Forensic Review: Numerical Simulations of Natural Convection in a Cylindrical Envelope with Slotted Internal Cylinder (Zhang, Yang, Zhang, 2014)

Academic fraud report · Geng Detector

Summary

This forensic assessment examines a 2014 Technical Note in the International Journal of Heat and Mass Transfer (DOI: 10.1016/j.ijheatmasstransfer.2013.11.015) authored by Kun Zhang, Mo Yang, and Yuwen Zhang, concerning 2D and 3D CFD simulations of natural convection. The overall verdict is that no evidence of academic fraud was identified. Four independent checks were performed: (1) last-digit distribution analysis of equivalent thermal conductivity values in Table 1 showed natural variation (0,1,2,3,4,9), inconsistent with fabricated data; (2) reported computational times (7.8 s for 2D, 112.5 s for 3D) are physically plausible for the stated grid sizes on circa-2013 hardware; (3) citation of a 2014 reference ([9]) is explained by Elsevier's Article-in-Press workflow and is internally consistent; (4) the authors honestly acknowledge 2D model limitations at high Rayleigh numbers rather than overstating agreement with experiment. Limitations: only the text was reviewed; no pixel-level image analysis was possible, and no raw simulation output was independently verified.

Verdict

Clean (no fraud indicators detected). The paper passes all textual forensic checks applied. No manipulation, fabrication, or suspicious patterns were found. Pixel-level image forensics could not be performed because the original figures were not available for inspection.

Key findings

Last-digit distribution of Table 1 data is natural. Sampled values (2.299, 3.261, 3.770, 4.134, 4.549, 6.103, 6.999, 7.672, 8.372) end in 0,1,2,3,4,9 with no signs of rounding bias, arithmetic progression, or generator-style regularity.
Monotonic physical trend preserved. Equivalent thermal conductivity K_eq increases monotonically with Rayleigh number, consistent with convection-enhanced transport.
Plausible computational cost disclosure. Reported wall times of 7.8 s (2D, grid 52×102) and 112.5 s (3D, grid 52×52×102) scale roughly with the grid-volume ratio (~14×), matching expectations for finite-volume transient solvers of that era.
Self-consistent citation timeline. Reference [9], listed as 2014, is an in-press companion experimental paper by the same group; authors explicitly state the experimental apparatus is detailed elsewhere [9].
Intellectual honesty in validation. Authors openly state that 2D simulations fail to capture the nonlinear phenomenon observed experimentally and that the critical Rayleigh number is lower in 3D, rather than force-fitting 2D results to experiment.

Evidence highlights

Table 1 (p. 437): K_eq values at varying slot angles θ and Ra numbers; last-digit set {0,1,2,3,4,9}.
Section 3 (p. 436): "The computational time for two- and three-dimensional simulations is 7.8 and 112.5 s, respectively."
Section 3 (p. 436): Authors' admission that 2D solutions "can also not predict the non-linear phenomenon in our experiment," with a lower critical Rayleigh number in 3D.
Reference [9]: Cited as published in 2014 within a paper accepted November 2013, explained by Article-in-Press status.

Notes

Scope of review: Textual content only. Original numerical output, mesh files, and high-resolution figures were not available, so no pixel-level or statistical reproducibility check was conducted.
Confidence: High for absence of textual fraud indicators; medium for absence of figure-based fraud, given the figure-level review was not possible.
Recommendation: No further action warranted on the basis of the materials reviewed. A formal institutional investigation would be needed to reach a definitive determination, and this tool explicitly does not claim certainty.