Pith. sign in

REVIEW 3 major objections 5 minor 77 references

A scalable algorithm for identifying multiple sensor faults using disentangled RNNs

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A disentangled RNN with a covariance penalty and a greedy search replaces combinatorial sensor fault isolation with a linear-time procedure.

desk verdict A useful industrial idea buried under a self-referential evaluation: the disentangled RNN and greedy isolation heuristic are plausible, but GreedyIso as written depends on a probability it declares unavailable. read the letter →

arxiv 1909.02449 v1 pith:YPGGRDIW submitted 2019-09-04 eess.SP

David Haldimann , Marco Guerriero , Yannick Maret , Nunzio Bonavita , Gregorio Ciarlo , Marta Sabbadin This is my paper · ORCID
classification eess.SP
keywords sensorfaultdetectionisolationrecurrentneuralnetworksdisentanglementsmearing-outeffectgreedyalgorithmanalyticalredundancymultiplefaults
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that two ideas together solve a scalability bottleneck in sensor fault isolation. The first is a training-time covariance penalty on the predictions of a recurrent neural network, which stops a faulty sensor's reading from contaminating the model's predictions for healthy sensors; that contamination is the smearing-out effect that misleads classic contribution plots. The second is a greedy algorithm, GreedyIso, that builds a list of faulty sensors one at a time, correcting each candidate's readings and keeping the correction only when the probabilistic detection score and the mean residual both drop. The claim is that this identifies multiple simultaneous faults in worst-case time linear in the number of sensors, with no prior knowledge of how many sensors are faulty. The paper validates the architecture on a real 8-sensor petrochemical plant dataset with injected bias faults and reports that the disentangled model isolates faults at smaller offsets than an unregularized model.

What carries the argument

The central object is the covariance-penalized prediction model, with total loss $L_{\mathrm{tot}} = L_{\mathrm{MSE}} + \lambda L_C$, where $L_C = \frac{1}{S^2}\|C\|_1$ forces the RNN's per-sensor predictions to be weakly cross-correlated, reducing the paths by which one sensor's fault leaks into predictions of healthy sensors. GreedyIso then converts the disentangled residuals into a fault list: contribution scores order the candidates, bias correction removes each candidate's estimated offset, and the decrease of $P_D$ together with the mean residual norm decides whether the correction is real. A targeted variant, $L_{C_s} = \frac{1}{S}\|C_s\|_1$, penalizes only one sensor's covariance row and raises sensitivity for that sensor at the cost of others.

What would settle it

Run GreedyIso on the 8-sensor petrochemical dataset with two injected faults whose offsets fall in the reported [5%, 30%] range, and log the accept/reject decision and the computed $P_D$ at every step. Because $P_D$ under the fault hypothesis is not computable from the paper's assumptions, any implementation must estimate it; if different reasonable estimates change the output fault list, the algorithm's isolation result is not well defined. A sharper test is the sensor-0 case: the paper reports that offsets in that range are undetectable for sensor 0, so a correctly working isolation algorithm must either expose that limitation through the residual-decrease test or fail in a way that reveals the missing $P_D$.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a disentangled RNN removes enough cross-sensor coupling to make per-sensor residuals trustworthy, and that a greedy search over those residuals can replace the combinatorial maximum-likelihood search over all $2^S$ fault subsets. The regularizer adds $(1/S^2)\|C\|_1$ to the mean-squared-error loss, with $C$ the covariance matrix of the network's predictions across sensors, so the network is discouraged from representing one sensor's behavior in another sensor's output. For isolation, GreedyIso selects the sensor with the largest contribution score, estimates its bias as the time-averaged prediction error, subtracts that bias from its measurements, and accepts the sensor as faulty only if the resulting probability of detection $P_D$ and the mean residual norm do not increase; it stops when $P_D$ reaches zero. The paper demonstrates that a two-stage architecture, with an unregularized model for detection and a disentangled model for isolation, gives faster detection than a single regularized model while improving single-fault accuracy and multiple-fault mIoU.

Load-bearing premise

The algorithm's stopping and acceptance rules both require the probability of detection $P_D$, which is defined from the residual density under the fault hypothesis; the paper explicitly says that density is unknown and cannot be learned because faulty sensor data are not available.

Editorial extensions

If this is right

  • Fault isolation in large sensor networks stops being exponential: with $S$ sensors, GreedyIso's worst-case work is $O(S)$ instead of the combinatorial $O(2^S)$ maximum-likelihood search, making per-sensor monitoring feasible at plant scale.
  • The two-stage architecture decouples detection from isolation, so a plant can keep a high-sensitivity unregularized detector and still gain the isolation benefits of disentanglement.
  • Because the regularizer acts on the prediction layer rather than on a specific architecture, the same smearing-out reduction should transfer to other learned predictors, including feedforward models, if their predictions are used as residuals.
  • GreedyIsoSparse inherits the same greedy acceptance rule and stays stable across a wide range of the sparsity parameter $\eta$, degrading only under very strong regularization.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A practical deployment would need an operational proxy for $P_D$, since the paper states the residual density under the fault hypothesis is unknown; one testable extension is to estimate $P_D$ empirically from residuals after correcting candidate faults and check whether GreedyIso's accept/reject decisions change.
  • The accept-if-residual-drops rule does not depend on the network being recurrent; applying the same covariance penalty and greedy search to PCA- or autoencoder-based analytical redundancy could give similar smearing-out suppression at lower computational cost.
  • The paper's sensor-0 result suggests a natural stress test: vary the offset level for every sensor separately and plot the minimum detectable offset for isolation, which would tell operators which sensors cannot be trusted in low-offset regimes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper addresses sensor fault detection and isolation (SFD-SFI) using a data-driven analytical-redundancy approach. The authors propose a "disentangled" RNN that penalizes covariance among the predicted sensor outputs, with the goal of mitigating the smearing-out effect in residual-based contribution scores. They also introduce a probabilistic residual model under the fault-free hypothesis H0, a K-out-of-M decision-fusion detector for SFD, and a greedy isolation algorithm called GreedyIso (plus a sparse variant, GreedyIsoSparse) that is claimed to identify multiple faulty sensors in O(S) time without knowing the number of faults. The method is evaluated on a real petrochemical-plant data set with injected biases of varying magnitude.

Significance. If the algorithmic gap were closed, the paper would offer a practical and potentially generalizable idea: a covariance-penalty regularizer against smearing-out, and a linear-time alternative to the combinatorial SFI search. The writing is generally clear, and the authors are explicit about their working hypothesis that only fault-free training data are available. The use of a real industrial data set is a strength, as is the explicit acknowledgment that sensor 0 faults are undetectable in the tested range. However, the central contribution, GreedyIso, depends on a quantity that the paper itself declares unavailable and unlearnable, and the numerical evaluation does not clarify how that quantity is obtained. As a result, the core algorithmic claim is not operational as stated, and the empirical evidence cannot be interpreted as validating the proposed unsupervised pipeline.

major comments (3)
  1. [Section V, Algorithm 1 (lines 1, 3, 13-15)] The control flow of GreedyIso depends on PD computed from Eq. (15), which in turn requires pd from Eq. (9) and therefore the pdf f(Rt|H1). Section III states that "The pdf f(Rt|H1) is obviously unknown (and it cannot be learned from data since sensors faulty data are not available)." This makes the acceptance criterion (PD,new <= PD), the rejection criterion, and the stopping condition (PD > 0) undefined in the very setting the paper targets, namely unsupervised learning from fault-free training data. This is a load-bearing inconsistency, not a missing implementation detail.
  2. [Section VII (numerical evaluation)] The experiments never state how PD is instantiated in GreedyIso. In the only possible source of H1 samples, faults are artificially injected into fault-free data, so PD can be estimated empirically from the injected faults. If that is indeed what the experiments do, then the algorithm is run with access to the fault distribution that the method declares unavailable in deployment, making the evaluation circular. The reported accuracy and mIoU therefore do not demonstrate that GreedyIso can operate under the paper's stated working hypothesis.
  3. [Section V, Algorithm 1 line 3 and surrounding text] The pseudocode says "while PD > 0 OR s <= S," while the text says the loop repeats "until either PD = 0 or all sensors have been processed at least once." With the OR operator, the loop always runs through all S sensors as long as s <= S, so a PD-based early stop never triggers; with the intended AND operator, the loop still depends on the unavailable PD. Either way, the pseudocode and the prose are inconsistent, and the claimed early-termination behavior is not what is actually specified.
minor comments (5)
  1. [Section II-B, Eq. (4)] Equation (4) contains an apparent typo: "||xt - hat x2 t||2" should be "||xt - hat xt||2." Also, "stochastic gradient decent" should be "stochastic gradient descent."
  2. [Section IV, Eq. (17)] The notation r_s^n in Eq. (17) is not defined; it appears to mean the residual of sensor s at some time index, but the superscript should be clarified (e.g., r_s,t or r_s,n).
  3. [Section IV, first paragraph] The text says "by exploding the at least one sensor fault hypothesis" but the intended word is likely "exploiting." Also, "miss-classified" should be "misclassified."
  4. [Section III and Table I] The notation pfa and Pfa is used with two different meanings (per-sample false-alarm probability and fused system-level false-alarm probability) but the distinction is not consistently maintained; the remark at the end of Section III conflates them. Please define both symbols once and use them consistently.
  5. [Section VII-D/E and Table II] The reported raw metrics (83% ACC, 83.3% mIoU) are renormalized to "approximately 95%" by dividing by the best attainable value of 87.5%, which excludes the undetectable sensor 0 faults. The renormalization is disclosed, but the raw numbers should be the primary reported comparison so readers can judge performance on all injected faults.

Circularity Check

1 steps flagged · score 6.0 of 10

GreedyIso's stopping and acceptance rules depend on PD, which the paper itself declares unlearnable; in experiments PD is estimated from injected faults, making the evaluation partially self-referential.

  1. fitted input called prediction [Section III (eq. 9 and following paragraph); Algorithm 1 lines 1, 3, 13-15; Section VII-B/E injection setup]
    "The pdf f(Rt|H1) is obviously unknown (and it cannot be learned from data since sensors faulty data are not available). Nevertheless, equation (9) would become very useful for: ... guiding in a principled way the greedy algorithm described in Section V."

    Algorithm 1 line 1 requires PD from eq. (15), line 3 stops only when 'PD > 0' fails, and lines 13-15 accept a candidate only if 'PD,new≤PD'. Eq. (15) is built from pd in eq. (9), which needs f(Rt|H1) — the very pdf the paper declares unknown and unlearnable. The only way the algorithm is run in Section VII is after injecting faults into faultless data, so f(Rt|H1) and PD are estimated from the same injected-fault ground truth the isolation procedure is supposed to recover. The claimed principled guide is thus an input fitted from the target fault distribution, not a first-principles prediction.

full rationale

The disentanglement contribution (covariance regularization, eqs. 19-22) is a genuine architectural modification and is evaluated against an external, self-contained comparison; no self-citation or definitional circularity is involved there. The circularity is concentrated in GreedyIso: its control flow is governed by PD, and PD is not computable under the paper's stated unsupervised assumption (only H0 training data). In the experimental section, H1 samples are manufactured by injecting biases into the data, so PD can be obtained only by using the fault distribution that constitutes the ground truth for the SFI task. This makes the isolation algorithm's acceptance and stopping criteria depend on labeled fault knowledge, partially reducing the claimed 'probabilistic model guides SFI' to a fitted input. The O(S) complexity and the disentangled-RNN results remain independent, but the central SFI algorithm as specified is not executable in deployment and its evaluation is self-referential.

Assumptions & free parameters 7 free parameters · 7 assumptions · 0 invented entities

The central method rests on a small number of hand-chosen hyperparameters and several unproven heuristics. The most serious is the assumed availability of PD under the fault hypothesis, which the paper itself states is unknown.

free parameters (7)
  • λ (disentanglement regularization weight) = 0.01
    Chosen by hand; balances prediction loss (LMSE) against output covariance penalty (LC) in Eq. (22).
  • γ (detection threshold) or pfa = pfa = 0.01
    Set by the user to achieve a desired false-alarm rate; defines the decision threshold in Eq. (8).
  • M (decision fusion window size) = 60
    Batch size for the K-out-of-M decision rule (Eq. 10); set in Table I.
  • L (SFI integration interval) = 60
    Length of the window used to estimate the bias vector and contribution scores; set in Table I.
  • η (GreedyIsoSparse regularization) = varied 0, 0.01, 1, 10, 100, 500, 1000
    Regularization strength in Eq. (25); the paper shows performance declines for large η but does not give a selection rule.
  • K (K-out-of-M threshold) = not stated in table
    Part of the decision rule in Eq. (10); the paper discusses choosing K via a randomized test but does not list the value used.
  • PD (probability of detection used in GreedyIso) = empirical, from injected faults in experiments
    Algorithm 1 requires PD from eq. (15), but the H1 distribution is unknown; in the experiments it must be estimated from the same injected faults used for evaluation.
assumptions (7)
  • domain assumption Faults are additive biases: xt~ = xt + Δ (Eq. 1).
    Used throughout; restricts the method to bias-type faults and excludes drifts or multiplicative faults.
  • domain assumption Training data are fault-free and cover the operating points seen at test time.
    Stated in Section III; violations cause false alarms on novel but healthy operating points.
  • standard math The K-out-of-M decision statistic is approximately Binomial (Eq. 12).
    The paper acknowledges residual dependence over time makes this an approximation.
  • ad hoc to paper The sensor with the largest contribution score is a faulty sensor (Algorithm 1 line 5).
    Greedy selection is justified only by the toy example in Fig. 3, not by a formal statement.
  • ad hoc to paper A candidate fault is accepted only if corrected data lower PD and residual norm (Algorithm 1 lines 15-16).
    No proof that this criterion finds the true fault set; it is a heuristic with possible local minima.
  • ad hoc to paper PD under H1 is available to the algorithm (Algorithm 1 line 1).
    The paper explicitly says f(Rt|H1) is unknown and cannot be learned from data, yet GreedyIso requires PD from eq. (15).
  • ad hoc to paper Penalizing the covariance of RNN outputs reduces smearing-out (Eq. 22).
    Borrowed from disentangled activations [76]; no theoretical guarantee that it transfers to residual-based fault isolation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A scalable algorithm for identifying multiple sensor faults using disentangled RNNs." pith.science (2026). https://pith.science/paper/YPGGRDIW

@misc{pith2026190902449,
  author       = {Pith},
  title        = {Pith review of: A scalable algorithm for identifying multiple sensor faults using disentangled RNNs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YPGGRDIW}},
  note         = {Machine review of arXiv:1909.02449}
}
read the original abstract

The problem of detecting and identifying sensor faults is critical for efficient, safe, regulatory-compliant and sustainable operations of modern systems. Their increasing complexity brings new challenges for the Sensor Fault Detection and Isolation (SFD-SFI) tasks. One of the key enablers for any SFD-SFI methods employed in modern complex sensor systems, is the so-called analytical redundancy, which is nothing but building an analytical model of the sensors observations (either derived from first principles or identified from historical data in a data-driven fashion). In a nutshell, SFD amounts to generate and to monitor residuals by comparing the sensor measurements with the model predictions with the idea that the faulty sensors will result in large residuals (i.e. the defective sensors generate measurement that are inconsistent with their expected behavior represented by the model). In this paper we introduce a disentangled Recurrent Neural Network (RNN) with the objective to cope with the \textit{smearing-out} effect, i.e. the propagation of a sensor fault to the non-faulty sensors resulting in large misleading residuals. Moreover, the introduction of a probabilistic model for the residual generation allows us to develop a novel procedure for the identification of the faulty sensors. The computational complexity of the proposed algorithm is linear in the number of sensors as opposed to the combinatorial nature of the SFI problem. Finally, we empirically verify the performances of the proposed SFD-SFI architecture using a real data set collected at a petrochemical plant.

Figures

Figures reproduced from arXiv: 1909.02449 by the authors.

Figure 1
Figure 1. Pictorial illustration of both the RNN and FFNN architectures deployed to generate the residual [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The SFD system processes a batch of size [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 2
Figure 2. The SFI problem can be formalized as the following multi￾hypothesis testing problem [9] by exploding the at least one sensor fault hypothesis H1 into multiple S = 2S − 1 hypotheses: H11 : r 1 t ≈ large, r2 t ≈ small, . . . , rS t ≈ small H12 : r 1 t ≈ small, r2 t ≈ large, . . . , rS t ≈ small . . . H1S : r 1 t ≈ large, r2 t ≈ large, . . . , rS t ≈ large for t = tI , . . . , tI + L (16) Clearly the hypothesis H11 ref… view at source ↗
Figures from the paper (4 more)
Figure 3
Figure 3. Figure 3: Comparing contribution plots in order to study the impact of disentanglement in challenging situations where the [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 4
Figure 4. Figure 4: Comparison diagram for one-stage vs two-stage [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Probability of detection PD as function of the offset level β for the two models, which are the RNN and the FFNN. The faster the detection rate PD goes to 1, the better the model is. faultless training data set according to the model as described in Equation (1). All m…
Figure 7
Figure 7. Figure 7: Probability of detection PD as function of the offset level β for the the one-stage and two-stage architectures. The faster the detection rate PD goes to 1, the better the model is. a random offset, which is uniformly distributed in the range [5%, 30%], is added into o…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

77 extracted references · 74 canonical work pages

  1. [36]

    Estimation of faults in dc electrical power system

    Dimitry Gorinevsky, Stephen Boyd, and Scott Poll. “Estimation of faults in dc electrical power system”. In: 2009 American Control Conference . IEEE. 2009, pp. 4334–4339

    work page 2009

  2. [39]

    Relaxed maximum a posteriori fault identification

    Argyrios Zymnis, Stephen Boyd, and Dimitry Gorinevsky. “Relaxed maximum a posteriori fault identification”. In: Signal Processing 89.6 (2009), pp. 989–999

    work page 2009

  3. [1]

    Distribution de la flore alpine dans le bassin des Dranses et dans quelques r ´egions voisines

    Paul Jaccard. “Distribution de la flore alpine dans le bassin des Dranses et dans quelques r ´egions voisines”. In: Bulletin de la Soci ´et´e Vaudoise des Sciences Na- turelles 37 (1901), pp. 241–272

    work page 1901

  4. [2]

    Distance between sets

    Michael Levandowsky and David Winter. “Distance between sets”. In: Nature 234.5323 (1971), p. 34

    work page 1971

  5. [3]

    Analytical redundancy and the design of robust failure detection systems

    Eyey Chow and Alan Willsky. “Analytical redundancy and the design of robust failure detection systems”. In: IEEE Transactions on automatic control 29.7 (1984), pp. 603–614

    work page 1984

  6. [4]

    Finding structure in time

    Jeffrey L Elman. “Finding structure in time”. In: Cog- nitive science 14.2 (1990), pp. 179–211

    work page 1990

  7. [5]

    Handbook of theoretical computer science

    Jan Van Leeuwen and Jan Leeuwen. Handbook of theoretical computer science . V ol. 1. Elsevier, 1990

    work page 1990

  8. [6]

    Analytical redundancy methods in fault detection and isolation-survey and synthesis

    Janos Gertler. “Analytical redundancy methods in fault detection and isolation-survey and synthesis”. In: IFAC Proceedings Volumes 24.6 (1991), pp. 9–21

    work page 1991

Show all 77 references
  1. [7]

    Learning complex, extended se- quences using the principle of history compression

    J ¨urgen Schmidhuber. “Learning complex, extended se- quences using the principle of history compression”. In: Neural Computation 4.2 (1992), pp. 234–242

    1992

  2. [8]

    Detection of abrupt changes: theory and application

    Mich `ele Basseville, Igor V Nikiforov, et al. Detection of abrupt changes: theory and application . V ol. 104. Prentice Hall Englewood Cliffs, 1993

    1993

  3. [9]

    Fundamentals of statistical signal pro- cessing

    Steven M Kay. Fundamentals of statistical signal pro- cessing. Prentice Hall PTR, 1993

    1993

  4. [10]

    Integration of fault detection and di- agnosis methods

    Rolf Isermann. “Integration of fault detection and di- agnosis methods”. In: IFAC Proceedings Volumes 27.5 (1994), pp. 575–590

    1994

  5. [11]

    Kernel smoothing

    Matt P Wand and M Chris Jones. Kernel smoothing . Chapman and Hall/CRC, 1994. REFERENCES 13

    1994

  6. [12]

    Optimal simulta- neous detection and estimation under a false alarm con- straint

    Bulent Baygun and Alfred O Hero. “Optimal simulta- neous detection and estimation under a false alarm con- straint”. In: IEEE Transactions on Information Theory 41.3 (1995), pp. 688–703

    1995

  7. [13]

    Phoneme recognition using time-delay neural networks

    Alexander Waibel, Toshiyuki Hanazawa, Geoffrey Hin- ton, Kiyohiro Shikano, and Kevin J Lang. “Phoneme recognition using time-delay neural networks”. In: Backpropagation: Theory, Architectures and Applica- tions (1995), pp. 35–61

    1995

  8. [14]

    Identification of faulty sensors using principal component analysis

    Ricardo Dunia, S Joe Qin, Thomas F Edgar, and Thomas J McAvoy. “Identification of faulty sensors using principal component analysis”. In: AIChE Journal 42.10 (1996), pp. 2797–2812

    1996

  9. [15]

    Long short- term memory

    Sepp Hochreiter and J ¨urgen Schmidhuber. “Long short- term memory”. In: Neural computation 9.8 (1997), pp. 1735–1780

    1997

  10. [16]

    Trends in the appli- cation of model-based fault detection and diagnosis of technical processes

    Rolf Isermann and Peter Balle. “Trends in the appli- cation of model-based fault detection and diagnosis of technical processes”. In: Control engineering practice 5.5 (1997), pp. 709–719

    1997

  11. [17]

    Serial order: A parallel distributed processing approach

    Michael I Jordan. “Serial order: A parallel distributed processing approach”. In: Advances in psychology . V ol. 121. Elsevier, 1997, pp. 471–495

    1997

  12. [18]

    A neural- network approach to fault detection and diagnosis in industrial processes

    Yunosuke Maki and Kenneth A Loparo. “A neural- network approach to fault detection and diagnosis in industrial processes”. In: IEEE Transactions on Control Systems Technology 5.6 (1997), pp. 529–541

    1997

  13. [19]

    Robust detection and isolation of process faults using neural networks

    Teodor Marcu and Letitia Mirea. “Robust detection and isolation of process faults using neural networks”. In: IEEE Control Systems Magazine 17.5 (1997), pp. 72– 79

    1997

  14. [20]

    Distributed detection with multiple sensors: Part I-fundamentals

    Ramanarayanan Viswanathan, Pramod K Varshney, et al. “Distributed detection with multiple sensors: Part I-fundamentals”. In: Proceedings of the IEEE 85.1 (1997), pp. 54–63

    1997

  15. [21]

    Neural network based fault detection in robotic manipulators

    Arun T Vemuri, Marios M Polycarpou, and Sotiris A Diakourtis. “Neural network based fault detection in robotic manipulators”. In: IEEE Transactions on Robotics and Automation 14.2 (1998), pp. 342–348

    1998

  16. [22]

    Fault diagnosis in chemical processes using Fisher discriminant analysis, discriminant partial least squares, and principal component analysis

    Leo H Chiang, Evan L Russell, and Richard D Braatz. “Fault diagnosis in chemical processes using Fisher discriminant analysis, discriminant partial least squares, and principal component analysis”. In: Chemometrics and intelligent laboratory systems 50.2 (2000), pp. 243– 252

    2000

  17. [23]

    Continuous emission monitoring

    James A Jahnke. Continuous emission monitoring. John Wiley & Sons, 2000

    2000

  18. [24]

    Di- agnosis techniques for sensor faults of industrial pro- cesses

    Silvio Simani, Cesare Fantuzzi, and S Beghelli. “Di- agnosis techniques for sensor faults of industrial pro- cesses”. In: IEEE Transactions on Control Systems Technology 8 (Oct. 2000), pp. 848–855

    2000

  19. [25]

    Generalized contribution plots in multivariate statistical process monitoring

    Johan A Westerhuis, Stephen P Gurden, and Age K Smilde. “Generalized contribution plots in multivariate statistical process monitoring”. In: Chemometrics and intelligent laboratory systems 51.1 (2000), pp. 95–114

    2000

  20. [26]

    The elements of statistical learning

    Jerome Friedman, Trevor Hastie, and Robert Tibshirani. The elements of statistical learning. V ol. 1. 10. Springer series in statistics New York, 2001

    2001

  21. [27]

    Consistent dynamic PCA based on errors-in-variables subspace identification

    Weihua Li and S Joe Qin. “Consistent dynamic PCA based on errors-in-variables subspace identification”. In: Journal of Process Control 11.6 (2001), pp. 661–678

    2001

  22. [28]

    Proba- bility, random variables, and stochastic processes

    Athanasios Papoulis and S Unnikrishna Pillai. Proba- bility, random variables, and stochastic processes . Tata McGraw-Hill Education, 2002

    2002

  23. [29]

    Novelty Detection: A Review — Part 1: Statistical Approaches

    Markos Markou and Sameer Singh. “Novelty Detection: A Review — Part 1: Statistical Approaches”. In: Signal Process. 83.12 (Dec. 2003), pp. 2481–2497

    2003

  24. [30]

    Model-based fault-detection and diagnosis–status and applications

    Rolf Isermann. “Model-based fault-detection and diagnosis–status and applications”. In: Annual Reviews in control 29.1 (2005), pp. 71–85

    2005

  25. [31]

    An improved PCA scheme for sensor FDI: Application to an air quality monitoring network

    Mohamed-Faouzi Harkat, Gilles Mourot, and Jos ´e Ragot. “An improved PCA scheme for sensor FDI: Application to an air quality monitoring network”. In: Journal of Process Control 16.6 (2006), pp. 625–634

    2006

  26. [32]

    Soft sensors for monitoring and control of industrial processes

    Luigi Fortuna, Salvatore Graziani, Alessandro Rizzo, and Maria Gabriella Xibilia. Soft sensors for monitoring and control of industrial processes. Springer Science & Business Media, 2007

    2007

  27. [33]

    Application of computational intelligence for sensor fault detection and isolation

    Al Jabbari, R Jedermann, and W Lang. “Application of computational intelligence for sensor fault detection and isolation”. In: World academy of science, engineering and technology 33 (2007), pp. 265–270

    2007

  28. [34]

    Multiple faults diagnosis for sensors in air handling unit using Fisher discrimi- nant analysis

    Zhimin Du and Xinqiao Jin. “Multiple faults diagnosis for sensors in air handling unit using Fisher discrimi- nant analysis”. In: Energy Conversion and Management 49.12 (2008), pp. 3654–3665

    2008

  29. [35]

    Reconstruction-based contribution for process monitoring

    Carlos F Alcala and S Joe Qin. “Reconstruction-based contribution for process monitoring”. In: Automatica 45.7 (2009), pp. 1593–1600

    2009

  30. [37]

    Dis- tributed target detection in sensor networks using scan statistics

    Marco Guerriero, Peter Willett, and Joseph Glaz. “Dis- tributed target detection in sensor networks using scan statistics”. In: IEEE Transactions on Signal Processing 57.7 (2009), pp. 2629–2639

    2009

  31. [38]

    Sensor network data fault types

    Kevin Ni, Nithya Ramanathan, Mohamed Nabil Hajj Chehade, Laura Balzano, Sheela Nair, Sadaf Zahedi, Eddie Kohler, Greg Pottie, Mark Hansen, and Mani Srivastava. “Sensor network data fault types”. In: ACM Transactions on Sensor Networks (TOSN) 5.3 (2009), p. 25

    2009

  32. [40]

    Tracking and data fusion

    Yaakov Bar-Shalom, Peter K Willett, and Xin Tian. Tracking and data fusion . YBS publishing Storrs, CT, USA: 2011

    2011

  33. [41]

    Scan statistics and applications

    Joseph Glaz and Narayanaswamy Balakrishnan. Scan statistics and applications. Springer Science & Business Media, 2012

    2012

  34. [42]

    Signal detection in non-Gaussian noise

    Saleem A Kassam. Signal detection in non-Gaussian noise. Springer Science & Business Media, 2012. REFERENCES 14

    2012

  35. [43]

    Survey on data-driven industrial process monitoring and diagnosis

    S Joe Qin. “Survey on data-driven industrial process monitoring and diagnosis”. In: Annual reviews in con- trol 36.2 (2012), pp. 220–234

    2012

  36. [44]

    A knowledge-based system approach for sensor fault modeling, detection and mit- igation

    Jonny Carlos da Silva, Abhinav Saxena, Edward Bal- aban, and Kai Goebel. “A knowledge-based system approach for sensor fault modeling, detection and mit- igation”. In: Expert Systems with Applications 39.12 (2012), pp. 10977–10989

    2012

  37. [45]

    Distributed detection and data fusion

    Pramod K Varshney. Distributed detection and data fusion. Springer Science & Business Media, 2012

    2012

  38. [46]

    Representation learning: A review and new perspec- tives

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. “Representation learning: A review and new perspec- tives”. In: IEEE transactions on pattern analysis and machine intelligence 35.8 (2013), pp. 1798–1828

    2013

  39. [47]

    Bootstrap inference for network construction with an application to a breast cancer microarray study

    Shuang Li, Li Hsu, Jie Peng, and Pei Wang. “Bootstrap inference for network construction with an application to a breast cancer microarray study”. In: The annals of applied statistics 7.1 (2013), p. 391

    2013

  40. [48]

    An introduction to signal detection and estimation

    H Vincent Poor. An introduction to signal detection and estimation. Springer Science & Business Media, 2013

    2013

  41. [49]

    Adaptive approximation for multiple sen- sor fault detection and isolation of nonlinear uncertain systems

    Vasso Reppa, Marios M Polycarpou, and Christos G Panayiotou. “Adaptive approximation for multiple sen- sor fault detection and isolation of nonlinear uncertain systems”. In: IEEE Transactions on Neural Networks and Learning Systems 25.1 (2013), pp. 137–153

    2013

  42. [50]

    Sequential analysis: tests and confi- dence intervals

    David Siegmund. Sequential analysis: tests and confi- dence intervals . Springer Science & Business Media, 2013

    2013

  43. [51]

    Contribution plots for statistical process control: Analysis of the smearing-out effect

    Pieter Van den Kerkhof, Jef Vanlaer, Geert Gins, and Jan FM Van Impe. “Contribution plots for statistical process control: Analysis of the smearing-out effect”. In: 2013 European Control Conference (ECC) . IEEE. 2013, pp. 428–433

    2013

  44. [52]

    Sequential multi-sensor change-point detection

    Yao Xie and David Siegmund. “Sequential multi-sensor change-point detection”. In: The Annals of Statistics 41.2 (2013), pp. 670–692

    2013

  45. [53]

    A review on multi-label learning algorithms

    Min-Ling Zhang and Zhi-Hua Zhou. “A review on multi-label learning algorithms”. In: IEEE transac- tions on knowledge and data engineering 26.8 (2013), pp. 1819–1837

    2013

  46. [54]

    On the properties of neural machine translation: Encoder-decoder approaches

    Kyunghyun Cho, Bart Van Merri ¨enboer, Dzmitry Bah- danau, and Yoshua Bengio. “On the properties of neural machine translation: Encoder-decoder approaches”. In: arXiv preprint arXiv:1409.1259 (2014)

    2014 arXiv

  47. [55]

    Towards end-to-end speech recognition with recurrent neural networks

    Alex Graves and Navdeep Jaitly. “Towards end-to-end speech recognition with recurrent neural networks”. In: International conference on machine learning . 2014, pp. 1764–1772

    2014

  48. [56]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. “Adam: A method for stochastic optimization”. In: arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  49. [57]

    Sensor fault detection by sparsity optimization

    Bingxuan Li, Hang Yu, Justin Dauwels, and Kay Soon Low. “Sensor fault detection by sparsity optimization”. In: 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE. 2014, pp. 7614–7618

    2014

  50. [58]

    Fault identification in distributed sensor networks based on universal probabilistic mod- eling

    Stavros Ntalampiras. “Fault identification in distributed sensor networks based on universal probabilistic mod- eling”. In: IEEE transactions on neural networks and learning systems 26.9 (2014), pp. 1939–1949

    2014

  51. [59]

    Distributed Fault Detection in Sensor Net- works using a Recurrent Neural Network

    Oliver Obst. “Distributed Fault Detection in Sensor Net- works using a Recurrent Neural Network”. In: Neural Processing Letters 40.3 (Dec. 2014), pp. 261–273

    2014

  52. [60]

    Data- driven control and process monitoring for industrial ap- plications—Part I

    Shen Yin, Huijun Gao, and Okyay Kaynak. “Data- driven control and process monitoring for industrial ap- plications—Part I”. In: IEEE Transactions on Industrial Electronics 61.11 (2014), pp. 6356–6359

    2014

  53. [61]

    Anomaly detection in ECG time signals via deep long short-term memory net- works

    S. Chauhan and L. Vig. “Anomaly detection in ECG time signals via deep long short-term memory net- works”. In: 2015 IEEE International Conference on Data Science and Advanced Analytics (DSAA) . Oct. 2015, pp. 1–7

    2015

  54. [62]

    Fault isolation in data-driven multivariate process monitoring

    Dimitry Gorinevsky. “Fault isolation in data-driven multivariate process monitoring”. In:IEEE Transactions on Control Systems Technology 23.5 (2015), pp. 1840– 1852

    2015

  55. [63]

    Change detection with an unknown sensor subset: More information is not always better

    M. Guerriero, D. Huang, J. Unnikrishnan, M. Lexa, S. Iyengar, and F. Wheeler. “Change detection with an unknown sensor subset: More information is not always better”. In: 2015 18th International Conference on Information Fusion (FUSION) . IEEE. July 2015, pp. 1388–1394

    2015

  56. [64]

    Recurrent convolutional neural network for object recognition

    Ming Liang and Xiaolin Hu. “Recurrent convolutional neural network for object recognition”. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2015, pp. 3367–3375

    2015

  57. [65]

    Long Short Term Memory Networks for Anomaly Detection in Time Series

    Pankaj Malhotra, Lovekesh Vig, Gautam Shroff, and Puneet Agarwal. “Long Short Term Memory Networks for Anomaly Detection in Time Series”. In: 23rd Euro- pean Symposium on Artificial Neural Networks, Com- putational Intelligence and Machine Learning . 2015

    2015

  58. [66]

    Industry 4.0: the industrial internet of things

    Alasdair Gilchrist. Industry 4.0: the industrial internet of things. Apress, 2016

    2016

  59. [67]

    Deep learning

    Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep learning. MIT press, 2016

    2016

  60. [68]

    Learning distance metrics for multi-label classifica- tion

    Henry Gouk, Bernhard Pfahringer, and Michael J Cree. “Learning distance metrics for multi-label classifica- tion”. In: 8th Asian conference on machine learning . V ol. 63. 2016, pp. 318–333

    2016

  61. [69]

    The Visual Object Tracking VOT2016 Challenge Results

    Matej Kristan et al. “The Visual Object Tracking VOT2016 Challenge Results”. In: Computer Vision – ECCV 2016 Workshops . Ed. by Gang Hua and Herv ´e J´egou. Cham: Springer International Publishing, 2016, pp. 777–823. ISBN : 978-3-319-48881-3

    2016

  62. [70]

    LSTM-based Encoder-Decoder for Multi- sensor Anomaly Detection

    Pankaj Malhotra, Anusha Ramakrishnan, Gaurangi Anand, Lovekesh Vig, Puneet Agarwal, and Gau- tam Shroff. “LSTM-based Encoder-Decoder for Multi- sensor Anomaly Detection”. In: Anomaly Detection Workshop at 33rd International Conference on Machine Learning (ICML 2016) . July 2016

    2016

  63. [71]

    Measurement, instru- mentation, and sensors handbook: spatial, mechanical, thermal, and radiation measurement

    John G Webster and Halit Eren. Measurement, instru- mentation, and sensors handbook: spatial, mechanical, thermal, and radiation measurement . CRC press, 2016

    2016

  64. [72]

    Assessment and testing of sensor validation algorithms IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 15 for environmental monitoring applications

    G Ciarlo, E Bonica, Barbara Bosio, and N Bonavita. “Assessment and testing of sensor validation algorithms IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 15 for environmental monitoring applications”. In: Chemi- cal Engineering Transactions 57 (2017), pp. 331–336

    2017

  65. [73]

    Sensor Data Validation and Reconstruction

    Joseba Quevedo, Diego Garcia, Vicenc ¸ Puig, Jordi Saludes, Miquel Angel Cuguer´o, Santiago Espin, Jaume Roquet, and Fernando Valero. “Sensor Data Validation and Reconstruction”. In: Real-time Monitoring and Op- erational Control of Drinking-Water Systems . Ed. by Vicenc ¸ Pui...

    2017

  66. [74]

    A KPI-based process monitoring and fault detection framework for large-scale processes

    Kai Zhang, Yuri AW Shardt, Zhiwen Chen, Xu Yang, Steven X Ding, and Kaixiang Peng. “A KPI-based process monitoring and fault detection framework for large-scale processes”. In: ISA transactions 68 (2017), pp. 276–286

    2017

  67. [75]

    A Sensor Fault-Resilient Framework for Predictive Emission Monitoring Sys- tems

    Daniele Angelosante, Marco Guerriero, Gregorio Cia- rlo, and Nunzio Bonavita. “A Sensor Fault-Resilient Framework for Predictive Emission Monitoring Sys- tems”. In: 2018 21st International Conference on Infor- mation Fusion (FUSION) . IEEE. 2018, pp. 557–564

    2018

  68. [76]

    Disentangled ac- tivations in deep networks

    Mikael K ˚ageb¨ack and Olof Mogren. “Disentangled ac- tivations in deep networks”. In: (2018). URL: https:// openreview.net/forum?id=ByzvHagA-

    2018

  69. [77]

    Data-driven fault detection, isolation and estimation of aircraft gas turbine engine actuator and sensors

    Esmaeil Naderi and Khashayar Khorasani. “Data-driven fault detection, isolation and estimation of aircraft gas turbine engine actuator and sensors”. In: Mechanical Systems and Signal Processing 100 (2018), pp. 415– 438

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.