pith. sign in

arxiv: 2604.24880 · v1 · submitted 2026-04-27 · 📡 eess.SP · cs.LG· physics.soc-ph

Monitoring exposure-length variations in submarine power cables using distributed fiber-optic sensing

Sakiko Mishima , Yoshiyuki Yajima , Noriyuki Tonami , Tomoyuki Hino , Shugo Aibe , Junichiro Saikawa , Koji Mizuguchi This is my paper

Pith reviewed 2026-05-08 01:47 UTC · model grok-4.3

classification 📡 eess.SP cs.LGphysics.soc-ph
keywords distributed acoustic sensingsubmarine power cablesfree-span monitoringanomaly detectionone-class SVMexposure lengthfiber-optic sensingwave-tank experiments
0
0 comments X

The pith

A regression-based feature extraction method with one-class SVM on distributed acoustic sensing data detects exposure-length variations in submarine cables using only small training sets.

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

The paper develops an anomaly detection system to track how much submarine power cable is suspended above the seabed, known as free-span or exposure length. It extracts compact features from fiber-optic vibration signals that emphasize patterns tied to length while ignoring wave and current effects. These features then train a one-class support vector machine to spot when the length has changed, without needing examples of every possible fault. Wave-tank tests covering 2-to-10-meter lengths produced anomaly scores that fell steadily as exposure increased, reaching a correlation of -0.83, and the detector reached an F1 score of 0.82 even with limited data. The approach addresses the practical problem that real offshore cable monitoring rarely supplies large labeled datasets.

Core claim

The authors establish that regression-based feature extraction from DAS signals produces low-dimensional latent representations preserving exposure-length-dependent vibration characteristics while suppressing environmental influences, enabling one-class SVM anomaly detection that yields anomaly scores decreasing approximately monotonically with increasing exposure-length change (r = -0.83) and binary classification F1 score of 0.82 on small-sample wave-tank datasets.

What carries the argument

Regression-based feature extraction deriving low-dimensional latent representations that retain exposure length-dependent vibration characteristics while suppressing environmental influences, followed by one-class SVM anomaly detection.

If this is right

  • Anomaly scores decrease approximately monotonically with increasing exposure-length change.
  • The scores exhibit a strong negative correlation (r = -0.83) with exposure-length variations.
  • Binary classification reaches an F1 score of 0.82 despite training on only small-sample datasets.
  • Exposure-length variations can be reliably detected under severe data limitations typical of offshore environments.

Where Pith is reading between the lines

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

  • The same feature-extraction step could be reused to monitor additional cable conditions such as changes in burial depth without collecting new labeled data for each case.
  • Deployment on existing DAS-equipped cables would allow continuous rather than periodic inspections of free-span risk.
  • Performance in real ocean conditions remains an open test, since wave-tank results may understate the influence of tides, currents, and sediment movement.

Load-bearing premise

The wave-tank experiments with controlled exposure lengths from 2 to 10 m sufficiently capture the vibration signatures and environmental variability present in actual offshore deployments.

What would settle it

If field measurements on operating submarine cables show that anomaly scores do not decrease monotonically with independently verified exposure-length changes or yield correlation weaker than r = -0.83, the detection framework would fail to generalize beyond the lab setting.

Figures

Figures reproduced from arXiv: 2604.24880 by Junichiro Saikawa, Koji Mizuguchi, Noriyuki Tonami, Sakiko Mishima, Shugo Aibe, Tomoyuki Hino, Yoshiyuki Yajima.

Figure 1
Figure 1. Figure 1: illustrates typical environmental vibrations acting on submarine cables and the cable condition changes that can potentially be monitored using DAS. There is considerable interest in using DAS for early risk detection, such as identifying anomalies that might cause cable failures. One application uses DAS's ability to monitor specific localized points, enabling the detection of localized structural changes… view at source ↗
read the original abstract

This study proposes an anomaly-detection framework for monitoring exposure-length variations in submarine free-span cables using Distributed Acoustic Sensing (DAS), which is one of the distributed fiber-optic sensing technologies. To address environmental variability and limited training data in offshore environments, a regression-based feature extraction method was introduced to derive low-dimensional latent representations that retain exposure length-dependent vibration characteristics while suppressing environmental influences. The extracted features were used for one-class Support Vector Machine (SVM)-based anomaly detection. The proposed framework was evaluated through wave-tank experiments with exposure lengths ranging from 2 to 10 m. Experimental results showed that anomaly scores decreased approximately monotonically with increasing exposure-length change, exhibiting a strong correlation ($r = -0.83$). The binary classification achieved an F1 score of 0.82 despite training with only small-sample datasets. These findings demonstrate that exposure-length variations can be reliably detected under severe data limitations, supporting the potential of DAS-based cable condition monitoring.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a regression-based feature extraction pipeline to derive low-dimensional latent representations from DAS data that retain exposure-length-dependent vibration signatures while suppressing environmental variability, followed by one-class SVM anomaly detection. The method is evaluated on wave-tank experiments with controlled exposure lengths of 2–10 m; results show anomaly scores decreasing monotonically with exposure change (r = −0.83) and binary classification achieving F1 = 0.82 when trained on small-sample datasets.

Significance. If the reported monotonic trend and classification performance generalize beyond the tank setting, the work would offer a practical route to DAS-based free-span monitoring under the severe data limitations typical of offshore deployments. The explicit use of regression to isolate exposure-dependent features and the demonstration of usable performance with limited training data are concrete strengths that address a recognized operational constraint.

major comments (2)
  1. [Experimental Results] Experimental Results section: the reported correlation r = −0.83 and F1 = 0.82 are presented without error bars, confidence intervals, number of independent trials, or specification of the regression model (type, features, hyperparameters, or loss) used to obtain the latent representations. These omissions are load-bearing for the central claim that the pipeline reliably detects exposure-length variations under small-sample conditions.
  2. [Discussion] Discussion / Conclusion: the assertion that the framework supports offshore cable monitoring rests on the assumption that wave-tank conditions with fixed 2–10 m exposures reproduce the vibration signatures encountered under stochastic hydrodynamic forcing, variable currents, and seabed interactions. No sensitivity test (e.g., addition of non-stationary noise or variable wave spectra to the tank data) is reported to quantify how such unmodeled effects would shift the latent features or degrade the observed correlation.
minor comments (2)
  1. [Abstract] Abstract: the phrase “approximately monotonically” is imprecise; the exact functional form or statistical test used to establish monotonicity should be stated.
  2. [Method] The manuscript would benefit from a brief comparison against at least one baseline feature set (e.g., raw statistical moments or frequency-domain summaries) to demonstrate the added value of the regression step.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive and detailed review of our manuscript. We address each major comment point by point below, indicating the revisions we will make to strengthen the paper.

read point-by-point responses

  1. Referee: [Experimental Results] Experimental Results section: the reported correlation r = −0.83 and F1 = 0.82 are presented without error bars, confidence intervals, number of independent trials, or specification of the regression model (type, features, hyperparameters, or loss) used to obtain the latent representations. These omissions are load-bearing for the central claim that the pipeline reliably detects exposure-length variations under small-sample conditions.

    Authors: We agree that these statistical details and model specifications are essential for supporting the central claims. In the revised manuscript, we will add error bars and confidence intervals to the reported correlation and F1 score, specify the number of independent trials performed, and provide complete details on the regression model including its type, selected features, hyperparameters, and loss function. These elements were part of our analysis but were not fully documented in the original submission. revision: yes

  2. Referee: [Discussion] Discussion / Conclusion: the assertion that the framework supports offshore cable monitoring rests on the assumption that wave-tank conditions with fixed 2–10 m exposures reproduce the vibration signatures encountered under stochastic hydrodynamic forcing, variable currents, and seabed interactions. No sensitivity test (e.g., addition of non-stationary noise or variable wave spectra to the tank data) is reported to quantify how such unmodeled effects would shift the latent features or degrade the observed correlation.

    Authors: We acknowledge that wave-tank experiments with fixed exposures provide a controlled but simplified representation of offshore conditions and do not capture the full range of stochastic hydrodynamic forcing, variable currents, or seabed interactions. In the revised manuscript, we will expand the Discussion and Conclusion sections to explicitly articulate these assumptions, their potential impact on the latent features, and the resulting limitations on generalizability. However, performing the suggested sensitivity tests would require new experimental data that is not available from the current study. revision: partial

standing simulated objections not resolved
  • Additional sensitivity tests to non-stationary noise or variable wave spectra cannot be provided, as they would require new wave-tank experiments beyond the scope and data of the present work.

Circularity Check

0 steps flagged

No circularity: empirical results measured on held-out tank data

full rationale

The paper describes a standard machine-learning pipeline: regression-based feature extraction to obtain low-dimensional latent representations, followed by one-class SVM anomaly detection. Anomaly scores and F1 performance are computed directly against held-out wave-tank experiments with known exposure lengths (2-10 m). No equation or step reduces by construction to its own inputs; the correlation (r = -0.83) and classification metric are external measurements on experimental outcomes rather than tautological re-statements of fitted parameters. No load-bearing self-citations or uniqueness theorems are invoked. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard supervised and unsupervised machine-learning assumptions rather than new physical axioms or invented entities. No free parameters are explicitly fitted beyond routine SVM regularization; no new particles or forces are postulated.

axioms (2)
  • domain assumption Vibration signatures in DAS data contain separable components attributable to exposure length versus environmental noise
    Invoked to justify the regression-based feature extraction step that retains exposure-dependent characteristics while suppressing environmental influences.
  • domain assumption One-class SVM trained on limited normal samples can reliably flag exposure-length anomalies
    Central modeling choice for the anomaly-detection stage.

pith-pipeline@v0.9.0 · 5493 in / 1374 out tokens · 58375 ms · 2026-05-08T01:47:39.428377+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

  1. [1]

    Gulski E, Anders G J, Jongen R A, Parciak J, Siemin ski J, Piesowicz E, Paszkiewicz S and Irska I 2021 Discussion of electrical and thermal aspects of offshore wind farms’ power cables reliability Renewable and Sustainable Energy Reviews 151 111580 (doi: 10.1016/j.rser.2021.111580)

    work page doi:10.1016/j.rser.2021.111580 2021

  2. [2]

    Mou J, Jia X, Chen P and Chen L 2021 Research on operation safety of offshore wind farms Journal of Marine Science and Engineering 9(8) 881 (doi: 10.3390/jmse9080881)

    work page doi:10.3390/jmse9080881 2021

  3. [3]

    Hartog A H 2017 An introduction to distributed optical fibre sensors Boca Raton, FL: CRC Press (doi:10.1201 /9781315119014)

    work page 2017

  4. [4]

    Power Delivery 34 pp 1079–1086 (doi: 10.1109/TPWRD.2018.2881770)

    Lux J, Olschewski M, Scha fer P and Hill W 2019 Real-time determination of depth of burial profiles for submarine power cables IEEE Trans. Power Delivery 34 pp 1079–1086 (doi: 10.1109/TPWRD.2018.2881770)

    work page doi:10.1109/tpwrd.2018.2881770 2019

  5. [5]

    Li Q, Hao Y, Zhang P, Tan H and Yang L 2024 Study on the burial depth calculation method for AC submarine cable based on surface optical fiber monitoring temperature IET Gener. Transm. Distrib. 18 pp 1893–1904 (doi: 10.1049/gtd2.13243)

    work page doi:10.1049/gtd2.13243 2024

  6. [6]

    Simon N, Bour O, Lavenant N, Porel G, Nauleau B, Pouladi B and Longuevergne L 2020 A comparison of different methods to estimate the effective spatial resolution of FO-DTS measurements achieved during sandbox experiments Sensors 20 570 (doi: 10.3390/s20020570)

    work page doi:10.3390/s20020570 2020

  7. [7]

    O’Donnell Meininger T and Selker J S 2015 Bed conduction impact on fiber optic distributed temperature sensing water temperature measurements Geosci. Instrum. Method. Data Syst. 4 pp 19–22 (doi: 10.5194/gi-4-19- 2015)

    work page doi:10.5194/gi-4-19- 2015

  8. [8]

    Duthe G, L ’Homme Y , Abdallah I and Chatzi E 2024 Towards unsupervised fault detection for offshore wind turbine cable protection systems using contrastive learning J. Phys. Conf. Ser. 2767 032038 (doi: 10.1088/1742- 6596/2767/3/032038)

    work page doi:10.1088/1742- 2024

  9. [9]

    Mizuguchi K, Mishima S and Ding Y 2025 Monitoring submarine power transmission cable conditions with optical fiber sensing technology for offshore wind power generation NEC Technical Journal 18 1

    work page 2025

  10. [10]

    Tang X, Wang Z, He Q, Liu J and Ying Z 2020 Latent feature extraction for process data via multidimensional scaling Journal of Educational and Behavioral Statistics 85 2 pp 378–397 (doi:10.1007/s11336-020-09708-3)

    work page doi:10.1007/s11336-020-09708-3 2020

  11. [11]

    Pirouz D M 2006 An overview of partial least squares SSRN Electronic Journal (doi:10.2139/ssrn.1631359)

    work page doi:10.2139/ssrn.1631359 2006

  12. [12]

    Scho lkopf B, Platt J C, Shawe-Taylor J, Smola A J and Williamson R C 2001 Estimating the support of a high - dimensional distribution Neural Computation 13 7 pp 1443–1471(doi: 10.1162/089976601750264965)

    work page doi:10.1162/089976601750264965 2001

  13. [13]

    Mann H B and Whitney D R 1947 On a test of whether one of two random variables is stochastically larger than the other Ann. Math. Stat. 18 1 pp 50–60

    work page 1947