretability of our method, while we finally discuss implications for decision support.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper builds a neural net with explicit structured effects fit by variational Bayes for RUL on engine data, and it actually shows the derivations plus both accuracy numbers and qualitative effect plots.

read the letter

The core contribution is a neural network that keeps additive structured effects on the inputs so the forecast can still be read as a sum of interpretable terms, while the rest of the network supplies the flexibility. They fit it with variational Bayesian inference and compare it to a plain probabilistic baseline on the aircraft engine dataset, reporting lower error and some plots that show how individual sensors shift the predicted life. That combination is new enough in this domain and the paper does the work of writing down the model and the ELBO objective rather than just asserting it. The experiments include both predictive metrics and qualitative checks on the effects, which is better than many hybrid papers that stop at accuracy. The main soft spot is exactly the one the abstract flags: the variational approximation could create hidden dependencies among the structured terms, and the paper only checks interpretability qualitatively rather than with a direct test of whether the effects remain cleanly separable after fitting. No other load-bearing gaps show up in the derivations or the data handling. This is for readers who need practical RUL models that are not fully black-box, especially in maintenance settings. It is coherent on its own terms and has enough formal and empirical grounding to go to referees, even if the interpretability claim will need tightening in revision.

Referee Report

1 major / 1 minor

Summary. The paper proposes a structured-effect neural network for remaining useful life (RUL) prediction that integrates the interpretability of classical structured models with the flexibility of deep learning. Parameters are estimated via variational Bayesian inference. The approach is compared to baselines on aircraft engine time-to-failure data and is claimed to deliver both high accountability and superior predictive performance for maintenance decision support.

Significance. If the central claim holds, the work would be significant for prognostics by addressing the black-box limitation of machine learning while retaining structured interpretability. The explicit derivation of structured effects combined with a stated variational objective and reported qualitative interpretability checks on real data represent a strength, as does the focus on accountable predictions for practical decision support.

major comments (1)

[Abstract / variational Bayesian section] Abstract and methods description of variational inference: the central claim that structured effects deliver accountability is load-bearing, yet it is unclear whether the variational approximation (as stated in the objective) introduces hidden dependencies among the structured effects that would undermine interpretability; a concrete check or bound showing preservation of the structured interpretation is needed.

minor comments (1)

[Abstract] Abstract: no quantitative metrics, baselines, or error bars are supplied to support the performance claim, even though the full manuscript reportedly includes them.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed review and constructive feedback. We address the major comment point-by-point below.

read point-by-point responses

Referee: [Abstract / variational Bayesian section] Abstract and methods description of variational inference: the central claim that structured effects deliver accountability is load-bearing, yet it is unclear whether the variational approximation (as stated in the objective) introduces hidden dependencies among the structured effects that would undermine interpretability; a concrete check or bound showing preservation of the structured interpretation is needed.

Authors: We appreciate the referee pointing out this potential issue with the variational approximation. Upon re-examination, the variational posterior in our model is a mean-field approximation that factorizes over the structured effects, thereby not introducing dependencies between them. The structured interpretation is thus preserved by design of the variational family. To make this clearer for readers, we will add an explicit statement and a short derivation in the methods section confirming that the ELBO does not couple the structured effects in a way that affects their individual interpretability. We will also include a supplementary check verifying low posterior correlations in the real data application. This constitutes a revision to the manuscript. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper defines a structured-effect neural network whose parameters are estimated by variational Bayesian inference and whose interpretability is attributed to the explicit structured-effect terms. No equation is shown to reduce the final RUL forecast to a fitted quantity by construction, no prediction is relabeled as such after fitting on the same data, and no load-bearing uniqueness theorem or ansatz is imported via self-citation. The derivation therefore remains self-contained against external benchmarks and does not match any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that sensor data encodes deterioration processes that can be captured by a neural network, plus the modeling choice that variational Bayesian inference yields interpretable parameters; no free parameters or invented entities are identifiable from the abstract alone.

axioms (1)

domain assumption Sensor data incorporates deterioration processes and environmental variables
Stated in the abstract as the reason machine learning can outperform simple lifetime distributions.

pith-pipeline@v0.9.0 · 5739 in / 1136 out tokens · 22744 ms · 2026-05-24T23:10:09.206478+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

SENN_Θ(t;X_t,...,X_1) = λ(t) + βᵀX_t + RNN_Θ(X_t,...,X_1,t) with variational ELBO
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Weibull / log-normal baseline + linear + LSTM components

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages · 1 internal anchor

[1]

Heidergott, T

B. Heidergott, T. Farenhorst-Yuan, Gradient estimation for multicomponent maintenance systems with age-replacement policy, Operations Research 58 (2010) 706–718

work page 2010
[2]

Groenevelt, L

H. Groenevelt, L. Pintelon, A. Seidmann, Production lot sizing with machine breakdowns, Management Science 38 (1992) 104–123

work page 1992
[3]

Dogramaci, N

A. Dogramaci, N. M. Fraiman, Replacement decisions with maintenance under uncertainty: An imbedded optimal control model, Operations Research 52 (2004) 785–794

work page 2004
[4]

H. L. Lee, M. J. Rosenblatt, Simultaneous determination of production cycle and inspection schedules in a production system, Management Science 33 (1987) 1125–1136

work page 1987
[5]

A. K. Jardine, D. Lin, D. Banjevic, A review on machinery diagnostics and prognostics implementing condition-based maintenance, Mechanical Systems and Signal Processing 20 (2006) 1483–1510

work page 2006
[6]

A. Heng, S. Zhang, A. C. Tan, J. Mathew, Rotating machinery prognostics: State of the art, challenges and opportunities, Mechanical Systems and Signal Processing 23 (2009) 724–739

work page 2009
[7]

X.-S. Si, W. Wang, C.-H. Hu, D.-H. Zhou, Remaining useful life estimation: A review on the statistical data driven approaches, European Journal of Operational Research 213 (2011) 1–14

work page 2011
[8]

Baptista, S

M. Baptista, S. Sankararaman, I. P. de Medeiros, C. Nascimento, H. Prendinger, E. M. Henriques, Forecasting fault events for predictive maintenance using data-driven techniques and ARMA modeling, Computers & Industrial Engineering 115 (2018) 41–53

work page 2018
[9]

Seera, C

M. Seera, C. P. Lim, S. Nahavandi, C. K. Loo, Condition monitoring of induction motors: A review and an application of an ensemble of hybrid intelligent models, Expert Systems with Applications 41 (2014) 4891–4903

work page 2014
[10]

Breiman, Statistical modeling: The two cultures, Statistical Science 16 (2001) 199–231

L. Breiman, Statistical modeling: The two cultures, Statistical Science 16 (2001) 199–231

work page 2001
[11]

Jang, A decision support framework for robust R & D budget allocation using machine learning and optimization, Decision Support Systems 121 (2019) 1–12

H. Jang, A decision support framework for robust R & D budget allocation using machine learning and optimization, Decision Support Systems 121 (2019) 1–12

work page 2019
[12]

H. S. Subramania, V. R. Khare, Pattern classiﬁcation driven enhancements for human-in-the-loop decision support systems, Decision Support Systems 50 (2011) 460–468

work page 2011
[13]

Delen, H

D. Delen, H. Zaim, C. Kuzey, S. Zaim, A comparative analysis of machine learning systems for measuring the impact of knowledge management practices, Decision Support Systems 54 (2013) 1150–1160

work page 2013
[14]

K. L. Reifsnider, S. W. Case, Damage tolerance and durability of material systems, Wiley Interscience, New York, NY, 2002

work page 2002
[15]

Y. Liu, B. Stratman, S. Mahadevan, Fatigue crack initiation life prediction of railroad wheels, Interna- tional Journal of Fatigue 28 (2006) 747–756. 30

work page 2006
[16]

Papakostas, P

N. Papakostas, P. Papachatzakis, V. Xanthakis, D. Mourtzis, G. Chryssolouris, An approach to opera- tional aircraft maintenance planning, Decision Support Systems 48 (2010) 604–612

work page 2010
[17]

Z. C. Lipton, The mythos of model interpretability, Communications of the ACM 61 (2018) 36–43

work page 2018
[18]

Rudin, Please stop explaining black box models for high stakes decisions, in: Conference on Neural Information Processing Systems, 2018

C. Rudin, Please stop explaining black box models for high stakes decisions, in: Conference on Neural Information Processing Systems, 2018

work page 2018
[19]

Y. Lou, R. Caruana, J. Gehrke, G. Hooker, Accurate intelligible models with pairwise interactions, in: SIGKDD International Conference on Knowledge Discovery and Data Mining, 2013, pp. 623–631

work page 2013
[20]

Saxena, K

A. Saxena, K. Goebel, C-mapss data set, NASA Ames Prognostics Data Repository (2008)

work page 2008
[21]

J. B. Butcher, D. Verstraeten, B. Schrauwen, C. R. Day, P. W. Haycock, Reservoir computing and extreme learning machines for non-linear time-series data analysis, Neural networks 38 (2013) 76–89

work page 2013
[22]

Dong, X.-Y

D. Dong, X.-Y. Li, F.-Q. Sun, Life prediction of jet engines based on LSTM-recurrent neural networks, in: Prognostics and System Health Management Conference, IEEE, 2017

work page 2017
[23]

L. Liao, F. Kottig, Review of hybrid prognostics approaches for remaining useful life prediction of engineered systems, and an application to battery life prediction, IEEE Transactions on Reliability 63 (2014) 191–207

work page 2014
[24]

Navarro, T

J. Navarro, T. Rychlik, Comparisons and bounds for expected lifetimes of reliability systems, European Journal of Operational Research 207 (2010) 309–317

work page 2010
[25]

M. Dong, D. He, Hidden semi-Markov model-based methodology for multi-sensor equipment health diagnosis and prognosis, European Journal of Operational Research 178 (2007) 858–878

work page 2007
[26]

M. J. Kim, V. Makis, Joint optimization of sampling and control of partially observable failing systems, Operations Research 61 (2013) 777–790

work page 2013
[27]

Plitsos, P

S. Plitsos, P. P. Repoussis, I. Mourtos, C. D. Tarantilis, Energy-aware decision support for production scheduling, Decision Support Systems 93 (2017) 88–97

work page 2017
[28]

Ghosh, R

D. Ghosh, R. Sharman, H. Raghav Rao, S. Upadhyaya, Self-healing systems: survey and synthesis, Decision Support Systems 42 (2007) 2164–2185

work page 2007
[29]

Mazhar, S

M. Mazhar, S. Kara, H. Kabernick, Remaining life estimation of used components in consumer products: Life cycle data analysis by Weibull and artiﬁcial neural networks, Journal of Operations Management 25 (2007) 1184–1193

work page 2007
[30]

Cox, Regression models and life-tables, Journal of the Royal Statistical Society 34 (1972) 187–220

work page 1972
[31]

Y. Wang, S. Wang, Y. Fang, P. Y. Chau, Store survival in online marketplace: An empirical investigation, Decision Support Systems 56 (2013) 482–493

work page 2013
[32]

Zihajehzadeh, E

S. Zihajehzadeh, E. J. Park, Regression model-based walking speed estimation using wrist-worn inertial 31 sensor, PLOS ONE 11 (2016) e0165211

work page 2016
[33]

A. Riad, H. Elminir, H. Elattar, Evaluation of neural networks in the subject of prognostics as compared to linear regression model, International Journal of Engineering & Technology 10 (2010) 52–58

work page 2010
[34]

Mosallam, K

A. Mosallam, K. Medjaher, N. Zerhouni, Nonparametric time series modelling for industrial prognostics and health management, The International Journal of Advanced Manufacturing Technology 69 (2013) 1685–1699

work page 2013
[35]

G. S. Babu, P. Zhao, X.-L. Li, Deep convolutional neural network based regression approach for estimation of remaining useful life, in: Database Systems for Advanced Applications, volume 9642, Springer, 2016, pp. 214–228

work page 2016
[36]

Y. Wu, M. Yuan, S. Dong, L. Lin, Y. Liu, Remaining useful life estimation of engineered systems using vanilla LSTM neural networks, Neurocomputing 275 (2018) 167–179

work page 2018
[37]

Zheng, K

S. Zheng, K. Ristovski, A. Farahat, C. Gupta, Long short-term memory network for remaining useful life estimation, in: IEEE International Conference on Prognostics and Health Management, IEEE, 2017, pp. 88–95

work page 2017
[38]

Verification for Machine Learning, Autonomy, and Neural Networks Survey

W. Xiang, P. Musau, A. A. Wild, D. M. Lopez, N. Hamilton, X. Yang, J. Rosenfeld, T. T. Johnson, Veriﬁcation for machine learning, autonomy, and neural networks survey, arXiv preprint arXiv:1810.01989 (2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018
[39]

Athalye, N

A. Athalye, N. Carlini, D. Wagner, Obfuscated gradients give a false sense of security: Circumventing defenses to adversarial examples, in: International Conference on Machine Learning, 2018

work page 2018
[40]

Singh, T

G. Singh, T. Gehr, M. Mirman, M. P¨ uschel, M. Vechev, Fast and eﬀective robustness certiﬁcation, in: Conference on Neural Information Processing Systems, 2018, pp. 10802–10813

work page 2018
[41]

M. T. Ribeiro, S. Singh, C. Guestrin, Why should i trust you? explaining the predictions of any classiﬁer, in: SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 1135–1144

work page 2016
[42]

S. M. Lundberg, S.-I. Lee, A uniﬁed approach to interpreting model predictions, in: Advances in Neural Information Processing Systems, 2017, pp. 4765–4774

work page 2017
[43]

van der Maaten, G

L. van der Maaten, G. Hinton, Visualizing data using t-SNE, Journal of Machine Learning Research 9 (2008) 2579–2605

work page 2008
[44]

W. J. Murdoch, C. Singh, K. Kumbier, R. Abbasi-Asl, B. Yu, Interpretable machine learning: Deﬁnitions, methods, and applications, arXiv preprint (2019)

work page 2019
[45]

P. J. Vlok, J. L. Coetzee, D. Banjevic, A. K. S. Jardine, V. Makis, Optimal component replacement decisions using vibration monitoring and the proportional-hazards model, Journal of the Operational Research Society 53 (2002) 193–202. 32

work page 2002
[46]

Hastie, R

T. Hastie, R. Tibshirani, J. H. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed., Springer, New York, NY, 2009

work page 2009
[47]

Goodfellow, Y

I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, MIT Press, Cambridge, MA, 2017

work page 2017
[48]

Hochreiter, J

S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural Computation 9 (1997) 1735–1780

work page 1997
[49]

Fischer, C

T. Fischer, C. Krauss, Deep learning with long short-term memory networks for ﬁnancial market predic- tions, European Journal of Operational Research 270 (2018) 654–669

work page 2018
[50]

Kraus, S

M. Kraus, S. Feuerriegel, A. Oztekin, Deep learning in business analytics and operations research: Models, applications and managerial implications, arXiv preprint (2018)

work page 2018
[51]

Srivastava, S

S. Srivastava, S. Lessmann, A comparative study of LSTM neural networks in forecasting day-ahead global horizontal irradiance with satellite data, Solar Energy 162 (2018) 232–247

work page 2018
[52]

Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society

R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society. Series B (Methodological) 58 (1996) 267–288

work page 1996
[53]

Ramasso, Investigating computational geometry for failure prognostics, International Journal of Prognostics and Health Management 5 (2014) 1–18

E. Ramasso, Investigating computational geometry for failure prognostics, International Journal of Prognostics and Health Management 5 (2014) 1–18

work page 2014
[54]

Bring, How to standardize regression coeﬃcients, The American Statistician 48 (1994) 209–213

J. Bring, How to standardize regression coeﬃcients, The American Statistician 48 (1994) 209–213

work page 1994
[55]

Evermann, J.-R

J. Evermann, J.-R. Rehse, P. Fettke, Predicting process behaviour using deep learning, Decision Support Systems 100 (2017) 129–140

work page 2017
[56]

Kraus, S

M. Kraus, S. Feuerriegel, Decision support from ﬁnancial disclosures with deep neural networks and transfer learning, Decision Support Systems 104 (2017) 38–48

work page 2017
[57]

Mahmoudi, P

N. Mahmoudi, P. Docherty, P. Moscato, Deep neural networks understand investors better, Decision Support Systems 112 (2018) 23–34

work page 2018
[58]

Rabatel, S

J. Rabatel, S. Bringay, P. Poncelet, Anomaly detection in monitoring sensor data for preventive mainte- nance, Expert Systems with Applications 38 (2011) 7003–7015

work page 2011
[59]

Y. Gal, Z. Ghahramani, Dropout as a bayesian approximation: Representing model uncertainty in deep learning, in: International Conference on International Conference on Machine Learning, 2016, pp. 1050– 1059. 33

work page 2016

[1] [1]

Heidergott, T

B. Heidergott, T. Farenhorst-Yuan, Gradient estimation for multicomponent maintenance systems with age-replacement policy, Operations Research 58 (2010) 706–718

work page 2010

[2] [2]

Groenevelt, L

H. Groenevelt, L. Pintelon, A. Seidmann, Production lot sizing with machine breakdowns, Management Science 38 (1992) 104–123

work page 1992

[3] [3]

Dogramaci, N

A. Dogramaci, N. M. Fraiman, Replacement decisions with maintenance under uncertainty: An imbedded optimal control model, Operations Research 52 (2004) 785–794

work page 2004

[4] [4]

H. L. Lee, M. J. Rosenblatt, Simultaneous determination of production cycle and inspection schedules in a production system, Management Science 33 (1987) 1125–1136

work page 1987

[5] [5]

A. K. Jardine, D. Lin, D. Banjevic, A review on machinery diagnostics and prognostics implementing condition-based maintenance, Mechanical Systems and Signal Processing 20 (2006) 1483–1510

work page 2006

[6] [6]

A. Heng, S. Zhang, A. C. Tan, J. Mathew, Rotating machinery prognostics: State of the art, challenges and opportunities, Mechanical Systems and Signal Processing 23 (2009) 724–739

work page 2009

[7] [7]

X.-S. Si, W. Wang, C.-H. Hu, D.-H. Zhou, Remaining useful life estimation: A review on the statistical data driven approaches, European Journal of Operational Research 213 (2011) 1–14

work page 2011

[8] [8]

Baptista, S

M. Baptista, S. Sankararaman, I. P. de Medeiros, C. Nascimento, H. Prendinger, E. M. Henriques, Forecasting fault events for predictive maintenance using data-driven techniques and ARMA modeling, Computers & Industrial Engineering 115 (2018) 41–53

work page 2018

[9] [9]

Seera, C

M. Seera, C. P. Lim, S. Nahavandi, C. K. Loo, Condition monitoring of induction motors: A review and an application of an ensemble of hybrid intelligent models, Expert Systems with Applications 41 (2014) 4891–4903

work page 2014

[10] [10]

Breiman, Statistical modeling: The two cultures, Statistical Science 16 (2001) 199–231

L. Breiman, Statistical modeling: The two cultures, Statistical Science 16 (2001) 199–231

work page 2001

[11] [11]

Jang, A decision support framework for robust R & D budget allocation using machine learning and optimization, Decision Support Systems 121 (2019) 1–12

H. Jang, A decision support framework for robust R & D budget allocation using machine learning and optimization, Decision Support Systems 121 (2019) 1–12

work page 2019

[12] [12]

H. S. Subramania, V. R. Khare, Pattern classiﬁcation driven enhancements for human-in-the-loop decision support systems, Decision Support Systems 50 (2011) 460–468

work page 2011

[13] [13]

Delen, H

D. Delen, H. Zaim, C. Kuzey, S. Zaim, A comparative analysis of machine learning systems for measuring the impact of knowledge management practices, Decision Support Systems 54 (2013) 1150–1160

work page 2013

[14] [14]

K. L. Reifsnider, S. W. Case, Damage tolerance and durability of material systems, Wiley Interscience, New York, NY, 2002

work page 2002

[15] [15]

Y. Liu, B. Stratman, S. Mahadevan, Fatigue crack initiation life prediction of railroad wheels, Interna- tional Journal of Fatigue 28 (2006) 747–756. 30

work page 2006

[16] [16]

Papakostas, P

N. Papakostas, P. Papachatzakis, V. Xanthakis, D. Mourtzis, G. Chryssolouris, An approach to opera- tional aircraft maintenance planning, Decision Support Systems 48 (2010) 604–612

work page 2010

[17] [17]

Z. C. Lipton, The mythos of model interpretability, Communications of the ACM 61 (2018) 36–43

work page 2018

[18] [18]

Rudin, Please stop explaining black box models for high stakes decisions, in: Conference on Neural Information Processing Systems, 2018

C. Rudin, Please stop explaining black box models for high stakes decisions, in: Conference on Neural Information Processing Systems, 2018

work page 2018

[19] [19]

Y. Lou, R. Caruana, J. Gehrke, G. Hooker, Accurate intelligible models with pairwise interactions, in: SIGKDD International Conference on Knowledge Discovery and Data Mining, 2013, pp. 623–631

work page 2013

[20] [20]

Saxena, K

A. Saxena, K. Goebel, C-mapss data set, NASA Ames Prognostics Data Repository (2008)

work page 2008

[21] [21]

J. B. Butcher, D. Verstraeten, B. Schrauwen, C. R. Day, P. W. Haycock, Reservoir computing and extreme learning machines for non-linear time-series data analysis, Neural networks 38 (2013) 76–89

work page 2013

[22] [22]

Dong, X.-Y

D. Dong, X.-Y. Li, F.-Q. Sun, Life prediction of jet engines based on LSTM-recurrent neural networks, in: Prognostics and System Health Management Conference, IEEE, 2017

work page 2017

[23] [23]

L. Liao, F. Kottig, Review of hybrid prognostics approaches for remaining useful life prediction of engineered systems, and an application to battery life prediction, IEEE Transactions on Reliability 63 (2014) 191–207

work page 2014

[24] [24]

Navarro, T

J. Navarro, T. Rychlik, Comparisons and bounds for expected lifetimes of reliability systems, European Journal of Operational Research 207 (2010) 309–317

work page 2010

[25] [25]

M. Dong, D. He, Hidden semi-Markov model-based methodology for multi-sensor equipment health diagnosis and prognosis, European Journal of Operational Research 178 (2007) 858–878

work page 2007

[26] [26]

M. J. Kim, V. Makis, Joint optimization of sampling and control of partially observable failing systems, Operations Research 61 (2013) 777–790

work page 2013

[27] [27]

Plitsos, P

S. Plitsos, P. P. Repoussis, I. Mourtos, C. D. Tarantilis, Energy-aware decision support for production scheduling, Decision Support Systems 93 (2017) 88–97

work page 2017

[28] [28]

Ghosh, R

D. Ghosh, R. Sharman, H. Raghav Rao, S. Upadhyaya, Self-healing systems: survey and synthesis, Decision Support Systems 42 (2007) 2164–2185

work page 2007

[29] [29]

Mazhar, S

M. Mazhar, S. Kara, H. Kabernick, Remaining life estimation of used components in consumer products: Life cycle data analysis by Weibull and artiﬁcial neural networks, Journal of Operations Management 25 (2007) 1184–1193

work page 2007

[30] [30]

Cox, Regression models and life-tables, Journal of the Royal Statistical Society 34 (1972) 187–220

work page 1972

[31] [31]

Y. Wang, S. Wang, Y. Fang, P. Y. Chau, Store survival in online marketplace: An empirical investigation, Decision Support Systems 56 (2013) 482–493

work page 2013

[32] [32]

Zihajehzadeh, E

S. Zihajehzadeh, E. J. Park, Regression model-based walking speed estimation using wrist-worn inertial 31 sensor, PLOS ONE 11 (2016) e0165211

work page 2016

[33] [33]

A. Riad, H. Elminir, H. Elattar, Evaluation of neural networks in the subject of prognostics as compared to linear regression model, International Journal of Engineering & Technology 10 (2010) 52–58

work page 2010

[34] [34]

Mosallam, K

A. Mosallam, K. Medjaher, N. Zerhouni, Nonparametric time series modelling for industrial prognostics and health management, The International Journal of Advanced Manufacturing Technology 69 (2013) 1685–1699

work page 2013

[35] [35]

G. S. Babu, P. Zhao, X.-L. Li, Deep convolutional neural network based regression approach for estimation of remaining useful life, in: Database Systems for Advanced Applications, volume 9642, Springer, 2016, pp. 214–228

work page 2016

[36] [36]

Y. Wu, M. Yuan, S. Dong, L. Lin, Y. Liu, Remaining useful life estimation of engineered systems using vanilla LSTM neural networks, Neurocomputing 275 (2018) 167–179

work page 2018

[37] [37]

Zheng, K

S. Zheng, K. Ristovski, A. Farahat, C. Gupta, Long short-term memory network for remaining useful life estimation, in: IEEE International Conference on Prognostics and Health Management, IEEE, 2017, pp. 88–95

work page 2017

[38] [38]

Verification for Machine Learning, Autonomy, and Neural Networks Survey

W. Xiang, P. Musau, A. A. Wild, D. M. Lopez, N. Hamilton, X. Yang, J. Rosenfeld, T. T. Johnson, Veriﬁcation for machine learning, autonomy, and neural networks survey, arXiv preprint arXiv:1810.01989 (2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018

[39] [39]

Athalye, N

A. Athalye, N. Carlini, D. Wagner, Obfuscated gradients give a false sense of security: Circumventing defenses to adversarial examples, in: International Conference on Machine Learning, 2018

work page 2018

[40] [40]

Singh, T

G. Singh, T. Gehr, M. Mirman, M. P¨ uschel, M. Vechev, Fast and eﬀective robustness certiﬁcation, in: Conference on Neural Information Processing Systems, 2018, pp. 10802–10813

work page 2018

[41] [41]

M. T. Ribeiro, S. Singh, C. Guestrin, Why should i trust you? explaining the predictions of any classiﬁer, in: SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 1135–1144

work page 2016

[42] [42]

S. M. Lundberg, S.-I. Lee, A uniﬁed approach to interpreting model predictions, in: Advances in Neural Information Processing Systems, 2017, pp. 4765–4774

work page 2017

[43] [43]

van der Maaten, G

L. van der Maaten, G. Hinton, Visualizing data using t-SNE, Journal of Machine Learning Research 9 (2008) 2579–2605

work page 2008

[44] [44]

W. J. Murdoch, C. Singh, K. Kumbier, R. Abbasi-Asl, B. Yu, Interpretable machine learning: Deﬁnitions, methods, and applications, arXiv preprint (2019)

work page 2019

[45] [45]

P. J. Vlok, J. L. Coetzee, D. Banjevic, A. K. S. Jardine, V. Makis, Optimal component replacement decisions using vibration monitoring and the proportional-hazards model, Journal of the Operational Research Society 53 (2002) 193–202. 32

work page 2002

[46] [46]

Hastie, R

T. Hastie, R. Tibshirani, J. H. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed., Springer, New York, NY, 2009

work page 2009

[47] [47]

Goodfellow, Y

I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, MIT Press, Cambridge, MA, 2017

work page 2017

[48] [48]

Hochreiter, J

S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural Computation 9 (1997) 1735–1780

work page 1997

[49] [49]

Fischer, C

T. Fischer, C. Krauss, Deep learning with long short-term memory networks for ﬁnancial market predic- tions, European Journal of Operational Research 270 (2018) 654–669

work page 2018

[50] [50]

Kraus, S

M. Kraus, S. Feuerriegel, A. Oztekin, Deep learning in business analytics and operations research: Models, applications and managerial implications, arXiv preprint (2018)

work page 2018

[51] [51]

Srivastava, S

S. Srivastava, S. Lessmann, A comparative study of LSTM neural networks in forecasting day-ahead global horizontal irradiance with satellite data, Solar Energy 162 (2018) 232–247

work page 2018

[52] [52]

Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society

R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society. Series B (Methodological) 58 (1996) 267–288

work page 1996

[53] [53]

Ramasso, Investigating computational geometry for failure prognostics, International Journal of Prognostics and Health Management 5 (2014) 1–18

E. Ramasso, Investigating computational geometry for failure prognostics, International Journal of Prognostics and Health Management 5 (2014) 1–18

work page 2014

[54] [54]

Bring, How to standardize regression coeﬃcients, The American Statistician 48 (1994) 209–213

J. Bring, How to standardize regression coeﬃcients, The American Statistician 48 (1994) 209–213

work page 1994

[55] [55]

Evermann, J.-R

J. Evermann, J.-R. Rehse, P. Fettke, Predicting process behaviour using deep learning, Decision Support Systems 100 (2017) 129–140

work page 2017

[56] [56]

Kraus, S

M. Kraus, S. Feuerriegel, Decision support from ﬁnancial disclosures with deep neural networks and transfer learning, Decision Support Systems 104 (2017) 38–48

work page 2017

[57] [57]

Mahmoudi, P

N. Mahmoudi, P. Docherty, P. Moscato, Deep neural networks understand investors better, Decision Support Systems 112 (2018) 23–34

work page 2018

[58] [58]

Rabatel, S

J. Rabatel, S. Bringay, P. Poncelet, Anomaly detection in monitoring sensor data for preventive mainte- nance, Expert Systems with Applications 38 (2011) 7003–7015

work page 2011

[59] [59]

Y. Gal, Z. Ghahramani, Dropout as a bayesian approximation: Representing model uncertainty in deep learning, in: International Conference on International Conference on Machine Learning, 2016, pp. 1050– 1059. 33

work page 2016