Research misconduct in engineering is fabrication, falsification, or plagiarism involving structural test data, materials properties, or simulation results — but unlike biomedical misconduct, it is judged against physical, re-testable evidence and can trigger professional-licence discipline, not just retraction. Where biomedical fraud is typically uncovered through statistical forensics on data that cannot be re-run on the same human subjects, engineering fraud is frequently caught — and proven — by re-testing the actual material, structure, or component in question.
Most misconduct coverage focuses on biomedicine and psychology, where retractions and clinical-trial scandals dominate. Engineering misconduct is less visible but carries a distinct evidentiary logic, a distinct enforcement path through licensing bodies, and a distinct risk profile: infrastructure and physical safety rather than patient health. This analysis sets out what makes engineering cases different.
- What is research misconduct in engineering?
- Why engineering’s evidentiary standards differ from biomedicine’s
- Types of misconduct most relevant to engineering research
- Structural-testing and materials-data fraud: four real cases
- Common questions and what comes next
What Is Research Misconduct in Engineering?
Research misconduct is fabrication, falsification, or plagiarism (FFP) in proposing, performing, or reviewing research, or in reporting results, as defined by the US Office of Research Integrity (ORI), the federal body overseeing Public Health Service-funded research. This definition applies without modification to engineering research; what changes is the object being faked.
In engineering, misconduct usually touches quantifiable physical properties: yield strength, fatigue-cycle counts, thermal tolerance, load-bearing capacity, corrosion resistance, or finite-element simulation outputs. A fabricated result in this domain is not an abstract statistical artefact — it is a claim about whether a material or structure will hold under real load. That distinction shapes every part of how allegations are investigated and resolved.
The Committee on Publication Ethics (COPE) supplies the procedural backbone most engineering journals use to handle allegations, via its published core practices and case-specific flowcharts covering fabricated data, image manipulation, plagiarism, authorship disputes, and undisclosed conflicts of interest.
Why Engineering’s Evidentiary Standards Differ From Biomedicine’s
Engineering misconduct investigations lean on re-testable physical evidence far more than biomedical ones do. A disputed tensile-strength figure can, in principle, be checked by re-machining a sample and re-running the test rig; a disputed clinical-trial outcome in a now-treated or deceased patient population usually cannot be re-run at all. This single fact reshapes the entire evidentiary standard.
Three structural differences follow from it:
- Physical re-testability. Engineering claims about materials, structures, and components can often be independently re-verified against the original artefact, lab notebook, or calibration log — a forensic route rarely available in human-subjects research.
- Professional licensure exposure. Many engineering academics also hold a Professional Engineer (PE) or Chartered Engineer licence. A misconduct finding can trigger parallel discipline from a state or national licensing board — a structural check with no direct academic equivalent for most non-clinician biomedical researchers.
- Public-safety framing. The National Society of Professional Engineers’ Code of Ethics states that engineers “shall hold paramount the safety, health, and welfare of the public” as its first fundamental canon. Biomedical research ethics is instead anchored in the Belmont Report’s principles of respect for persons, beneficence, and justice — a subject-protection frame rather than an infrastructure-safety frame.
Prevalence data reflects the same underlying pattern. A widely cited 2009 meta-analysis (Fanelli, published via PLOS ONE) found that close to 2% of scientists admitted to fabricating or falsifying data at least once, with up to a third admitting other questionable research practices — figures drawn predominantly from biomedical and life-science samples. A 2021 systematic review and meta-analysis published in Science and Engineering Ethics (Xie et al.) updated pooled prevalence estimates for both research misconduct and questionable research practices, underlining that engineering-specific base rates remain comparatively under-studied against biomedicine’s much larger evidence base.
Types of Misconduct Most Relevant to Engineering Research
The core FFP taxonomy applies across disciplines, but its practical expression in engineering research differs from its expression in biomedicine.
| Misconduct type | Typical biomedical expression | Typical engineering expression |
|---|---|---|
| Fabrication | Invented clinical outcomes, patient counts, or assay readings | Invented structural-test results, fatigue-cycle counts, or simulation outputs |
| Falsification | Selective omission of adverse trial data; altered statistical models | Altered materials-strength certificates; suppressed failed load tests |
| Image manipulation | Reused or altered western blots, microscopy, or gel images | Altered micrographs, stress-map renders, or non-destructive-test scans |
| Plagiarism | Copied text, methods, or literature review sections | Copied methodology or design specifications without attribution |
Image manipulation as research misconduct deserves particular attention: COPE guidance treats any enhancing, obscuring, moving, or adding of image features as misconduct, while proportionate brightness/contrast adjustments applied equally across an image (and its controls) remain acceptable. In engineering, the equivalent images are typically micrographs, non-destructive-testing scans, or stress-distribution renders — evidence that, unlike a clinical image, can sometimes be regenerated from the original physical specimen if it still exists.
Structural-Testing and Materials-Data Fraud: Four Real Cases
Research misconduct case studies in engineering rarely make front-page news the way clinical-trial scandals do, but several documented cases illustrate the pattern.
- Kobe Steel (2017). The Japanese manufacturer admitted to falsifying quality-inspection data on the strength and durability of aluminium, copper, and steel products, which had been supplied into automotive, rail, and aerospace supply chains — a case that, while industrial rather than academic, shows how falsified materials data propagates once it enters downstream engineering use.
- Ranga Dias superconductivity claims. A University of Rochester investigation concluded that physicist Ranga Dias had committed research misconduct, including data fabrication and falsification, in connection with room-temperature superconductivity claims published in Nature. Multiple papers were retracted and Dias was dismissed — a rare case where fabricated materials-property data was caught partly through failed independent replication attempts by other labs.
- Falsified precast-concrete inspection records. A case reported through American Society of Civil Engineers (ASCE) channels involved a materials-testing firm supplying falsified paperwork claiming that required inspections of precast concrete units had been carried out when they had not — misconduct that, had it gone undetected, would have compromised a live construction project rather than merely a journal record.
- Forged structural engineering seals. Separately reported cases have involved individuals using stolen or copied software to forge a licensed engineer’s professional seal on structural plans, bypassing the licensure check that engineering — uniquely among the research disciplines discussed here — relies on as a second line of defence beyond peer review.
The common thread: in three of the four cases, the fraud was exposed through re-inspection of a physical artefact — steel stock, a concrete unit, a stamped drawing — rather than statistical anomaly detection alone.
Common Questions and What Comes Next
What are the types of research misconduct?
The three recognised types are fabrication (inventing data or results), falsification (manipulating data, equipment, or processes to misrepresent findings), and plagiarism (using others’ work without attribution). Related but distinct issues — undisclosed conflicts of interest, authorship disputes, and citation manipulation — are handled under separate publication-ethics procedures rather than the core misconduct definition.
What is the difference between fabrication and falsification?
Fabrication means inventing data or results that were never actually produced by an experiment or test. Falsification means manipulating real research materials, equipment, or processes, or altering/omitting genuine data, so that the record no longer accurately reflects what happened. Both are treated as equally serious under the ORI’s FFP standard.
Is image manipulation considered research misconduct?
Yes — inappropriate image manipulation is treated as a form of falsification under COPE guidance. Enhancing, removing, moving, or adding specific image features is prohibited; uniform brightness or contrast adjustments applied equally to an image and its controls are acceptable, provided nothing is obscured or misrepresented.
What is an example of research misconduct in engineering?
Documented examples include the Kobe Steel falsified materials-certification scandal, the Ranga Dias room-temperature superconductivity fabrication finding at the University of Rochester, and falsified precast-concrete inspection paperwork reported through ASCE channels. Each involved misrepresenting physical-material or structural-test data rather than clinical or behavioural data.
For research administrators, the implication is practical: engineering integrity offices should maintain re-testing and sample-retention protocols alongside the statistical and plagiarism-detection tools built for biomedical and social-science misconduct. A materials sample or calibration log retained past publication is often the single most decisive piece of evidence in resolving an engineering allegation — a resource with no direct biomedical equivalent once a clinical study has closed. As more engineering journals adopt COPE’s flowcharts and licensing boards sharpen data-retention expectations, expect engineering cases to be resolved increasingly through re-testable physical evidence rather than statistical inference alone.