Pith. sign in

REVIEW 3 major objections 6 minor 48 references

RSMI-IPCA makes the batch errors-in-variables subspace identification method recursive, so a MIMO state-space model, its order, and input/output noise variances can be tracked online from a fixed-length lag window.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A recursive extension of the SMI-IPCA algorithm that updates state-space model, model order, and input/output noise variances online in the errors-in-variables setting.

T0 review reviewed 2026-08-01 challenge →

load-bearing objection RSMI-IPCA is a genuinely new recursive EIV subspace identification algorithm with solid simulations, but the unweighted cumulative covariance update does not support the claim of tracking abrupt changes late in the record, and input-noise variance estimates are biased in two of three case studies. the 3 major comments →

arxiv 2607.17065 v1 pith:AXV3TRAN submitted 2026-07-19 eess.SY cs.SY

A recursive subspace based method for errors-in-variables model identification of time-varying systems

Deepanjhan Das , Shankar Narasimhan This is my paper
classification eess.SY cs.SY
keywords recursive subspace identificationerrors-in-variablestime-varying systemsiterative principal component analysisnoise variance estimationonline system identificationprocess order estimationMIMO state-space models
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 the theoretically rigorous batch method SMI-IPCA can be converted into a fully recursive algorithm, RSMI-IPCA, that identifies a linear state-space model of a MIMO process when both inputs and outputs are corrupted by measurement noise, and does so online without storing the full history. The recursive update runs on a lagged sample covariance matrix of fixed size; from this matrix the algorithm re-estimates the number of constraints (process order), the state-space matrices, and the input/output noise variances at every new sample. The authors argue this fills a gap in the literature: prior recursive subspace methods either ignore input noise, require a known model order, or cannot estimate noise variances, so none can track the simultaneous changes in dynamics, sensor degradation, and model structure that occur in real processes. The paper's simulations show tracking of gradual sensor degradation, changing operating conditions, and a structural change that adds an output variable, with unbiased pole and zero estimates over 50 trials.

Core claim

The central claim is that the iterative optimization for the measurement-noise covariance matrix in SMI-IPCA can be reformulated, using the cyclic property of the trace, so that its objective depends only on the recursively updated lagged sample covariance matrix and the current estimate of the dynamic constraint matrix, rather than on the full data matrix. With this reformulation, the batch algorithm's two-step iteration—scaling by the Cholesky factor of the current noise covariance, extracting the constraint subspace by eigendecomposition, testing the smallest eigenvalues to fix the order, and re-solving for the noise variances—can be run at each sampling instant using only the previous co

What carries the argument

The recursively updated lagged sample covariance matrix, computed from the previous covariance and the current lagged measurement vector, is the load-bearing object: it replaces the stored history. The other key piece is the trace-based reformulation of the variance-estimation objective, which turns a sum over all past samples into an expression involving only the lagged covariance matrix and the constraint-model estimate, so the noise-covariance optimization can be solved online. Order estimation uses the eigendecomposition of the Cholesky-scaled covariance and a hypothesis test for equality of the smallest eigenvalues, the same statistical test as the batch method.

Load-bearing premise

The recursive covariance update is a plain cumulative average with no forgetting factor, so as the sample count grows, each new sample's influence on the covariance estimate shrinks to zero—meaning the method's ability to track changes that occur after a long history is not guaranteed, and the paper's 'abrupt change' claim is only tested for changes early in the record.

What would settle it

Run RSMI-IPCA on a system whose measurement-noise variance or a model parameter steps to a new value at a late time (e.g., after 20,000 samples of a 20,000-sample simulation) and check whether the estimates move toward the new values within a few hundred samples; the paper provides no convergence or tracking-error result for the recursive estimator, so a failure to adapt in that regime, or a formal bound showing the adaptation rate tends to zero, would refute the claim of general time-varying tracking.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A process with slowly degrading sensors can be monitored online: the estimated measurement-noise standard deviations follow the true values, enabling predictive maintenance without storing historical data.
  • Changes in operating conditions are reflected in the estimated poles; the pole estimates converge to the new values, so the model can be kept current for observer and controller design.
  • Structural modifications, such as adding a new output sensor, are handled by extending the lagged covariance matrix; the process-order estimate later converges to the new order.
  • Because memory depends on the lagged dimension and not the history length, the method is deployable on embedded or edge devices with limited storage.
  • When the system is time-invariant, the recursive estimates converge toward the batch SMI-IPCA estimates, so the theoretical guarantees of the batch method carry over in that regime.

Where Pith is reading between the lines

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

  • The unweighted cumulative average in the covariance update means each new sample's weight shrinks as the sample count grows; the paper provides no forgetting factor or tracking-error bound, so its claim of adapting to abrupt changes is only demonstrated for changes that occur early in the record, and a step change late in a long run is an untested regime where the estimates may barely move.
  • Introducing a forgetting factor into the covariance update would trade tracking agility against asymptotic variance; a natural extension is an adaptive forgetting factor that detects change points (for instance via the order-test residual) and resets the effective memory.
  • The similarity transformation that sets the process-noise covariance to identity means the method identifies a particular state-space coordinate system; the paper leaves the non-full-rank process-noise (innovations form) case as future work, so the practical scope is systems with full-rank process noise.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes RSMI-IPCA, a recursive extension of the batch SMI-IPCA algorithm for errors-in-variables (EIV) identification of time-varying linear state-space models. The method updates a sample covariance matrix of stacked input/output vectors using Eq. (32), then alternates between estimating the constraint matrix via eigendecomposition of a scaled covariance and estimating input/output measurement noise variances via the objective in Eq. (34), with process order determined by an eigenvalue-equality hypothesis test. Three simulation studies (sensor degradation, changing valve coefficients, and addition of an output sensor) are used to claim tracking of noise variances, model matrices, and process order. The advertised contribution is the first recursive EIV subspace method that simultaneously updates state-space matrices, process order, and noise variances from a fixed-length lag memory.

Significance. If the claimed properties are correct, the method would fill a genuine gap: existing recursive subspace methods either ignore input noise or require a known order, while SMI-IPCA/RIPCA are batch/static. The algebraic reformulation of the variance estimation objective in covariance-recursive form (Eq. 34) is elegant, and the simulation protocol (50 trials, confidence intervals, pole/zero comparisons) is transparent. However, the central claims are not yet supported. Eq. (32) is an unweighted cumulative average with no forgetting factor, no tracking bound is provided, the recursive use of the batch hypothesis test is not proven, and two of the three case studies show large biases in input-noise variance estimates. These gaps are load-bearing for the paper's claims of online adaptation to abrupt changes and preservation of batch theoretical rigor.

major comments (3)
  1. [§3.1, Eq. (32)] The recursive covariance update is S_Zf,k = (N_{k-1}/N_k) S_Zf,k-1 + (1/N_k) z_f(k)z_f(k)^T, which is a cumulative average with no forgetting factor. For large N_k, each new sample has weight 1/N_k → 0, so a step change occurring late in the record (e.g., at sample 5000 of 6095) changes the covariance only infinitesimally per new sample; reaching even 50% weight on the new regime requires roughly N_old new samples. All three case studies introduce changes by sample 1495 at the latest, so the new data can eventually dominate. The abstract's 'fixed length lag window' is therefore not realized in the data-influence sense: the stored object has constant dimension, but the estimator's memory is the full history. The claim of adaptive, abrupt tracking is unsupported unless a forgetting factor is introduced or a tracking-error bound is supplied, and it must be tested with a late-onset change.
  2. [§3.1, Eq. (34); Algorithm 1, step 8] The advertised theoretical properties — the eigenvalue-equality hypothesis test for order determination and the likelihood objective for variance estimation — are imported from the batch SMI-IPCA paper [39] and RIPCA [40] without a proof that they remain valid for the recursively updated S_Zf,k under time-varying model and noise. Eq. (34) is derived from a log-likelihood that assumes a single noise covariance over all N_k observations; when Σ_e changes over time, the cumulative S_Zf,k averages data generated under different noise covariances, so Eq. (34) is not the correct likelihood for the time-varying problem. No convergence or tracking-error result is provided, so the Introduction's claim that RSMI-IPCA preserves 'theoretical rigor of the batch method' is not supported. A formal treatment, or at minimum a clear statement of the approximation and its domain of validity, is needed.
  3. [Table 1 and §4.2.1] In two of the three case studies, the input-noise variance estimates are substantially biased. Table 1 reports σ̂_eu2 = 0.8259 for a true value 0.4472, with a 95% CI approximately [0.6307, 1.0211] that excludes the truth — a factor of about 1.85. Section 4.2.1 reports σ̂_Fi = 0.3314 ± 0.0408 against a true value 0.1581, a factor of about 2.1; the text states that more samples are needed, but no convergence is demonstrated within N = 6095. Since simultaneous noise-variance tracking is a central advertised capability, this bias needs to be explained or remedied (e.g., by testing larger lags f, as suggested in [39], or by reporting longer-horizon experiments).
minor comments (6)
  1. [Eq. (32) and Algorithm 1, line 1] The indexing of z_f(k) and N_k should be clarified. If k indexes sampling instants, the number of lagged observations is N_k - f + 1, not N_k, so the normalization in Eq. (32) is inconsistent with the definition of Z_f in Eq. (20). If k indexes lagged vectors, this should be stated explicitly.
  2. [Algorithm 1, step 8] The notation 'hypothesis test on S_λ' is undefined. The paper should specify the statistic and its distribution, or point to the exact equation in [39].
  3. [Eq. (11) and Eq. (34)] The running index k is used both as a summation index and as the current time index. Use a different symbol for the summation index to avoid confusion.
  4. [Abstract and §3.1] The phrase 'fixed length lag window' should be reconciled with the cumulative covariance update. If the claim is only that storage is constant-dimensional, that should be stated precisely; the current wording implies a finite-memory estimator, which Eq. (32) does not provide.
  5. [§4.2.1, Figure 6(c)] The caption says the input SD estimate shows 'initial divergence, indicating biased estimate', but the final estimate remains strongly biased. Please update the wording and discussion to reflect the actual final bias.
  6. [General] A comparison with a baseline recursive method (e.g., RPBSID or recursive MOESP with EIV) would help quantify the improvement; the current study is self-contained but lacks a benchmark.

Circularity Check

0 steps flagged

No significant circularity: RSMI-IPCA is a direct recursive extension of previously published and independently validated batch/IPCA methods; the main weaknesses (unweighted covariance update, imported rigor) are support/correctness concerns, not input-equivalence.

full rationale

The paper's derivation chain is a standard extension: Eq. (32) rewrites the RPCA covariance update of Li et al. (Eq. 4) for the lagged vector z_f(k), and Eq. (34) rewrites the SMI-IPCA/RIPCA noise-variance objective (Eqs. 29, 8, 11) so that it depends only on S_Zf,k. These are algebraic substitutions, not definitions that presuppose the stated outputs. The claimed outputs — model order, noise variances, and state-space matrices — are compared in simulations against independent ground-truth systems (Eqs. 36-37), so the empirical validation does not reduce to the fitted inputs. The self-citations [39] and [40] supply the batch SMI-IPCA and RIPCA ingredients; both are prior published methods with their own independent development, and the paper does not invoke them to forbid alternatives or to define the recursive estimate into existence. The more serious defects are not circularity: Eq. (32) is an unweighted cumulative average with no forgetting factor, so the method's advertised ability to track abrupt changes late in the record is not established; and the 'theoretical rigor' claim imports convergence/identifiability results without re-proving them in the recursive setting. Those are correctness/robustness concerns, which under the reviewing rules do not raise the circularity score.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

No new physical entities are introduced; the contribution is algorithmic. The main uncharged premises are the inherited EIV/noise assumptions and the unproven tracking behavior of the unweighted covariance recursion.

free parameters (4)
  • Lag f = 6 in all case studies
    Chosen by hand; the paper itself notes in Section 4.1.1 that larger lag improves noise-variance accuracy, so reported biases depend on this choice.
  • Initialization window length = 400 samples (case studies 1-2); 200 samples (case study 3)
    Batch SMI-IPCA is run on the first window to initialize the recursive estimator; the window length is not justified beyond convenience.
  • Initial noise variance for newly added output h2 = Small fraction of the (2,2) diagonal element of z_f z_f^T / 1501
    Ad hoc initialization in Section 4.2.2 needed to extend the covariance matrix when a structural change adds a variable.
  • Inner-loop tolerance and maximum iterations (epsilon_lambda, i_max) = Not specified numerically
    Algorithm 1 configuration parameters; no guidance is given on the values used, which affects reproducibility and convergence behavior.
axioms (5)
  • domain assumption Measurement noises v(k), w(k) and process noise p(k) are mutually independent zero-mean Gaussian white sequences with diagonal covariance matrices.
    Section 2.3, Eqs. (12)-(13); the likelihood-based variance estimator and the eigenvalue equality test rely on normality and independence.
  • domain assumption Process noise covariance is full rank, allowing a state transformation T = Sigma_p^{-1/2} that makes the transformed process noise covariance identity; process noise variances are therefore not estimated.
    Section 2.3 after Eq. (27); if process noise is rank-deficient, the paper explicitly leaves extension to future work in Section 5.
  • domain assumption Identifiability condition d_min(d_min+1) >= 2(m+ell) holds, i.e., the lag f is large enough.
    Section 2.3, Eq. (30); when this condition fails, noise variances are not uniquely identifiable from the data.
  • ad hoc to paper The unweighted recursive covariance update in Eq. (32) still tracks slow or abrupt time variations.
    Section 3.1, Eq. (32); this is asserted, not proven, and is questionable at large N_k because each new sample receives weight 1/N_k.
  • ad hoc to paper The eigenvalue-equality hypothesis test from [39] remains valid for recursively updated finite-sample covariance matrices.
    Algorithm 1 step 8 defers to [39]; no finite-sample or asymptotic justification is given in the recursive setting.

reviewed 2026-08-01 · how reviews work

0 comments
Cite this review

Pith. "Pith review of A recursive subspace based method for errors-in-variables model identification of time-varying systems." pith.science (2026). https://pith.science/paper/AXV3TRAN

@misc{pith2026260717065,
  author       = {Pith},
  title        = {Pith review of: A recursive subspace based method for errors-in-variables model identification of time-varying systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AXV3TRAN}},
  note         = {Machine review of arXiv:2607.17065}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The Subspace-based Model Identification algorithm using a modified Iterative Principal Component Analysis (SMI-IPCA) is a theoretically rigorous method for identifying a linear state-space model of a multi-input multi-output (MIMO) process, in an errors-in-variables (EIV) setting. The method can simultaneously estimate unknown heteroskedastic noise variances corrupting the input and output measurements, along with the state space model. This work proposes a recursive formulation of SMI-IPCA (RSMI-IPCA) enabling online identification and adaptive model updates as and when new data arrive. By maintaining a fixed length lag window rather than storing the complete historical data, RSMI-IPCA estimates measurement noise variances, process order, while simultaneously identifying the state-space matrices, making it suitable to monitor time-varying systems, whether the induced changes are slow or abrupt. The algorithm gradually adapts to slow sensor degradation (time-varying noise variances), changes in process operating conditions (time-varying model parameters), and structural modifications (varying model order). Simulation studies are presented to demonstrate the efficacy and practical applicability of the proposed algorithm.

Figures

Figures reproduced from arXiv: 2607.17065 by Deepanjhan Das, Shankar Narasimhan.

Figure 1
Figure 1. Figure 1: Simulation of gradual sensor degradation, i.e., increment in the noise standard deviations of input and output measurements in a quadratic manner from 695’th instant to 1095’th instant. 0 500 1000 1500 2000 2500 3000 3500 4000 Sampling Instances (s) -1 0 1 2 3 4 5 6 j^2 ! 2j Mean Error Mean + 1.96 SD Mean - 1.96 SD (a) Absolute difference between the estimated and true process orders of the fourth order sy… view at source ↗
Figure 2
Figure 2. Figure 2: Performance analysis of RSMI-IPCA for the LTI fourth-order system as described in Eq. (36) under gradual sensor degradation. RSMI-IPCA is still able to maintain accurate tracking of the changes as we see in Figure 2b. This is due to re-estimating the noise variances at each step as new data is received. The final estimates of the measurement noise standard deviations are provided in [PITH_FULL_IMAGE:figur… view at source ↗
Figure 3
Figure 3. Figure 3: The mean of the estimates of poles and zeros of the system defined in Eq. (36) are observed to be pretty close to the respective true values [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Schematic of the non-interacting two-tanks in series. 4.2.1 Case study 2: Tracking changes in process operating conditions In this case study, we simulate the scenario of changing process operating conditions by varying the valve coefficients and demonstrate the applicability of RSMI-IPCA to track these changes. At first, the system is brought to steady state and thereafter excited with the designed input.… view at source ↗
Figure 5
Figure 5. Figure 5: Simulation of variational process operational conditions, where the valve coefficients are changed gradually in a quadratic manner from 695’th instant to 1095’th instant. 0 1000 2000 3000 4000 5000 6000 Sampling Instances (s) -1 -0.5 0 0.5 1 1.5 2 j^2 ! 2j Mean Error Mean + 1.96 SD Mean - 1.96 SD (a) The process order estimate converges gradually to its true value η = 2 around 3500’th sampling instant 0 10… view at source ↗
Figure 6
Figure 6. Figure 6: Performance of RSMI-IPCA on the system, defined in Eq. (37) under changing operating condi￾tion. samples to converge to the true value. The final estimate of the input noise SD is σˆFi = 0.3314 ± 0.0408, which still has a bias and require more samples for RSMI-IPCA to converge to an estimate close to the corresponding true value, whereas the estimated output SD σˆh2 = 0.1314 ± 0.0028 is fairly accurate. Fu… view at source ↗
Figure 7
Figure 7. Figure 7: Performance of RSMI-IPCA on the system, defined in Eq. (37) in presence of a structural variation [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗

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

48 extracted references · 18 canonical work pages

  1. [1]

    E.; Prett, D

    García, C. E.; Prett, D. M.; Morari, M. Model predictive control: Theory and practice—A survey. Automatica1989,25, 335–348, DOI:https://doi.org/10.1016/0005-1098(89)90002-2. 19

    doi:10.1016/0005-1098(89)90002-2

  2. [2]

    Kim, S.; Tullu, A.; Jung, S. Linearized State-Space Model-Based Attitude Control for Rocket with Four Controllable Fins—Part 1-1: Basic Modeling and Identification.IEEE Access2023,11, 146014–146029, DOI:https://doi.org/10.1109/ACCESS.2023.3338445

    arXiv 2023

  3. [3]

    Monticelli, A. J.State Estimation in Electric Power Systems: A Generalized Approach, 1st ed.; Power Electronics and Power Systems; Springer New York, NY, 1999; pp XVII, 394, DOI: https://doi.org/10.1007/978-1-4615-4999-4

    work page doi:10.1007/978-1-4615-4999-4 1999

  4. [4]

    Söderström, T.; Stoica, P.System Identification; International series in systems and control engineering; Prentice-Hall, 1989

    1989

  5. [5]

    Ljung, L.System Identification: Theory for the User, 2nd ed.; Prentice Hall PTR: Upper Saddle River, NJ, 1999

    1999

  6. [6]

    Tangirala, A. K.Principles of System Identification: Theory and Practice, 1st ed.; Principles of System Identification; CRC Press: Boca Raton, 2015; p 908, DOI:https://doi.org/10.1201/9781315222509

    work page doi:10.1201/9781315222509 2015

  7. [7]

    Subspace Model Identification Part 1

    Verhaegen, M.; Dewilde, P. Subspace Model Identification Part 1. The Output-Error State-Space Model Identification Class of Algorithms.International Journal of Control1992,56, 1187–1210, DOI: https://doi.org/10.1080/00207179208934363

    doi:10.1080/00207179208934363

  8. [8]

    Subspace Model Identification Part 2

    Verhaegen, M.; Dewilde, P. Subspace Model Identification Part 2. Analysis of the Elementary Output- Error State-Space Model Identification Algorithm.International Journal of Control1992,56, 1211– 1241, DOI:https://doi.org/10.1080/00207179208934364

    work page doi:10.1080/00207179208934364

  9. [9]

    Van Overschee, P.; De Moor, B. N4SID: Subspace Algorithms for the Identifica- tion of Combined Deterministic-Stochastic Systems.Automatica1994,30, 75–93, DOI: https://doi.org/10.1016/0005-1098(94)90230-5, Special issue on statistical signal processing and control

    doi:10.1016/0005-1098(94)90230-5

  10. [10]

    Identification of Stochastic Linear Systems in presence of Input Noise.Automatica1981, 17, 713–725, DOI:https://doi.org/10.1016/0005-1098(81)90018-2

    Söderström, T. Identification of Stochastic Linear Systems in presence of Input Noise.Automatica1981, 17, 713–725, DOI:https://doi.org/10.1016/0005-1098(81)90018-2

    work page doi:10.1016/0005-1098(81)90018-2

  11. [11]

    Kalman, R. E. A New Approach to Linear Filtering and Prediction Problems.Journal of Basic Engi- neering1960,82, 35–45, DOI:https://doi.org/10.1115/1.3662552

    doi:10.1115/1.3662552

  12. [12]

    S.; Andrews, A

    Grewal, M. S.; Andrews, A. P.Kalman Filtering: Theory and Practice Using MATLAB, 4th ed.; Wiley- IEEE Press: Hoboken, NJ, 2015; DOI:https://doi.org/10.1002/9781118984987

    work page doi:10.1002/9781118984987 2015

  13. [13]

    J.Introduction to Stochastic Control Theory; Mathematics in Science and Engineering; Academic Press: New York, 1970; Vol

    Åström, K. J.Introduction to Stochastic Control Theory; Mathematics in Science and Engineering; Academic Press: New York, 1970; Vol. 70

    1970

  14. [14]

    Stengel, R. F.Optimal Control and Estimation; Dover Books on Mathematics; Dover Publications: New York, 1994; Unabridged republication of Stochastic Optimal Control: Theory and Application, 1986

    1994

  15. [15]

    Söderström, T.Errors-in-Variables Methods in System Identification, 1st ed.; Com- munications and Control Engineering; Springer Cham, 2018; pp XXVII, 485, DOI: https://doi.org/10.1007/978-3-319-75001-9

    doi:10.1007/978-3-319-75001-9 2018

  16. [16]

    Stoica, P.; Cedervall, M.; Eriksson, A. Combined Instrumental Variable and Subspace Fitting Approach to Parameter Estimation of Noisy Input-Output Systems.IEEE Transactions on Signal Processing 1995,43, 2386–2397, DOI:https://doi.org/10.1109/78.469852

    doi:10.1109/78.469852 1995

  17. [17]

    Chou, C.; Verhaegen, M. Subspace Algorithms for the Identification of Multivari- able Dynamic Errors-in-Variables Models.Automatica1997,33, 1857–1869, DOI: https://doi.org/10.1016/S0005-1098(97)00092-7, This paper was not presented at any IFAC meeting. This paper was recommended for publication in revised form by Associate Editor H. Hjalmarsson under the ...

    work page doi:10.1016/s0005-1098(97)00092-7

  18. [18]

    Consistent Dynamic PCA based on Errors-in-Variables Subspace Identification.Journal of Process Control2001,11, 661–678, DOI:https://doi.org/10.1016/S0959-1524(00)00041-X

    Li, W.; Qin, S. Consistent Dynamic PCA based on Errors-in-Variables Subspace Identification.Journal of Process Control2001,11, 661–678, DOI:https://doi.org/10.1016/S0959-1524(00)00041-X

    work page doi:10.1016/s0959-1524(00)00041-x

  19. [19]

    A new Subspace Identification Approach based on Princi- pal Component Analysis.Journal of Process Control2002,12, 841–855, DOI: https://doi.org/10.1016/S0959-1524(02)00016-1

    Wang, J.; Qin, S. A new Subspace Identification Approach based on Princi- pal Component Analysis.Journal of Process Control2002,12, 841–855, DOI: https://doi.org/10.1016/S0959-1524(02)00016-1

    doi:10.1016/s0959-1524(02)00016-1

  20. [20]

    Zheng, W. X. A Bias Correction Method for Identification of Linear Dynamic Errors-in- Variables Models.IEEE Transactions on Automatic Control2002,47, 1142–1147, DOI: https://doi.org/10.1109/TAC.2002.800661

    arXiv 2002

  21. [21]

    Ikenoue, M.; Kanae, S.; Yang, Z.-J.; Wada, K. Identfication of Noisy Input-Output System us- ing Bias-Compensated Least-Squares Method.IFAC Proceedings Volumes2005,38, 803–808, DOI: https://doi.org/10.3182/20050703-6-CZ-1902.00135, 16th IFAC World Congress

    Pith/arXiv arXiv 1902

  22. [22]

    Söderström, T. A Generalised Instrumental Variable Estimator for Multivariable Errors-in- Variables Identification Problems.International Journal of Control2012,85, 287–303, DOI: https://doi.org/10.1080/00207179.2011.647086

    arXiv 2011

  23. [23]

    Akaike, H. InSelected Papers of Hirotugu Akaike; Parzen, E., Tanabe, K., Kita- gawa, G., Eds.; Springer New York: New York, NY, 1998; pp 199–213, DOI: https://doi.org/10.1007/978-1-4612-1694-0_15

    doi:10.1007/978-1-4612-1694-0_15 1998

  24. [24]

    Estimating the Dimension of a Model.The Annals of Statistics1978,6, 461 – 464, DOI: https://doi.org/10.1214/aos/1176344136

    Schwarz, G. Estimating the Dimension of a Model.The Annals of Statistics1978,6, 461 – 464, DOI: https://doi.org/10.1214/aos/1176344136

    arXiv

  25. [25]

    Hashlamon, I.; Erbatur, K. An Improved Real-Time Adaptive Kalman Filter with Recursive Noise Covariance Updating Rules.Turkish Journal of Electrical Engineering and Computer Sciences2016, 24, 524–540, DOI:https://doi.org/10.3906/elk-1309-60

    work page doi:10.3906/elk-1309-60

  26. [26]

    Hajiyev, C. Robust Kalman Filter with Recursive Measurement Noise Covariance Estimation Against Measurement Faults.International Journal of Prognostics and Health Management2025,16, 1–11, DOI:https://doi.org/10.36001/ijphm.2025.v16i1.4204

    work page doi:10.36001/ijphm.2025.v16i1.4204 2025

  27. [27]

    Predictive Maintenance with Sensor Data Analytics on a Raspberry Pi-Based Experimental Platform.Sensors2019,19, DOI: https://doi.org/10.3390/s19183884

    Chuang, S.-Y.; Sahoo, N.; Lin, H.-W.; Chang, Y.-H. Predictive Maintenance with Sensor Data Analytics on a Raspberry Pi-Based Experimental Platform.Sensors2019,19, DOI: https://doi.org/10.3390/s19183884

    work page doi:10.3390/s19183884

  28. [28]

    K.; Koppolu, H

    Peruthambi, V.; Pandiri, L.; Kaulwar, P. K.; Koppolu, H. K. R.; Adusupalli, B.; Pamisetty, A. Big Data-Driven Predictive Maintenance for Industrial IoT (IIoT) Systems.Metallurgical and Materials Engineering2025,31, 21–30, DOI:https://doi.org/10.63278/1316

    work page doi:10.63278/1316

  29. [29]

    de la Fuente, R.; Radrigan, L.; Morales, A. S. Enhancing Predictive Maintenance in Mining Mobile Machinery Through a Hierarchical Inference Network.IEEE Access2025,13, 59480–59504, DOI: https://doi.org/10.1109/ACCESS.2025.3557405

    arXiv 2025

  30. [30]

    Recursive Subspace Identification of Lin- ear and Non-linear Wiener State-Space Models.Automatica2000,36, 1639–1650, DOI: https://doi.org/10.1016/S0005-1098(00)00103-5

    Lovera, M.; Gustafsson, T.; Verhaegen, M. Recursive Subspace Identification of Lin- ear and Non-linear Wiener State-Space Models.Automatica2000,36, 1639–1650, DOI: https://doi.org/10.1016/S0005-1098(00)00103-5

    work page doi:10.1016/s0005-1098(00)00103-5

  31. [31]

    Jiang, Y.; Fang, H. Recursive Subspace Identification Algorithm for Closed- loop Stochastic Systems.IFAC Proceedings Volumes2009,42, 116–121, DOI: https://doi.org/10.3182/20090706-3-FR-2004.00019, 15th IFAC Symposium on System Identifi- cation

    Pith/arXiv arXiv 2004

  32. [32]

    Weng, J.-H.; Loh, C.-H. Recursive Subspace Identification for On-line Tracking of Struc- tural Modal Parameter.Mechanical Systems and Signal Processing2011,25, 2923–2937, DOI: https://doi.org/10.1016/j.ymssp.2011.05.013. 21

    work page doi:10.1016/j.ymssp.2011.05.013 2011

  33. [33]

    Recursive Subspace Identification for in Flight Modal Analysis of Airplanes

    De Cock, K.; Mercère, G.; De Moor, B. Recursive Subspace Identification for in Flight Modal Analysis of Airplanes. International Conference on Noise and Vibration Engineering (ISMA 2006). Leuven, Belgium, 2006; p CDROM, HAL Id: hal-00097999, Available at https://hal.science/hal-00097999v1

    2006

  34. [34]

    Alenany, A.; Shang, H. Recursive Subspace Identification with Prior Information using the Con- strained Least Squares Approach.Computers & Chemical Engineering2013,54, 174–180, DOI: https://doi.org/10.1016/j.compchemeng.2013.03.016

    work page doi:10.1016/j.compchemeng.2013.03.016 2013

  35. [35]

    Change Detection in the Dynamics with Recursive Subspace Identification

    Oku, H.; Nijsse, G.; Verhaegen, M.; Verdult, V. Change Detection in the Dynamics with Recursive Subspace Identification. Proceedings of the 40th IEEE Conference on Decision and Control (Cat. No.01CH37228). 2001; pp 2297–2302, DOI:https://doi.org/10.1109/CDC.2001.980602

    arXiv 2001

  36. [36]

    Houtzager, I.; van Wingerden, J.-W.; Verhaegen, M. Recursive Predictor-Based Subspace Identification With Application to the Real-Time Closed-Loop Tracking of Flutter.IEEE Transactions on Control Systems Technology2012,20, 934–949, DOI:https://doi.org/10.1109/TCST.2011.2157694

    arXiv 2011

  37. [37]

    Hou, J.; Liu, T.; Wang, Q.-G. Recursive Subspace Identification subject to Relatively Slow Time-Varying Load Disturbance.International Journal of Control2017,91, 622–638, DOI: https://doi.org/10.1080/00207179.2017.1286538

    arXiv 2017

  38. [38]

    An Approach to Recursive Subspace Identification

    Bathelt, A.; Soffker, D.; Jelali, M. An Approach to Recursive Subspace Identification. 2017 IEEE 56th Annual Conference on Decision and Control (CDC). 2017; pp 4638–4643, DOI: https://doi.org/10.1109/CDC.2017.8264344

    arXiv 2017

  39. [39]

    Ramnath, K.; Narasimhan, S. Identification of Errors in Variables Linear State Space Models using Iterative Principal Component Analysis.International Journal of Control2023,96, 2773–2786, DOI: https://doi.org/10.1080/00207179.2022.2112089

    arXiv 2022

  40. [40]

    Recursive Iterative Principal Component Analysis.Computers & Chemical Engineering2025,201, 109170, DOI:https://doi.org/10.1016/j.compchemeng.2025.109170

    Pradeep, M.; Narasimhan, S. Recursive Iterative Principal Component Analysis.Computers & Chemical Engineering2025,201, 109170, DOI:https://doi.org/10.1016/j.compchemeng.2025.109170

    arXiv 2025

  41. [41]

    Abdi, H.; Williams, L. J. Principal Component Analysis.Wiley Interdisciplinary Reviews: Computa- tional Statistics2010,2, 433–459, DOI:https://doi.org/10.1002/wics.101

    doi:10.1002/wics.101

  42. [42]

    PCA-based Dimensionality Reduction Method for User Information in Universal Network

    Dai, Y.; Guan, J.; Quan, W.; Xu, C.; Zhang, H. PCA-based Dimensionality Reduction Method for User Information in Universal Network. 2012 IEEE 2nd International Conference on Cloud Computing and Intelligence Systems. 2012; pp 70–74, DOI:https://doi.org/10.1109/CCIS.2012.6664370

    arXiv 2012

  43. [43]

    T.Principal Component Analysis, 2nd ed.; Springer Series in Statistics; Springer: New York, NY, 2002; pp xxx + 488, DOI:https://doi.org/10.1007/b98835

    Jolliffe, I. T.Principal Component Analysis, 2nd ed.; Springer Series in Statistics; Springer: New York, NY, 2002; pp xxx + 488, DOI:https://doi.org/10.1007/b98835

    doi:10.1007/b98835 2002

  44. [44]

    Narasimhan, S.; Shah, S. L. Model Identification and Error Covariance Matrix Estima- tion from Noisy Data using PCA.Control Engineering Practice2008,16, 146–155, DOI: https://doi.org/10.1016/j.conengprac.2007.04.006

    work page doi:10.1016/j.conengprac.2007.04.006 2007

  45. [45]

    Recursive PCA for Adaptive Process Monitoring.Journal of Process Control2000,10, 471–486, DOI:https://doi.org/10.1016/S0959-1524(00)00022-6

    Li, W.; Yue, H.; Valle-Cervantes, S.; Qin, S. Recursive PCA for Adaptive Process Monitoring.Journal of Process Control2000,10, 471–486, DOI:https://doi.org/10.1016/S0959-1524(00)00022-6

    work page doi:10.1016/s0959-1524(00)00022-6

  46. [46]

    Golub, G. H. Some Modified Matrix Eigenvalue Problems.SIAM Review1973,15, 318–334, DOI: https://doi.org/10.1137/1015032

    work page doi:10.1137/1015032

  47. [47]

    R.; Nielsen, C

    Bunch, J. R.; Nielsen, C. P.; Sorensen, D. C. Rank-One Modification of the Symmetric Eigenproblem. Numerische Mathematik1978,31, 31–48, DOI:https://doi.org/10.1007/BF01396012

    doi:10.1007/bf01396012

  48. [48]

    D.; Andrews, D

    Wentzell, P. D.; Andrews, D. T.; Hamilton, D. C.; Faber, K.; Kowalski, B. R. Maximum Likelihood Principal Component Analysis.Journal of Chemometrics1997,11, 339–366, DOI: https://doi.org/10.1002/(SICI)1099-128X(199707)11:4<339::AID-CEM476>3.0.CO;2-L. 22

    work page doi:10.1002/(sici)1099-128x(199707)11:4

This paper was first reviewed by deepseek-v4-flash on August 1, 2026.