V — MSSP Resubmission Tracker (Consolidated)¶

Original submission: Mechanical Systems and Signal Processing (MSSP), 2026 Q1 — rejected with invitation to resubmit as a new paper after substantial revision. Resubmission path: Consolidated single paper targeting Engineering Structures (primary, Q1), with SHM SAGE and Smart Materials and Structures as backups. Target submission date: 2026-06-15.

See Papers / V for the scientific summary of the consolidated paper.

Progress¶

Consolidation confirmed · pipeline + benchmark + theory in single paper

Planning — 15 MSSP items, 10 V&V commissions Target: 2026-06-15 · Engineering Structures

Next steps:

V&V commission 1: re-run the 32-month compensation pipeline end-to-end; confirm numerics match the MSSP draft within ±0.5 %.
V&V commission 2: Johansen cointegration baseline implementation on the Gunsan feature matrix (new code — the load-bearing new engineering).
V&V commission 3: Gaussian-process regression baseline with leave-one-month-out tuning.
V&V commission 4: Gaussian-linear equivalence-proof derivation — if this fails, the paper reframes to four-method empirical comparison with a softer theoretical claim. Working draft at Workflow / V&V 6 — Equivalence proof.
Writing phase: §1–§3 theoretical core → §4–§5 pipeline and datasets → §6 benchmark → §7 field results → §8–§9 discussion and conclusions.
Pre-submission: acronym audit, reference validator sweep, one-command reproducibility script, cover letter drafting.

Consolidation decision (2026-04-18)¶

The split into V1 (JCSHM application) and V2 (SHM SAGE theory) was scrapped in favour of a single consolidated submission. The rationale:

What MSSP actually rejected. Two issues — methodology clarity (Reviewer 1) and novelty positioning (Reviewer 2). Both can be fixed in one paper. Splitting into two was a rhetorical solution to a clarity problem.

What the split would have cost. Roughly 4 weeks of writing and coordination effort: two introductions, two method sections, two response letters, companion-paper sequencing rules, duplicate-publication text mitigation, shared-dataset cross-citation management, and the risk that V2's central theoretical claim (the cointegration-equivalence proof) could collapse and leave V2 reframed to a weaker headline.

What consolidation buys. ~2.5 weeks of work instead of ~4. One submission cycle. Stronger scope per paper. One first-author paper with both the application depth and the theoretical derivation — and the equivalence-proof content positioned as a load-bearing contribution within that paper rather than the headline of a standalone paper whose failure mode is catastrophic. Engineering Structures (Q1, IF ~5.5, broad audience) avoids Ocean Engineering editor overlap with J2 and J3.

Trade-off summary:

Dimension	Split (rejected)	Consolidate (chosen)
Writing effort	~4 weeks	~2.5 weeks
Submission cycles	2	1
Risk diversification	Two journals, two chances	One chance in a stronger paper
V2 theoretical-proof risk	Load-bearing — if proof fails, V2 collapses	Absorbed — equivalence proof is a contribution, not the headline
Defense timing (2026-09-03)	Both under review (best case)	One under review by defense; more polish time
Duplicate-publication coordination	Required (shared dataset, cover-letter cross-refs, ~500 shared words)	Not needed
CV output	2 first-author papers	1 first-author paper of larger scope

Why Engineering Structures as primary target. Q1, IF ~5.5, accepts both methodology and application papers, broad offshore / structural audience, no editor overlap with Ocean Engineering. SHM SAGE and Smart Materials and Structures remain backups if Engineering Structures declines the scope fit.

Editor summary of the MSSP rejection¶

The editor's letter concurred with both reviewers and declined the paper in its current form. The rejection specifically allowed resubmission as a new paper with a substantial revision, a point-by-point reply, and modified text in a different colour. Both reviewers identified issues that — addressed together in a single paper — expand rather than fragment the scope.

Reviewer 1 (MSSP) — methodology clarity¶

Reviewer 1 clustered concerns around signal-processing exposition insufficient for MSSP's methodology bar: acronyms defined only in the abstract, feature-matrix definition scattered, block-Hankel / model-order treatment too brief, sensor orientation undefined in the figure, displacement-estimation under noise unexplained. Mostly clarity and completeness issues rather than fundamental method problems.

ID	Paraphrased concern	Where addressed	Action	Status
R1.a	Acronyms CUMSUM, RANSAC, MAC defined only in abstract, not main text	§1, §4	Define at first main-text use; glossary confirmed complete.	Planning
R1.b	Sensor sampling rates (1,500 Hz accelerometer, 500 Hz strain) unclear	§5.1	Replace "disparate rates" with explicit sampling-rate statement.	Planning
R1.c	Sensor orientation not specified; Fig 1(b) lacks Cartesian basis	§5.1 Fig 1(b)	Add N arrow, \(x/y/z\) triads, orientation of each triaxial accelerometer and strain gauge.	Planning
R1.d	Feature matrix not clearly defined; Table 4 too far from first mention	§5.1	Move feature-matrix definition near first mention.	Planning
R1.e	Figure 2 insufficiently explained	Pipeline figure caption + paragraph	Expand caption; add paragraph introducing it.	Planning
R1.f	"Native rates" ambiguous	Throughout	Replace with "instrument sampling rates".	Planning
R1.g	Displacement estimation from noisy accelerometer not explained	§4 new subsection	Document double-integration + Butterworth high-pass at 0.05 Hz + drift validation.	Planning
R1.h	SSI-COV under-described: block Hankel, model order	§4	Cite Peeters & De Roeck (1999) and Reynders (2012); develop one-paragraph summary of block Hankel construction and model-order selection. Consolidated paper can afford deeper treatment than a JCSHM-only version.	Planning
R1.i	MAC introduced with \(\phi_{\text{FE}}\) without definition	§4	Define MAC + \(\phi_{\text{FE}}\) at first use.	Planning
R1.j	Punctuation missing in equations	Throughout	Equation-punctuation sweep.	Planning
R1.k	Eq 2 environmental state vector components undefined	§2	Add component listing.	Planning
R1.l	Introduction lacks final paragraph outlining paper structure	§1	Add outline paragraph at end of §1.	Planning

Estimated effort on Reviewer 1 items: ~8 hours writing, mostly ≤ 10-line edits.

Reviewer 2 (MSSP) — novelty positioning, baselines, validation¶

Reviewer 2 judged the manuscript interesting but requested clearer novelty positioning relative to cointegration and physics-informed normalisation, deeper baseline benchmarking, and explicit acknowledgement of the synthetic damage-injection limitation.

ID	Paraphrased concern	Where addressed	Action	Status
R2.a	Novelty positioning vs. cointegration and physics-informed normalisation unclear	§1 + §2–§3	Introduction rewrites final paragraph with explicit four-bullet contribution statement citing Cross & Worden (2013) and Figueiredo (2010). §2 derives state-function from conservation arguments; §3 proves Johansen equivalence with explicit conditions and three counterexamples. The consolidated paper carries the theoretical positioning as a load-bearing section rather than a passing paragraph.	Planning
R2.b	Synthetic damage injection limitation	§8	Dedicated paragraph on injection-protocol limitations: linear approximation of nonlinear scour regime; J2 Winkler power-law as injection source; real scour is slower and spatially asymmetric. Supplemented by operational regime-shift stress tests using real-world transitions.	Planning
R2.c	PCA baseline benchmarking depth limited	§6	Full four-method comparison (state-function + cointegration + PCA + GP regression) on two datasets with ROC, F1, precision-recall, false-alarm curves per method.	Planning

V&V commissions (consolidated path)¶

Ten discrete sessions; roughly 5–8 working days of engineering before drafting.

Re-run the 32-month compensation pipeline end-to-end; confirm ±0.5 % against MSSP draft numerics.
PCA grid-search baseline (\(n_{\text{components}} \in \{2, 3, 5, 8, 10\}\)) with F1 and ROC per setting.
Johansen cointegration baseline implementation — new code on the Gunsan feature matrix under identical preprocessing and damage-injection protocol.
Gaussian-process regression baseline — GP hyperparameters tuned by leave-one-month-out cross-validation.
Displacement-estimation noise-floor validation — Butterworth order-4 high-pass at 0.05 Hz; residual drift < 1 mm over 10-minute window.
Gaussian-linear equivalence-proof derivation — LaTeX draft, peer-review-ready. Load-bearing: if the derivation fails, the paper reframes to four-method empirical comparison with a softer theoretical claim.
Three explicit counterexamples where state-function strictly outperforms cointegration: regime-shift boundary (startup/shutdown transients), non-Gaussian noise (heavy-tailed residuals), rank-deficient environmental vector.
Public SHM benchmark acquisition — Z-24 Bridge candidate; licensing; preprocessing; damage events labelled.
Acronym audit; CrossRef DOI validator sweep on references.
One-command reproducibility script covering all four baselines + synthetic injection + field result reproduction.

What has not changed from the MSSP draft¶

Core 32-month Gunsan dataset (22,617 parked-state windows).
Core compensation pipeline: regime-split RANSAC normalisation + exponentially weighted persistence filter (span 48).
Reported zero-false-alarm performance and 95 % detection probability at 0.2 % injected shift.
Authorship, affiliations, funding acknowledgements.

The revision is a repackaging and deepening, not a retraction of any prior claim.

Raw reviewer comments (verbatim from MSSP decision letter, MSSP26-1185)¶

Editor's rejection letter

Title: Double-Filter Framework for Physics-Informed Vibration-Based Scour Detection in Offshore Wind Turbine Foundations

Ref: MSSP26-1185

Dear Professor Kim,

Your manuscript has now been reviewed and the reviewers' comments are appended below for your attention. I have also evaluated your paper and concur with the reviewers' comments. I am afraid that I cannot recommend your paper for publication in Mechanical Systems and Signal Processing in its present form.

However, if you feel that you are able to overcome the reviewers' objections in a significant revision, then a re-submission as a new paper will be considered. In that case, please mention in the covering letter that it is a re-submission of paper MSSP26-1185 and include a point-by-point reply to all the reviewers' comments. Modified text should be printed in different color.

Yours sincerely, Mechanical Systems and Signal Processing

Reviewer #1 — verbatim

The manuscript MSSP26-1185 proposes a methodology for structural monitoring in wind turbines based on a Double-Filter framework combined with signal processing and physics-informed strategies. The topic is relevant to the scope of MSSP and addresses important challenges in the context of environmental and operational variability.

Nonetheless, the reviewer has several concerns regarding the clarity, completeness, and rigor of the manuscript, particularly in the description of the signal processing methodology and the overall contextualization of the work within the existing literature. These comments are detailed below:

In the introduction, adding equations in plain text is not appropriate, as some terms are not yet defined.
Acronyms for CUMSUM and RANSAC are not properly introduced in the main text (they appear only in the abstract).
The authors do a good job positioning their work within the context of wind turbines and clearly identifying research gaps. However, there is a lack of discussion on how the methodologies employed in this work (e.g., RANSAC) have been previously used in the context of EOV. As a result, despite a significant number of citations, the related work section does not sufficiently contextualize the proposed approach.
In the introduction, the concept of the Double-Filter framework becomes clear. However, this clarity is lacking in the abstract, which could be improved accordingly.
It is not clear what type of physics-informed constraint is used, nor whether similar constraints have been considered in related works. More generally, the final paragraph of the introduction, which is intended to describe the proposed solution to the identified research gaps, is too concise and reads more like a list of tasks than a properly contextualized description of the contribution.
The introduction would benefit from a final paragraph outlining the structure of the paper.

Section 2

In Section 2.1.2, when presenting the monitoring framework, the meaning of "disparate rates" is unclear. Additionally, the mention of accelerometers and strain gauges at 1500 Hz and 500 Hz, respectively, requires clarification (e.g., do these values correspond to sampling rates or maximum frequency ranges?).
When describing the strain gauges and accelerometers, their orientations should be specified (e.g., triaxial accelerometers or sensors aligned along specific directions). If directional measurements are involved, a Cartesian basis should be defined in Figure 1(b), or a schematic representation of the structure and sensor placement should be provided.
In Section 2.1.2, the feature matrix is not clearly defined. Furthermore, the description of the associated parameters is too distant from their introduction (page 7), with Table 4 only appearing on page 19. The MAC here introduced is also not defined, being its description given only later on the paper.
Figure 2 is not clearly explained.
The term "native rates" should be clarified (does it refer to sampling rates?).
The authors state that "[...] enables valid fusion of acceleration-derived displacement responses [...]". However, it is not explained how displacement estimation is performed in the presence of measurement noise.
The description of the covariance-driven stochastic subspace identification method is insufficient. It is unclear how the measured data are used, how the block Hankel matrix is constructed, and how the model order is selected or interpreted. For a journal such as MSSP, this signal processing component is not adequately described.
When defining the MAC, the term \(\phi_{FE}\) is introduced without definition.
Punctuation is missing in several equations.
In Equation 2, the environmental state vector is introduced without defining its components.

Due to the issues listed above, particularly those appearing at an early stage of the manuscript, the reviewer recommends rejection of the paper, with the possibility of resubmission after a substantial revision to fit within the MSSP standards. In particular, the signal processing methodology must be described in a clearer and more rigorous manner. The figures are of good quality, and the paper shows potential; however, in its current form, it is not possible to recommend even major revisions.

Reviewer #2 — verbatim

The manuscript proposes a physics-informed double-filter framework for vibration-based scour detection in offshore wind turbine foundations. The use of a long-term 32-month field dataset is a significant strength, and the integration of physics-based modeling with statistical detection techniques is timely and relevant to the structural health monitoring community.

Comments:

1. The manuscript presents an interesting approach, but the novelty and positioning relative to existing literature need to be more clearly articulated. While the separation of environmental variability from damage is a well-established problem, the authors introduce a state-function formulation and a double-filter architecture as key contributions. However, it remains somewhat unclear whether the main novelty lies in the theoretical formulation, the specific regression framework, or the overall pipeline integration. A clearer and more explicit statement of contributions would significantly strengthen the paper and help distinguish it from prior work on cointegration methods and physics-informed normalization approaches.

2. The availability of a 32-month field dataset is a major strength of the study; however, the validation relies heavily on synthetic damage injection due to the absence of direct ground truth measurements of scour. While this is understandable, the manuscript should more explicitly discuss the limitations of this approach. In particular, it is unclear how well the injected frequency shifts represent realistic scour evolution, and whether the detection performance would remain as strong under more complex or nonlinear damage scenarios. Any available indirect validation or comparison with field observations would strengthen the results.

3. The reported performance improvements, including significant variance reduction and high detection probability, are promising, but the benchmarking against baseline methods could be more rigorous. The description of the PCA-based comparison is relatively limited, and it is not clear whether competing methods were optimally tuned.

Where this links in the rest of the portfolio¶

Scientific page: Papers / V.
Downstream: the V compensation pipeline (residual stream) feeds Paper A's three-channel Bayesian fusion as the statistical-detection channel. A's decision layer depends on V's compensator selection.
Cross-cuts: V's four-method benchmark uses Paper B's candidate feature set and Paper J2's Winkler power-law for synthetic damage injection.