Investigation Report: 'Joint multi-objective optimization based on multitask and multi-fidelity Gaussian processes for flapping foil' (Ocean Engineering, 2024)

Academic fraud report · Geng Detector

Summary

This report assesses the 2024 Ocean Engineering paper by Wang, Yan, Zeng, Li, Cui, Liang, and Fan (DOI: 10.1016/j.oceaneng.2024.116862) and returns a verdict of 'Questionable' (存疑). Two substantive issues were verified. First, in Appendix A (Table A.5 surrounding text), the authors state that as the time step is reduced from 1.5dt to dt, the thrust coefficient rises from 1.232 to 1.274, while the lift coefficient 'decreases from 0.0144 to 0.0242'. Numerically, 0.0242 is greater than 0.0144, a clear internal contradiction between the descriptive text and the reported values. Second, the Introduction contains multiple malformed citations such as '(Wang et al., ; , )' and the fragment 'A comprehensive review by 2022)', indicating failed reference-manager rendering that escaped author and reviewer attention. A third concern — potential image or data fabrication in figures such as MAE convergence curves and Pareto fronts — could not be evaluated because high-resolution originals and CFD computation logs (reportedly 300 high-fidelity samples, 8 h each) were not available. The findings warrant editorial follow-up and requests for raw data and source code.

Verdict

🟡 Questionable — verified internal inconsistencies in text and reference formatting; image/data integrity could not be assessed.

Key findings

Mathematical/textual contradiction in Appendix A: Authors claim the lift coefficient 'decreases from 0.0144 to 0.0242' when the time step is reduced from 1.5dt to dt, but 0.0242 > 0.0144. The thrust coefficient description (1.232 → 1.274) is internally consistent, isolating the error to the lift-coefficient wording.
Severely corrupted citation formatting in the Introduction: Multiple in-text citations appear as (Wang et al., ; , ), (Wu et al., ; , ; , ), and the fragment A comprehensive review by 2022), suggesting a literature-manager rendering failure that was not caught during author proofreading or peer review.
Inability to verify figures and underlying CFD/ML data: Figures 1 through B.18 (vorticity fields, MAE convergence curves, Pareto fronts) could not be examined at pixel level; raw CFD logs and code for the 300-sample, 8-hour-per-sample high-fidelity runs were not provided.

Evidence highlights

Appendix A, Table A.5 surrounding paragraph: stated lift coefficient trajectory 0.0144 → 0.0242 described as a 'decrease', contradicting basic numerical comparison.
Introduction (pp. 1–2): at least three visibly broken citation strings with missing authors/years and stray punctuation.
Reported computational scope: ~300 high-fidelity CFD samples × ~8 h each, motivating a request for raw outputs and source code to substantiate ML training data.

Notes

The lift/thrust wording error may stem from a draft not updated after numerical revision; however, its presence in the published Appendix indicates inadequate proofreading.
Broken citations, while not equivalent to data fabrication, signal weak editorial and authorial quality control.
Confidence in findings 1 and 2: high (text directly verifiable). Confidence in finding 3: indeterminate due to absence of source images and data.
Recommended actions: post a PubPeer comment citing both issues; request raw CFD logs and code from the authors; notify the journal editorial office about reference-rendering quality.
DOI preserved as reported: 10.1016/j.oceaneng.2024.116862.