Geng academic fraud detection report: 'Multi-sensor fusion for high-precision air-gap estimation in electromagnetic maglev systems: an extended Kalman filter approach' (Zhang et al., 2026)

Academic fraud report · Geng Detector

Summary

Verdict: Strong evidence of serious scholarly misconduct (red-flagged, '实锤' status). The Detection report identifies multiple internal contradictions in the paper published in Measurement Science and Technology (DOI: 10.1088/1361-6501/ae4e03). Most critically, the abstract and conclusion report an air-gap RMSE of 0.12 mm and velocity RMSE of 4.2 mm s−1, while Section 8.3 and Table 2 give 0.182 mm and 4.860 mm s−1 respectively — an irreconcilable discrepancy suggesting fabricated headline figures. Additional issues include physically inconsistent sensor equations (mixing mm and m units, producing negative bases raised to a fractional power), a misattribution of zero-mean Gaussian noise as causing a 16% DC shift in inductance, identical acceleration RMSE (0.022) before and after fusion, and overstatement of practical applicability despite validation being entirely simulation-based. Confidence is high for the numerical contradictions and the unit/equation errors; image-based duplication checks were not performed due to lack of image input. The report warrants formal editorial investigation.

Verdict

🔴 Strong evidence of serious academic misconduct. Multiple irreconcilable internal contradictions in the core reported numbers and the underlying mathematical model. Formal investigation by the journal is recommended.

Key findings

Fabricated/inflated headline metrics: The Abstract and Conclusion (Section 9) claim a fusion EKF air-gap RMSE of 0.12 mm (33% reduction from 0.18 mm) and velocity RMSE of 4.2 mm s−1. Section 8.3 reports an RMSE of 0.1816 mm, and Table 2 reports Airgap RMSE = 0.182 mm and Velocity RMSE = 4.860 mm s−1. The headline numbers cannot be obtained from the paper's own table.
Physical/mathematical inconsistency in sensor model: Equation (16) uses a Hall-sensor fit B(g) = 1.09 / (g − 0.634)^0.445 − 0.151. With g in metres in the stated 2–8 mm range, (g − 0.634) becomes negative, and a negative base raised to the 0.445 power is undefined in the reals. Subsequent equations (17, 22) appear to mix mm and m, and the correction term exp(−50·(gk − 0.001)) collapses to ~0 if gk is in mm.
Misinterpretation of zero-mean Gaussian noise: Section 3.2 adds zero-mean Gaussian noise to the inductive sensor model, yet reports that mean measured inductance drops from 283.4 µH to 238.1 µH (≈16%) at a 2 mm air gap, attributed to 'magnetic saturation'. Zero-mean noise does not shift a DC mean.
Implausibly identical fusion result: Table 2 gives RMSE(a) = 0.022 for the accelerometer-only EKF and 0.022 for the multi-sensor fusion EKF, while the conclusion still claims 'smoother temporal behaviour and improved consistency' for the fused acceleration — i.e., fusion yields no measurable benefit on this channel.
Simulation-only validation misrepresented as practical: Abstract and Introduction frame the method as a 'practical and economical sensing solution… readily adaptable'; only the conclusion discloses that 'the current validation relies entirely on simulation data'. No hardware validation is presented.

Evidence highlights

Abstract vs. Table 2: 0.12 mm (Abstract) vs. 0.182 mm (Table 2) — air-gap RMSE.
Conclusion vs. Table 2: 4.2 mm s−1 (Conclusion) vs. 4.860 mm s−1 (Table 2) — velocity RMSE.
Section 8.3 narrative: 'near-overlapping trajectories and an RMSE of 0.1816 mm'.
Equation (16): fractional-power fit with g − 0.634 where g in SI units yields a negative base.
Table 2: RMSE(a) = 0.022 in both single- and multi-sensor rows.
Section 3.2: 283.4 µH → 238.1 µH mean shift under zero-mean Gaussian noise.
Conclusion caveat: 'the current validation relies entirely on simulation data'.

Notes

Image duplication and splicing checks were not performed; no source images were provided for analysis.
All numeric values above are quoted directly from the report and were not independently re-verified against the published PDF.
The report itself disclaims that final determination of misconduct requires an institutional or editorial investigation; findings here are presented as red flags, not a final adjudication.
DOI preserved exactly: 10.1088/1361-6501/ae4e03.