Geng Report: 'Prediction of Horizontal Deformation of Retaining Piles in Deep Excavation Based on Modal Decomposition Methods' (DOI: 10.16285/j.rsm.2023.1204)

Academic fraud report · Geng Detector

Summary

This Geng (耿) academic-integrity report raises high suspicion regarding a 2024 supplement paper in Rock and Soil Mechanics (岩土力学) by Li Tao, Shu Jiajun, Wang Yanlong, and Chen Qian. Although no image reuse or Photoshop splicing was detected (the source PDF provided only text), the report identifies multiple severe textual inconsistencies. The paper's title promises 'modal decomposition methods' (e.g., EMD/VMD), but the actual methodology is an AM-CNN-LSTM (Attention + CNN + LSTM) hybrid; no decomposition technique is mentioned anywhere in the body. The abstract cites a single Beijing deep excavation project, while the introduction and Section 3 reference three distinct sites (Jijiamiao, Yunjing East Road, Tongzhou New City). Most critically, the Conclusion duplicates identical RMSE (0.73, 0.70, 1.01 mm) and MAE (0.62, 0.60, 0.84 mm) figures for both 'maximum deformation' and 'single-point continuous deformation', whereas Section 4.2 actually shows RMSE of 0.59/0.67/0.68 and MAE of 0.48/0.56/0.57. The journal accepted the manuscript only 40 days after submission (2023-08-10 to 2023-09-20), suggesting minimal peer review. These patterns strongly indicate careless writing, possible template reuse, or batch production behavior. Confidence is moderate-high; final adjudication requires institutional investigation.

Verdict

🟠 Highly suspicious. No image-based fraud was detectable from the supplied text-only PDF, but the manuscript exhibits pervasive copy-paste artefacts, severe title-content mismatch, and internally contradictory numerical results, consistent with negligent drafting, template reuse, or commercial supplement production.

Key findings

🔴 Title–content mismatch (Issue 6): The title announces 'modal decomposition methods', yet the full text exclusively describes an AM-CNN-LSTM model. No occurrence of any modal decomposition technique (EMD, VMD, CEEMDAN, etc.) exists outside the title and English-title.
🟠 Inconsistent research scope (Issue 6): The Chinese abstract specifies a single Beijing deep excavation ('北京地区某深基坑工程'), whereas the introduction and main text explicitly name three projects: 纪家庙深基坑, 云景东路深基坑, and 通州新城深基坑.
🔴 Copy-paste error in the Conclusion (Issues 2/4): Conclusion item (3) reports identical performance metrics for two different experiments — maximum deformation prediction vs. single-point continuous deformation prediction — both stated as RMSE = 0.73, 0.70, 1.01 mm and MAE = 0.62, 0.60, 0.84 mm. Section 4.2 actually reports RMSE = 0.59, 0.67, 0.68 and MAE = 0.48, 0.56, 0.57 for the single-point continuous deformation case.
🟡 Abnormally rapid turnaround (Issue 5): Received 2023-08-10, accepted 2023-09-20 — a 40-day interval for a core Chinese geotechnical journal (even for a supplement), implying minimal external review and minimal proofreading.

Evidence highlights

Full-text keyword scan for '模态分解' / 'decomposition' yields hits only in the title, English title, and English abstract heading; the body text references solely CNN, LSTM, and Attention mechanisms.
Abstract: '以北京地区某深基坑工程为背景' (single project).
Introduction final paragraph: '以北京地区纪家庙深基坑、云景东路深基坑和通州新城深基坑工程为背景' (three projects).
Conclusion (3), first clause: '其桩体最大变形预测值的RMSE分别为0.73、0.70、1.01 mm，MAE为0.62、0.60、0.84 mm'.
Conclusion (3), second clause: '单点连续变形预测值的RMSE分别为0.73、0.70、1.01 mm，MAE为0.62、0.60、0.84 mm'.
Section 4.2 (true values for single-point continuous deformation): RMSE = 0.59, 0.67, 0.68; MAE = 0.48, 0.56, 0.57.
DOI: 10.16285/j.rsm.2023.1204; dates: received 2023-08-10, accepted 2023-09-20.

Notes

Detection limitation: only text was supplied, so pixel-level image-duplication (Issue 1) and splicing (Issue 3) analyses could not be performed.
The combination of a 40-day review window with multiple unresolved internal contradictions strongly suggests the supplement issue underwent limited external scrutiny.
Final determination of academic misconduct requires an official inquiry by the journal or the authors' institution; the publisher should be urged to issue a corrigendum addressing at minimum the conclusion's duplicated metrics and the title–content mismatch.