Academic Fraud Report: Stochastic Biomechanical Modeling of Human-Powered Electricity Generation

Academic fraud report · Geng Detector

Summary

Verdict: 🔴 Confirmed (实锤). This report identifies severe scientific and methodological flaws in the MDPI paper by Qirui Ding and Weicheng Cui (DOI: 10.3390/en18184821) titled 'Stochastic Biomechanical Modeling of Human-Powered Electricity Generation.' The study claims a metabolic efficiency of η_metabolic = 0.891 ± 0.016, which exceeds the well-established thermodynamic and physiological limit of 20–25% for human skeletal muscle, directly contradicting the authors' own cited literature. Additional inconsistencies include a 'Maximum Power' prototype producing only 175.5 W versus 316 ± 19 W for the 'Balanced' variant, conflicting time constants (τ_peripheral = 1347 s vs 1358 s) between Sections 2.3 and 4.4, unexplained arithmetic in trial counts (623 trials vs 267 × 7 points), and multiple RMSE values (3.52 W, 8.3 W, 23.7 W). Several references lack authors or titles, suggesting AI-generated citations. Limitations: no image analysis performed, and certain reference-related findings await stronger evidence.

Verdict

🔴 Confirmed — The paper exhibits fundamental thermodynamic violations, internal mathematical contradictions, and likely fabricated numerical outputs. The methodological and logical inconsistencies are severe enough to warrant an editorial investigation and potential retraction.

Key findings

Thermodynamic impossibility: Reported metabolic efficiency (η_metabolic = 0.891 ± 0.016) exceeds the established physiological limit of 20–25% for human skeletal muscle, contradicting the authors' own cited reference (Gad et al., 25%).
Logical inversion of optimality criteria: Table 1's 'Maximum Power' prototype outputs 175.5 W, which is lower than the 'Balanced Performance' prototype (316 ± 19 W).
Inconsistent fatigue-model parameters: Section 2.3 gives τ_peripheral = 1347 ± 89 s and τ_central = 3124 ± 178 s, while Section 4.4 gives τ_p = 1358 ± 94 s and τ_c = 3187 ± 167 s for the same dataset.
Arithmetic incoherence: The text states '623 experimental trials generated 1869 individual measurement points (267 × 7 participants)', yet 623/7 ≈ 89 (trials per participant), not 267. The relationship between 623, 267, 7, and 1869 is unexplained.
Conflicting RMSE values: Section 4.1 and the abstract report RMSE = 3.52 W, Appendix B cites 8.3 W, and Appendix B also mentions a reduction 'from 23.7 W to 8.3 W'.
Trial-count discrepancy: Main text claims 623 trials, while Appendix B mentions 89 experimental trials for validation.
Incomplete or ghost references: Entries such as [1] Nat. Rev. Methods Primers 2023, 3, 39; [8] Science 2006, 312, 242–246; [33] Energies 2017, 10, 1483; [36] Cogn. Comput. 2025, 17, 4; and [58] (SIAM 2024) lack author names and/or titles, consistent with LLM-generated citations.

Evidence highlights

Table 1 (Section 4.5): Maximum Power Power = 175.5 W < Balanced Performance Power = 316 ± 19 W.
Table 2 / Section 4.6 / 4.7: η_metabolic = η_bio = 0.891 ± 0.016.
Section 2.3 vs Section 4.4: τ_peripheral 1347 → 1358; τ_central 3124 → 3187.
Section 4.1: 623 trials / 1869 points (267 × 7) — arithmetic unclear.
Appendix B / Table A3: Values of 8.3 ± 1.7 W vs text's 3.52 W vs 23.7 W.
DOI: 10.3390/en18184821.

Notes

No images were available for analysis, so first- and third-form image duplication or photoshopping checks could not be performed.
The reference-related findings (Finding 5) are marked as ⚠️ Insufficient evidence pending manual verification against databases (Scopus, Web of Science, Google Scholar).
The arithmetic discrepancy in Finding 3 is reproducible from the text but the authors may provide an alternative explanation.
All verdicts are preliminary; formal determination requires institutional investigation.