CoChat
Sign Up
Cited by 4Structural Health Monitoring TechniquesEngineering and Test SystemsAeroelasticity and Vibration Control

Design of a Modularized IoT Multi-Functional Sensing System and Data Pipeline for Digital Twin-Oriented Real-Time Aircraft Structural Health Monitoring

Open access
Guo, Shengkai, West, Andrew, Papuga, Jan, Theodossiades, Stephanos, Jiang, Jingjing·2025·Sensors
DOI 10.3390/s25216531PMC PMC12608597
Open this paper in CoChat — free account

Abstract

A modular, multi-functional (encompassing data acquisition, management, preprocessing, and transmission) sensing (MMFS) system based upon the Internet of Things (IoT) paradigm is discussed in this paper with the goal of continuous real-time, multi-sensor and multi-location monitoring of aircraft (including drones) structural performances during flight. According to industrial and system requirements, a microcontroller and four sensors (strain, acceleration, vibration, and temperature) were selected and integrated into the system. To enable the determination of potential in-flight failures and estimates of the remaining useful service life of the aircraft, resistance strain gauge networks, piezoelectric sensors for capturing structural vibrations and impact, accelerometers, and thermistors have been integrated into the MMFS system. Real flight tests with Evektor’s Cobra VUT100i and SportStar RTC aircraft have been undertaken to demonstrate the features of recorded data and provide requirements for the MMFS functional design. Real flight test data were analysed, indicating that a sampling rate of 1000 Hz is necessary to balance representation of relevant features within the data and potential loss of quality in fatigue life estimation. The design and evaluation of the performance of a prototype (evaluated via representative stress/strain experiments using an Instron Hydraulic 250 kN machine within laboratories) are detailed in this paper.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper presents a modular IoT multi-functional sensing (MMFS) system and data pipeline for real-time aircraft structural health monitoring oriented toward digital twins. It integrates strain, acceleration, vibration/PZT, and temperature sensing, uses real flight data to recommend a sampling rate of at least 1000 Hz, and validates a prototype in laboratory static-load tests.

Key findings

  • The MMFS integrates an STM32L475VG MCU with four sensor types: BF350-3AA strain gauge, PZT sensor, EVAL-ADXL356CZ 3D accelerometer, and LM35 temperature sensor, supporting modular deployment across aircraft locations and cloud/digital-twin data transfer.
  • Real flight tests on Evektor Cobra VUT100i and SportStar RTC aircraft showed engine excitation near 200 Hz; sampling below 1000 Hz lost high-frequency acceleration features and reduced estimated fatigue life, leading to a recommended sampling frequency of no less than 1000 Hz. The prototype used 1094 Hz ADC sampling, 12-bit resolution, and end-to-end latency below 50 ms.
  • In laboratory tests on a 300 mm × 100 mm × 2 mm aluminium plate under static Instron loads from 0 to 15 kN, strain increased from 0 to 360 με between 10 s and 70 s, held at 360 με for 10 s, then reduced to 0 during 80–140 s. Hampel filtering reduced 3D acceleration noise to 0.1 g, with Y-axis around 1 g and X/Z around 0 g; temperature reflected ambient room temperature at 22 °C.
  • A data-volume example for two nodes with seven channels each, 1094 Hz sampling, 12-bit ADC, and a 1.5 h flight gives approximately 165 MB raw data, within 8 GB or 16 GB SD storage capacity.

Methods

  • The study designed and prototyped a modular MMFS architecture with analogue signal conditioning, ADC/DMA double-buffered acquisition, on-board preprocessing, and UART/BLE/Wi-Fi communication to an on-board gateway and cloud-based digital twin platform.
  • Real flight tests used an Evektor Cobra VUT100i with 18 accelerometer channels and a SportStar RTC; each aircraft flew three flights for approximately 3 h total across take-off, taxi, climb, cruise, parabolic flight, turns, and landing. Additional tail acceleration data were recorded at 10 kHz and decimated to 100, 200, 500, 1000, 2000, and 5000 Hz to compare maximum stress amplitude and predicted fatigue life.
  • Laboratory validation used an aluminium plate specimen with the prototype modules, loaded by an Instron Hydraulic 250 kN machine under static loads from 0 to 15 kN, with UART terminal logging and Hampel filtering using a three-sample sliding window and a threshold of three times the median absolute deviation.

Limitations

  • The paper states that the PZT output was scaled using a high-value resistor divider, which loads the high-impedance piezoelectric source and limits low-frequency response; a dedicated charge amplifier is being designed.
  • The paper states that temperature compensation for strain drift is proposed but requires further dedicated testing, and that per-sensor sampling frequencies could be optimised in future software work.
  • The paper states that higher-level digital twin analytics and remaining-useful-life algorithms are not yet integrated; the presented device is the physical data acquisition layer.
  • Evident from the text: prototype validation used static laboratory loads only, so dynamic PZT response was not exercised, and the full MMFS was not validated in real flight—the flight tests used accelerometers primarily to determine sampling requirements.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/s25216531

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

4 of 11 cited sources checked · 2 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Monitoring various locations of an aircraft can provide valuable information for assessing the aircraft's health condition.

    “Evidence suggests that monitoring various locations of an aircraft can provide valuable information for assessing the aircraft’s health condition [13,14].”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [13] Reymer, 2011, Flight Loads Acquisition for PZL-130 Orlik TCII Full Scale Fatigue Test, Fatigue Aircr. Struct. source
    • cites [14] Kim, 2015, Aircraft health and usage monitoring system for in-flight strain measurement of a wing structure, Smart Mater. Struct. source
  • Premisecitation

    Key aspects such as sub-system architecture, core modules, deployment models, and implementation methods remain insufficiently detailed for aerospace contexts.

    “As a result, key aspects such as sub-system architecture, core modules, deployment models, and implementation methods remain insufficiently detailed for aerospace contexts [40].”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [40] Vujic, 2015, Wireless sensor networks applications in aircraft structural health monitoring, Istraz. Proj. Za Privredu source

      The Sensor node architecture section describes node components and deployment models, and the Conclusion presents a general scheme for aircraft monitoring. Neither section says that sub-system architecture, core modules, deployment models, or implementation methods remain insufficiently detailed for aerospace contexts. Searches for that assessment found no passage stating it. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
  • Findingown data

    A sampling rate of 1000 Hz is necessary to balance representation of relevant features within the data and potential loss of quality in fatigue life estimation.

    “Real flight test data were analysed, indicating that a sampling rate of 1000 Hz is necessary to balance representation of relevant features within the data and potential loss of quality in fatigue life estimation.”
    Quote in the paperSubstantial claim
  • Findingreasoning

    For two nodes with seven channels each, sampling at 1094 Hz with a 12-bit ADC resolution during a 1.5 h flight, the total raw data volume is approximately 165 MB.

    “For example, considering two nodes with seven channels each, sampling at 1094 Hz (typical value used in this study) with a 12-bit ADC resolution during a 1.5 h flight (typical value used in this study), the total raw data volume is approximately (two Nodes * seven channels * 1094 Hz * 2 bytes * 1.5 H * 3600 s/H ≈ 165 MB).”
    Quote in the paperNot a substantial claim
  • Findingnone given

    Sampling frequencies below 1000 Hz result in the loss of high-frequency acceleration features.

    “Sampling frequencies below 1000 Hz result in the loss of high-frequency acceleration features.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 22
  • May combine two claims. 16
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Major aircraft manufacturers like Boeing and Airbus have proposed full-scale fatigue tests and diverse SHM technology in aircraft health monitoring, emphasising requirements such as large-scale deployments, lightweight sensors and processing equipment, low power consumption and continuous online monitoring.

    “Major aircraft manufacturers like Boeing and Airbus have proposed full-scale fatigue tests and diverse SHM technology [15,16] in aircraft health monitoring, emphasising requirements such as large-scale deployments, lightweight sensors and processing equipment, low power consumption and continuous online monitoring.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [15] Swindell, P., Doyle, J., and Roach, D. (2017, January 16). Integration of structural health monitoring solutions onto commercial aircraft via the Federal Aviation Administration structural health monitoring research program. Proceedings of the AIP Conference Proceedings, Atlanta, GA, USA. source

      Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [16] Pfeiffer, H., and Wevers, M. (2007, January 26). Aircraft Integrated Structural Health Assessment—Structural Health Monitoring and its implementation within the European project AISHA. Proceedings of the EU Project Meeting on Aircraft Integrated Structural Health Assessment (AISHA), Leuven, Belgium.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Findingnone given

    The hardware for the MMFS was designed, including an MCU, strain gauge modules, 3D accelerometers, PZT sensors, and thermistors.

    “The hardware for the MMFS was designed, including an MCU, strain gauge modules, 3D accelerometers, PZT sensors, and thermistors.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    The software has been designed to enable each sensor node to acquire data from the four types of sensors continuously and in real-time.

    “The software has been designed to enable each sensor node to acquire data from the four types of sensors continuously and in real-time.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    Each multi-functional sensor node in the system is connected to a processor and integrates four types of sensors (strain gauges, PZT sensor, accelerometer and thermistor).

    “each multi-functional sensor node in the system is connected to a processor and integrates four types of sensors (strain gauges, PZT sensor, accelerometer and thermistor).”
    Quote in the paperSubstantial claim

  • The quote only partly states this claim.

    Findingreasoning

    The signal loss from sampling below 1000 Hz converts to a decrease in the related fatigue life from the computational analysis.

    “The signal loss converts to a decrease in the related fatigue life from the computational analysis.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    After Hampel filtering, the amplitude of the 3D acceleration noise reduced to 0.1 g, with the Y-axis stabilising around 1 g and the X and Z axes stabilising around 0 g.

    “After Hampel filtering, the amplitude of the 3D acceleration noise reduced to 0.1 g, with the Y-axis stabilising around 1 g and the X and Z axes stabilising around 0 g.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The strain gauge amplification circuit has a low-pass filter in the analogue front end with a cutoff frequency of approximately 81 Hz.

    “For the employed strain gauge amplification circuit, a low-pass filter was implemented in the analogue front end, with a cutoff frequency of fc = 1/(2πR2C4) ≈ 81 Hz.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The integrated temperature sensor measurements primarily reflected the ambient room temperature (i.e., 22°).

    “The integrated temperature sensor measurements primarily reflected the ambient room temperature (i.e., 22°).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingnone given

    We successfully pre-processed the measured raw signals, reducing noise in the data.

    “In addition, we successfully pre-processed the measured raw signals, reducing noise in the data.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The strain gauge began detecting the static load applied by the Instron machine at around 10 s.

    “The strain gauge began detecting the static load applied by the Instron machine at around 10 s (see Figure 12b).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The static load was gradually reduced to 0 during 80 s to 140 s.

    “The static load was gradually reduced to 0 during 80 s to 140 s.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Between 10 s and 70 s, the strain increased from 0 to 360 με, followed by a static strain of 360 με for 10 s.

    “Between 10 s and 70 s, the strain increased from 0 to 360 με, followed by a static strain of 360 με for 10 s.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingreasoning

    The Y-axis of the accelerometer is parallel to the direction of gravity, resulting in a Y-axis measurement of 1 g.

    “Note: The Y-axis of the accelerometer is parallel to the direction of gravity, resulting in a Y-axis measurement of 1 g.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Due to the absence of dynamic loads, the output from the PZT remained largely unchanged during 0-160 s.

    “Due to the absence of dynamic loads, the output from the PZT remained largely unchanged during 0–160 s.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Engine failures, external impacts from objects, cracking at different locations of aircraft, deformation, and ageing are common issues that can affect aircraft structures.

    “Engine failures, external impacts from objects (such as birds and meteorites), cracking at different locations of aircraft, deformation, and ageing are common issues that can affect aircraft structures [17].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [17] Tavares, 2017, An overview of fatigue in aircraft structures, Fatigue Fract. Eng. Mater. Struct. source

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The sensing system can be utilised to monitor strain/stress values, dynamic characteristics, acceleration, acoustic emission and temperature simultaneously and continuously throughout complete flights.

    “A sensing system that can be utilised to monitor strain/stress values, dynamic characteristics, acceleration, acoustic emission and temperature simultaneously and continuously throughout complete flights.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Key features can be extracted and transmitted to the ground station, reducing the wireless backhaul demand.

    “Meanwhile, key features can be extracted and then transmitted to the ground station, reducing the wireless backhaul demand.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    In a full digital twin framework, the presented device represents the physical data acquisition layer.

    “In a full digital twin framework, the presented device represents the physical data acquisition layer.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Study design and conductnone given

    Double-buffered DMA minimises data latency and loss.

    “Double-buffered DMA minimises data latency and loss, while BLE and SD card layers use checksum mechanisms to ensure data integrity.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductreasoning

    The double-buffering mechanism ensures continuous data acquisition and maintains the real-time integrity of the sampled signals.

    “This approach ensures continuous data acquisition and maintains the real-time integrity of the sampled signals.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The current STM32L4 series MCU can only measure voltage input signals in the range of 0-3.3 V, so any signals outside this range should be scaled through conditioning circuits to ensure compatibility with the MCU's input limits.

    “As the current STM32L4 series Microcontroller Unit (MCU) can only measure voltage input signals in the range of 0–3.3 V, any signals outside of this range should be scaled through conditioning circuits to ensure compatibility with the MCU’s input limits [44].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [44] STMicroelectronics (2025, July 29). AN2834: How to Get the Best ADC Accuracy in STM32 Microcontrollers. Available online: https://www.st.com/resource/en/application_note/an2834-how-to-get-the-best-adc-accuracy-in-stm32-microcontrollers-stmicroelectronics.pdf.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The system is designed to interface seamlessly with higher-level analytics and modelling modules that constitute the complete digital twin.

    “Although this paper focuses on the hardware layer, the system is designed to interface seamlessly with higher-level analytics and modelling modules that constitute the complete digital twin.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationown data

    A sampling frequency of no less than 1000 Hz is recommended to ensure reliable aircraft damage assessment.

    “The determination of the sampling frequency for the structural health monitoring system of small aircraft digital twins has been demonstrated; that is, a sampling frequency of no less than 1000 Hz is recommended to ensure reliable aircraft damage assessment.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    This paper provides a feasible scheme for converting ADC readings into strain data through the modulation/amplification circuit of the strain sensor.

    “This paper provides a feasible scheme for converting ADC readings into strain data through the modulation/amplification circuit of the strain sensor.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationnone given

    Real-time acquisition of sensor data is crucial during aircraft flight, necessitating the operation of any DT functionality in real-time.

    “Unlike SHM approaches in other scenarios, real-time acquisition of sensor data is crucial during aircraft flight, necessitating the operation of any DT functionality in real-time.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The analysis of the results indicates that the proposed hardware and data acquisition system can be utilised as the MMFS, demonstrating its potential for effective real-time and continuous data collection in aircraft structural health monitoring.

    “The analysis of the results indicates that the proposed hardware and data acquisition system can be utilised as the MMFS, demonstrating its potential for effective real-time and continuous data collection in aircraft structural health monitoring.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The modularised, non-customised design approach can be adapted and applied to various forms of machines (including manned aircraft and drones).

    “A modularised, non-customised design approach for multifunctional sensing systems, which can be adapted and applied to various forms of machines (including manned aircraft and drones).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    This modular design facilitates expansion to include more areas of the aircraft fuselage within the SHM system.

    “This modular design facilitates expansion to include more areas of the aircraft fuselage within the SHM system.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The proposed hardware layer enables synchronised multi-sensor acquisition (strain, vibration, acceleration and temperature) with on-board pre-processing and real-time transmission, which are essential for model-driven digital twin updates.

    “It enables synchronised multi-sensor acquisition (strain, vibration, acceleration and temperature) with on-board pre-processing and real-time transmission, which are essential for model-driven digital twin updates.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The selected integrated temperature sensor LM35 has a temperature range of 0 °C to 150 °C and a linear scale factor of +10 mV/°C.

    “The selected integrated temperature sensor, LM35, is an analogue sensor with a temperature range of 0 °C to 150 °C and a linear scale factor of +10 mV/°C [51].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] Texas Instruments (2025, July 29). LM35 Precision Centigrade Temperature Sensors. Available online: https://www.ti.com/lit/ds/symlink/lm35.pdf.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    A 12-bit ADC is chosen to strike a balance between sampling precision (0.8 mV) and conversion time (12.5 ADC clock cycles per conversion).

    “In this study, a 12-bit ADC is chosen to strike a balance between sampling precision (i.e., 0.8 mV) and conversion time (i.e., 12.5 ADC clock cycles per conversion).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The high-value resistor divider in the current prototype inherently loads the high-impedance piezoelectric source and limits the low-frequency response.

    “However, it inherently loads the high-impedance piezoelectric source and limits the low-frequency response.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    BLE and SD card layers use checksum mechanisms to ensure data integrity.

    “Double-buffered DMA minimises data latency and loss, while BLE and SD card layers use checksum mechanisms to ensure data integrity.”
    Quote in the paperNot a substantial claim
33 of 33

  • The paper relies on this without support.

    Reported designs often lack validation using real flight test data; hence, multiple factors such as the sampling rate of sensors, the sensing measurement range at different flight stages and external impacts and damage are not addressed during the sensing node design stage.

    “Most importantly, reported designs often lack validation using real flight test data; hence, multiple factors such as the sampling rate of sensors, the sensing measurement range at different flight stages and external impacts and damage are not addressed during the sensing node design stage.”

  • The authors state this as an assumption.

    The temperature-dependent offset Voffset(T) is a linear function of temperature.

    “Specifically, the temperature-dependent offset Voffset(T) was modelled as a linear function:”

  • The paper relies on this without support.

    There is a linear relationship between strain and resistance change in the elastic working range of commercial strain gauges.

    “Since there is a linear relationship between strain and resistance change in the elastic working range of commercial strain gauges, the real-time strain value can be obtained based on the strain change and gauge factor parameters.”
3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. Pfeiffer, H., and Wevers, M. (2007, January 26). Aircraft Integrated Structural Health Assessment—Structural Health Monitoring and its implementation within the European project AISHA. Proceedings of the EU Project Meeting on Aircraft Integrated Structural Health Assessment (AISHA), Leuven, Belgium.
  2. Wang, 2013, Research and Calibration Experiment of Characteristic Parameters of High Temperature Resistance Strain Gauges, Sens. Transducers, 159, 324

Cited by

Sorted by citations

Authors
Guo, S., West, A., Papuga, J., Theodossiades, S., & Jiang, J.
Affiliations
  • Loughborough University (GB)
  • Evektor (Czechia) (CZ)
Corresponding author
Jiang, Jingjing
ORCID
Venue
Sensors
Publisher
MDPI AG
Published
2025-10-23
Volume
25
Issue
21
Pages
6531
ISSN
1424-8220 (ISSN-L)
DOI
10.3390/s25216531
PMC
PMC12608597
OpenAlex
W4415489824
Semantic Scholar
e65091c44b2a8d2fa84451536c020162a8d6985b
Cited by
4
References
54
Citations by year
2026: 4
FWCI
1.52
Type
journal article
Language
en
Open access
Licence
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • European Climate, Infrastructure and Environment Executive Agency (101096073)
  • UK Research and Innovation (10065739)
Topics
  • Physical Sciences › Engineering › Civil and Structural Engineering › Structural Health Monitoring Techniques
  • Physical Sciences › Engineering › Control and Systems Engineering › Engineering and Test Systems
  • Physical Sciences › Engineering › Aerospace Engineering › Aeroelasticity and Vibration Control
Keywords
aircraft structural health monitoring, Internet of Things, digital twin, real-time monitoring, multi-sensor data acquisition, piezoelectric vibration sensor, accelerometer, thermistor, remaining useful life prediction, fatigue life prediction, sampling frequency, flight testing, unmanned aerial vehicles

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review