Investigation Report: 'Accurate Detection for Zirconium Sheet Surface Scratches Based on Visible Light Images' (Sensors, 2023)

Academic fraud report · Geng Detector

Summary

Verdict: Substantiated misconduct (cattle-sheep substitution). The paper by Bin Xu et al. advertises a method specifically designed for zirconium sheet surface scratch detection and extensively justifies zirconium's unique physical/chemical properties in the introduction. However, Section 3 (Experiments) states that aluminum plate was used in place of zirconium sheet. Because aluminum and zirconium differ markedly in reflectivity, texture, and hardness, accuracy figures (reported as >85%, with selected values such as 87.44% and 92.17%) cannot be generalized to zirconium. Secondary concerns include an implausible data-availability refusal citing 'personal privacy' for industrial metal-surface images, a 36-day submission-to-acceptance window indicative of superficial peer review, and the absence of any statistical inference (no confidence intervals, standard deviations, or significance tests). Pixel-level image forensics was not performed because raw data were withheld. Confidence is high on the material-substitution issue based on a direct textual admission; statistical and review-quality concerns are medium confidence.

Verdict

Substantiated misconduct. The paper openly admits in its Experiments section to substituting aluminum plate for the zirconium sheet that the title, abstract, and introduction claim to address. This constitutes a fundamental mismatch between the claimed subject and the actual experimental material, invalidating the paper's stated conclusions on zirconium-specific detection accuracy. Additional concerns about data availability and statistical rigor compound the problem.

Key findings

Material substitution (severe): Aluminum plate was used in place of zirconium sheet in the experiments, despite the title and rationale emphasizing zirconium's unique properties.
Implausible data-availability refusal: Authors cite 'personal privacy' as the reason for withholding industrial metal-surface image data—an inappropriate justification that impedes independent verification.
Image integrity unverified: Figures 5–11 (scratch images and 3D grayscale maps) could not be subjected to source or Photoshop-level analysis because raw data were not released and metadata are insufficient; given the material fraud, authenticity of these figures is suspect.
Abnormally short review cycle: Received 12 July 2023; accepted 17 August 2023 (36 days), consistent with a cursory peer-review process that failed to catch the material mismatch.
Statistical deficiencies: Tables 1 and 2 report only single-run percentages (e.g., 87.44%, 92.17%) with no confidence intervals, standard deviations, or significance tests; claims of robustness are therefore unsupported.

Evidence highlights

Direct textual admission, Section 3 (Experiments, p. 8): *"Among them, aluminum plate is used to replace zirconium sheet."*
Data Availability Statement (p. 14): *"The data are not publicly available due to personal privacy."*—applied to industrial metal-surface imagery.
Timeline: Received 12 July 2023 → Accepted 17 August 2023 (36 days).
DOI: 10.3390/s23167291. Reported accuracy values include 87.44% and 92.17% in Table 2, presented without error bars or replication statistics.

Notes

The substitution is the authors' own written statement, so confidence in this finding is high.
The 'personal privacy' rationale is evaluated on its face as implausible for non-human imagery; intent cannot be inferred from the text alone.
Pixel-level image forensics were not conducted due to withheld data; conclusions about image manipulation are tentative.
Rapid MDPI review timelines are not inherently invalid but, in combination with the missed material mismatch, suggest inadequate editorial scrutiny in this case.
Final determination of misconduct requires investigation by the publisher (MDPI / Sensors editorial board) and/or the authors' institution.