Research libraryEdge to Dashboard
Chapter 06 of 14 · Part II · Draft web edition
Local inspection earns its place when a valid image becomes a traceable quality decision before the fabric leaves the decision point. Acquisition, persistence and an approved fallback matter as much as inference speed.
The cloth is woven from this chapter's 5 exhibits: each weft is one exhibit's own data, normalised; each warp group is a section. The probe reads the exhibit's own value where it crosses. Procedural visualisation, not a measurement.
Read-out
GFT proposed cell informed by the sources. Layers are responsibilities, not certified network zones. Protective machine functions are independent and outside this analytics path.
Unit: illustrative responsibility layers. Status: illustrative. Sources: S14, S16, S26, S17. Illustrative figures are teaching examples, not measured factory results.
Overture · 1 min
Keep the inspection decision beside the fabric when waiting for a remote service would make that decision late. GFT proposes a local chain from controlled acquisition to inference, durable evidence and an authorised quality action. The remote service receives selected results and supports later analysis. Locality alone does not establish accuracy, continuity or safety. Each depends on a tested operating boundary. This chapter specifies that chain as a reference design, not an account of an installed GFT system. W3C describes selective industrial edge collection, while ETSI provides an architecture for applications near the network edge.[S14][S16]
6.1
Start by defining what the station must distinguish. GFT recommends a defect catalogue with acceptance criteria, a physical location convention and examples agreed by the quality owner. Describe the product families and surfaces within scope. Separate a visible indication from a confirmed defect, and separate either from the decision to reject material. The acquisition literature identifies lighting variation and background noise as threats to detection. A promising model result therefore cannot stand in for a camera-cell acceptance record.[S26]
GFT's proposed commissioning record should bind the camera position, lens, focus, illumination arrangement, exposure, trigger source, fabric presentation and calibration reference to an approved configuration. Specify how the fabric is supported and where its position is measured. Make the reference image reproducible after cleaning or maintenance. Record changes to the light source and enclosure as configuration changes, because the station must be able to explain a changed image before it attributes a changed result to the fabric. These are design recommendations derived from the acquisition limitations, not performance guarantees.[S26]
Commissioning should prove the station's physical coverage before assessing its model. GFT recommends inspecting the edges of the intended field, transitions between camera views, folds and areas obscured by guides. Test the permissible fabric movement and presentation conditions against the defect definition. A station can report a technically valid inference while failing to see the material that matters. Record unseen or invalid areas explicitly as uninspected coverage. Do not let a camera heartbeat stand for complete fabric inspection.
Site validation should then use independently adjudicated material from the intended operating conditions. GFT recommends grouping evaluation evidence by fabric construction, colour, finish and relevant operating condition. Keep samples from the same physical defect together when separating development and evaluation material. Otherwise adjacent views may make the evaluation look more independent than it is. Document how a disputed reference label is resolved. The research scope includes plain, patternless cotton and linen; it does not establish transferable performance across every textile construction or camera arrangement.[S26]
GFT recommends an acquisition-validity gate before inference eligibility. The gate should reject an obscured field, lost trigger association, unacceptable exposure or an unapproved setup. It should return an explicit invalid state with a reason, rather than a confident defect classification on an unusable image. This recommendation makes the distinction operational: a classification concerns visible material; invalid acquisition concerns whether the station could observe that material adequately. Both conditions require a responsible response, but they support different conclusions.
The illustrative architecture in Exhibit 23 separates the observation, computation and decision responsibilities. Its layers describe an allocation of duties, not a purchased equipment list. The protective machine system remains outside the analytics chain. NIST's guidance considers OT performance, reliability and safety together; GFT uses that discipline to require a reviewed boundary between quality analytics and machine protection.[S17]
GFT recommends accepting physical coverage, acquisition validity and classification performance as separate findings. A failure in any finding should limit the permitted operating scope. Replacing a model cannot cure an unlit edge of the fabric, and changing a lamp should not silently preserve an old acceptance statement. The cell owner should retain the evidence and state which conditions require revalidation. This makes commissioning a repeatable measurement activity rather than a demonstration arranged around favourable samples.
For maintenance, GFT proposes a station handback sheet that begins with the physical intervention. State whether the camera mount moved, the lens was replaced, the lighting was changed or only the enclosure was cleaned. Compare the restored reference view with the approved setup and record the decision to resume. A maintenance ticket marked complete should not automatically clear an inspection restriction. Give the person responsible for quality a clear route to accept the restored scope or request further checks.
6.2
Image decision
Image decision: Remote data movement 5, Consequence of delay 5 · illustrative
Place computation according to the consequence of delay.
Operator alert
Operator alert: Remote data movement 1, Consequence of delay 4 · illustrative
Energy trend
Energy trend: Remote data movement 2, Consequence of delay 2 · illustrative
Model review
Model review: Remote data movement 4, Consequence of delay 1 · illustrative
The method behind it: Worked example: the physical deadline includes the action.
GFT assumed scores: higher x means greater data movement if processed remotely; higher y means greater consequence of waiting. Coordinates describe design judgement, not measurements or an economic ranking.
Unit: illustrative ordinal scores, 1 to 5. Status: illustrative. Sources: S16, S14, S0602. Illustrative figures are teaching examples, not measured factory results.
6.2 · the bodyExhibit 24 above; its figures are illustrative where marked.
Place computation according to the consequence of delay. ETSI describes software applications running on virtualisation infrastructure near the network edge. W3C's industrial example separates controller operation from selected data collection for monitoring or analysis. Neither source proves a latency for a particular inspection station. GFT's placement rule is therefore conditional: keep the necessary decision path local when the measured remote path cannot meet the decision deadline and interruption tolerance.[S16][S14]
Define that deadline at the physical action point. GFT recommends measuring the available travel distance from observation to diversion, marking, hold or operator intervention. Establish the speed envelope and the time needed by the receiving mechanism. Subtract acquisition, transfer within the cell, preprocessing, inference, evidence recording, decision handling and action time. Reserve explicit uncertainty for speed variation, scheduling and triggering. The resulting budget is a commissioning hypothesis until tested on the complete installed path.
GFT recommends measuring the whole decision path under competing workloads. Include image persistence, record synchronisation, log rotation, temperature-related throttling and restart activity in the test conditions. Report delayed and lost decisions alongside typical timing. A benchmark that excludes the result store or action acknowledgement does not test the same boundary as production. Where the required deadline remains unproven, restrict the station to advisory or retrospective use and retain the approved quality procedure.
The illustrative placement matrix in Exhibit 24 expresses consequence and evidence movement using ordinal design scores. The coordinates are GFT assumptions, not measured network or computing performance. The matrix is informed by edge placement and selective collection, and by the energy paper's abstract describing local storage, processing and selective transfer. It supports discussion of placement; it cannot select hardware or certify a deadline.[S16][S14][S0602]
GFT proposes a split between inference needed during production and analysis that can wait for synchronisation. Local software should hold the approved model, required configuration and operating context before the shift begins. Remote services may compare outcomes, curate review material and prepare a candidate update. Approval and deployment of that update remain a separate change. The placement argument does not require a complete MEC platform at every camera; the relevant requirement is a bounded local service with known dependencies.[S16]
For resource analytics, GFT recommends the same discipline without copying an image-inspection deadline. A local energy calculation can inform an operator while later reports reconcile production context. The energy study's publisher-deposited abstract describes a database at the edge followed by analytical processing. This supports the pattern, not a claimed saving or a guarantee against power failure. GFT would require the factory to identify the useful response interval and measure continuity at its own metering boundary.[S0602]
GFT also recommends an explicit rule for late results. Once the inspected fabric has passed the permitted action point, an inference may still support investigation or a material hold downstream. It should no longer issue the original time-sensitive instruction as though its deadline remained open. Record the missed deadline and the alternative action. This rule prevents a delayed but otherwise correct classification from being counted as a successful timely decision, and it makes the consequence of queue growth visible to the quality owner.
6.3
Related · chapter 099 paragraphs step it
Read-out
Illustrative GFT historian store architecture informed by ISO 22400 KPI structure (S08), EPCIS visibility event sharing (S15), ISO/IEC 27001 change control (S24) and edge database persistence (S29).
Unit: . Status: illustrative. Sources: S08, S15, S24, S29. Illustrative figures are teaching examples, not measured factory results.
Read this exhibit in its chapter ↗Separate decision evidence from image distribution. MQTT is a publish/subscribe messaging transport whose specification discusses constrained environments. W3C describes selected industrial data collection and edge preprocessing. GFT therefore recommends publishing compact quality events first, with images subject to a separate retention and transfer policy. A transport choice cannot determine whether a defect label is correct, whether its material identity is intact or whether a remote reviewer has enough evidence.[S0601][S14]
For each publication, GFT recommends retaining event identity, observed material identity, source time, position, defect taxonomy, model and configuration references, acquisition state and current decision state. Include a confidence score only with its declared meaning. Distinguish an image location from evidence that the image remains retrievable. Record what a receiver can do if the image is absent. The result should explain whether material was inspected and whether its release was authorised, even when the image transfer is delayed.
GFT recommends separate states for acquisition, inference execution, defect indication and human disposition. A valid image with no detected defect is different from an invalid image with no inference. A supervisor's acceptance of material is different from the model's indication. Keeping these states separate prevents a subscriber from treating an empty defect field as evidence of satisfactory quality. Require receiving applications to reject unknown state combinations rather than repair them through a hidden assumption.
Before sizing transfer, declare the event cadence. GFT recommends distinguishing frame-level observations, merged defect events and final dispositions. The event reduction rule must preserve the relationship between successive images of the same physical defect. If the rule merges indications, retain their references and spatial boundary. If it discards repeated observations, record that policy. Otherwise a link-efficient system may change the apparent defect count whenever the camera rate or merge window changes.
The illustrative calculation supports a transfer-policy test, not a universal reduction claim. GFT recommends reserving transfer capacity for current results and agreed review evidence before bulk historical images. Identify image classes that require retention, such as disputed decisions, invalid acquisitions and a representative audit sample. Define the sampling policy before deployment and preserve its version. Avoid selecting only convincing detections, because a reviewer also needs evidence about material the system judged acceptable.
GFT recommends minimising the scene to the fabric and necessary reference features. Where people or customer-sensitive patterns may appear, assign an owner for access and retention decisions before image collection begins. ISO's public description of its information-security management standard provides a governance framework; it does not supply this station's image permissions or retention duration. The factory must document those decisions and the means of enforcing them. This chapter does not offer a legal conclusion about a particular capture arrangement.[S24]
Test the published schema with a receiving application before accepting the link. GFT recommends injecting unknown taxonomy values, unavailable image references and delayed dispositions. Check that the receiver preserves the raw event and displays its quality state. A message arriving successfully proves transport arrival within the tested boundary. It does not prove physical inspection coverage, material release or a completed corrective action. The application acceptance record should make those distinctions visible.[S0601]
GFT proposes that the transfer contract distinguish unavailable evidence from restricted evidence. A missing file, an expired retention period and a denied access request have different remedies. The remote reviewer should see the applicable reason without receiving unnecessary sensitive content. The local operator should be able to locate the original material decision even if the reviewer has no image permission. This makes selective publication an evidence policy with accountable exceptions rather than an unexplained reduction in data.
Interlude · the cloth half-woven
Woven: sections 6.0 to 6.3Still to weave: 6.4 to 6.6
Read-out
GFT arithmetic from World Bank Enterprise Surveys Cameroon 2024 Country Profile (S32). Outage-incidence gap against Sub-Saharan Africa: 93.3 percent minus 69.7 percent equals 23.6 percentage points. Outage-incidence gap against lower-middle-income economies: 93.3 percent minus 55.0 percent equals 38.3 percentage points. Monthly frequency excess: 10.4 minus 6.9 equals 3.5 outages per typical month. Compares all surveyed firms, not textile factories.
Unit: percentage points and monthly events. Status: derived. Sources: S32.
Read this exhibit in its chapter ↗6.4
Observe and infer
Observe and infer. Starting point. Leads to Commit locally. · illustrative
Persist the evidence needed to explain the local decision before allowing remote delivery status to determine its fate.
Commit locally
Commit locally. From Observe and infer. Leads to Check host state. · illustrative
Check host state
Check host state. From Commit locally. Leads to Queue with limits (Offline), Publish when ready (Online). · illustrative
Queue with limits
Queue with limits. From Check host state (Offline). Leads to Publish when ready (Host returns). · illustrative
Publish when ready
Publish when ready. From Check host state (Online), Queue with limits (Host returns). Leads to Reconcile event IDs. · illustrative
Reconcile event IDs
Reconcile event IDs. From Publish when ready. End point. · illustrative
The method behind it: Worked example: capacity depends on the retention policy.
Proposed GFT sequence. Local continuation requires healthy acquisition, inference, storage and an approved procedure. Sparkplug does not guarantee storage capacity, persistence or uninterrupted inspection.
Unit: illustrative recovery sequence. Status: illustrative. Sources: S40, S0602, S22. Illustrative figures are teaching examples, not measured factory results.
6.4 · the bodyExhibit 25 above; its figures are illustrative where marked.
Persist the evidence needed to explain the local decision before allowing remote delivery status to determine its fate. The energy study's abstract describes local database storage followed by processing. Sparkplug describes a possible edge store-and-forward behaviour linked to primary-host state. These are useful patterns, but neither establishes this station's durable-write behaviour, outage capacity or replay correctness. GFT recommends testing those properties independently and keeping inspection eligibility separate from host availability.[S0602][S40]
GFT's proposed local journal has distinct observation, result, disposition and delivery records joined by stable identifiers. An observation can exist without a completed inference; a result can exist before a human disposition. Delivery acknowledgements should refer to the actual event version delivered. Preserve the order and meaning of revisions. If the service restarts between recording a result and publishing it, the unresolved delivery remains visible and eligible for reconciliation rather than being silently treated as completed.
The illustrative recovery flow in Exhibit 25 shows the intended boundary. Host loss changes remote synchronisation while a healthy, eligible local cell continues under its approved procedure. This is a GFT design condition, not automatic Sparkplug behaviour. A host may be essential in another application's architecture. Test the dependency explicitly by removing the actual remote services, including authentication, name resolution and image access, that the local process might otherwise use.[S40]
GFT recommends acknowledging remote delivery only at a defined receiving boundary. A broker acknowledgement and a committed business record are different events. Define which acknowledgement permits local deletion, how its identity survives restart and how lost acknowledgements are retried. The receiving store should recognise a repeat delivery without duplicating the production event. This is an application requirement beyond the fact that MQTT provides messaging transport. Test repeat delivery and conflicting revisions against the declared event policy.[S0601]
The illustrative storage comparison in Exhibit 26 uses the same inputs. Its common scale intentionally makes compact result storage appear small beside image storage. The exhibit does not compare reliability or evidence sufficiency. A policy that retains fewer images may save capacity while weakening the ability to resolve disputed decisions. GFT recommends documenting that trade-off in the quality acceptance record.
GFT recommends staged storage alarms tied to a calculated remaining capture horizon. Stop nonessential bulk transfer or discard only material already eligible under the approved retention policy. Never silently overwrite unresolved decisions. If the required evidence cannot be preserved, the station should declare itself ineligible for the affected quality function and invoke its fallback. Local persistence protects recoverability only while capacity, power and the recovery procedure satisfy the agreed boundary.
GFT recommends testing the recovery drain as well as the outage buffer. After a host returns, historical records and current production compete for the available link and receiving capacity. Give current decision visibility an explicit priority while preserving source time for historical records. Verify that the backlog shrinks under the tested production load. If it continues to grow, the system has reconnected without recovering. Record the remaining backlog and its oldest unresolved source time until reconciliation is complete.
6.5
Preserve enough context to reproduce the meaning of a decision. GFT recommends identifying the model artefact, preprocessing, threshold configuration, defect taxonomy and acquisition setup separately. Store their references with the result rather than relying on whichever versions happen to be installed during later review. The acquisition study motivates attention to input conditions; ISO's security-management overview supports governance at a broad level. The detailed inference-lineage requirement here is GFT's proposed engineering contract.[S26][S24]
Do not describe a confidence score as a measured probability of correctness unless that interpretation has been validated for the relevant operating scope. GFT recommends recording the score's definition, class association and threshold rule. An uncalibrated ranking may still help order human review, but its use must be declared. Preserve the distinction between a classification score and the station's eligibility state. A high score cannot turn an out-of-scope material or invalid acquisition into an accepted inspection.
The illustrative schema in Exhibit 27 groups related fields to keep the exhibit readable. Its groups are a GFT proposal, not fields mandated by the cited standards. Implementers should expand the contract into a typed schema with permitted values, null behaviour, units and compatibility rules. The visible table identifies the minimum questions that the receiving quality system must be able to answer.
Read-out
| Event and material | Stable identity and fabric reference |
|---|---|
| Time and position | Source time, clock state, location |
| Acquisition | Setup reference and validity reason |
| Inference | Model, preprocessing, threshold |
| Indication | Taxonomy, score and geometry |
| Evidence | Image reference and availability |
| Disposition | Decision, role and revision link |
| Delivery | Receiver state and replay identity |
GFT proposed schema, not a standard-mandated field list. Each group needs typed fields, explicit missing-value behaviour and a versioned contract. Acquisition and wireless sources motivate quality checks; they do not define model lineage.
Unit: illustrative field groups. Status: illustrative. Sources: S26, S31, S24. Illustrative figures are teaching examples, not measured factory results.
Wireless sensing research discusses synchronisation and radio sharing as data-integrity concerns. GFT recommends preserving those limitations where a wireless source contributes material position or operating context. A received value without a known association to the inspected material should remain uncertain. The model cannot repair a missing association merely by returning a class. Conversely, a wired camera does not make unrelated wireless context trustworthy. Validate the evidence chain at each relevant interface.[S31]
ISO's provenance standard for characteristic master data has a narrower scope than this proposed inference record. Its public abstract addresses provenance representation and exchange for characteristic master data, and explicitly excludes configuration management and change control from its scope. GFT may use governed asset and characteristic records as context, but should not describe that standard as a model-lineage specification. The inference record needs its own versioned contract and approved responsibility.[S41]
GFT recommends keeping later review as an appended decision, with a link to the original result and a reason for amendment. Review should state whether the disagreement concerns the reference label, acquisition, classification, location or release decision. That distinction determines the corrective action. A confirmed camera problem calls for cell intervention; a disputed taxonomy calls for quality governance; a missed process defect calls for production investigation. A generic override count would hide these different mechanisms.
For routine review, GFT proposes a paired evidence pack containing an observation reference, the original indication, the final disposition and any consequent action. Sample material judged acceptable as well as flagged material, using a declared selection rule. Keep unavailable images and unresolved cases in the denominator. Compare only periods whose operating scope and configuration are understood. This makes the record useful for production learning without implying that a dashboard trend alone demonstrates a change in model performance.
When production and quality teams disagree about an indication, GFT recommends preserving the disagreement as an unresolved case with an assigned owner. Do not overwrite the original label to make the records agree. State the additional inspection, sample examination or material identification needed to close it. Once resolved, retain the reason and the resulting material decision. This provides a usable review trail without pretending that the model output, the first operator response and the final disposition were always equivalent.
6.6
Related · chapter 139 paragraphs step it
Read-out
Illustrative recovery and reconciliation workflow informed by CISA isolated backup restoration guidance (S21), NIST SP 800-34r1 testing, training and exercise principles (S22), and NIST SP 800-61r3 incident response lifecycle alignment (S1301).
Unit: . Status: illustrative. Sources: S21, S22, S1301. Illustrative figures are teaching examples, not measured factory results.
Read this exhibit in its chapter ↗Define a quality fallback before giving the cell a production responsibility. NIST distinguishes recovery validation, personnel training and exercises that expose planning gaps. GFT recommends applying that discipline to inspection eligibility, not merely to server availability. Identify who can suspend automated reliance, who controls affected material and who authorises return to service. A functioning model process is insufficient if the station cannot preserve evidence or the operator cannot execute its decision.[S22]
GFT proposes explicit fallback entries for camera loss, an obscured lens, a lost trigger association, an unavailable model, unacceptable processing delay, full storage and missing remote services. For each, state the detection method, visible condition, immediate material treatment and responsible role. Remote-host loss can permit continued local inspection when all local conditions remain valid. Acquisition failure cannot receive that same permission. The procedure should make the difference apparent at the station.
Manual inspection should be a defined operating mode with a scope the quality owner has accepted. GFT recommends recording its capacity, inspection method, material identification and release authority. If the fallback cannot keep pace, the approved response may require slower production, segregated output or a quality hold. Do not label uninspected fabric acceptable simply to preserve an output counter. Keep fallback output identifiable so later reports distinguish normal digital coverage from material controlled by another procedure.
Protective action remains a separate engineering responsibility. NIST describes safety systems and the importance of considering their relationship with other OT functions. GFT's requirement is stricter for this proposed analytics path: inference availability, broker state and dashboard access must not disable guards, emergency stops or other protective functions. Any automated quality action needs its own reviewed machine interface and permitted behaviour. This chapter does not qualify an inference model as a protective safety function.[S17]
GFT recommends a staged restart that first restores the approved acquisition setup and operating context. Check the installed model and configuration references, available storage, clock state and a reference sample before enabling reliance on results. Reconcile the material that passed during the interruption separately from delayed event delivery. Replaying historical evidence should not replay physical line actions against current fabric. Record which results arrived late and which material remains on hold.
Acceptance should retain the injected fault, observed station state, affected material boundary, response taken and evidence of restoration. GFT recommends a witnessed exercise with the intended operators, including a shift handover while a quality hold remains open. A technical recovery can succeed while responsibility is lost between shifts. NIST's separation of testing, training and exercise provides the relevant discipline; the material-control checks are GFT's application to the inspection cell.[S22]
GFT recommends recording acceptance by permitted use. The quality owner may accept image capture for retrospective review while withholding permission for automated disposition. Another acceptance may permit local alerts but require manual release of affected material. State each permission, its operating scope and the evidence that supports it. Changes to the cell should be assessed against those permissions. This avoids turning a successful demonstration of one function into an implied approval for every function the software happens to expose.
The chapter's recommendation is deliberately bounded. GFT proposes local inference where its deadline and continuity requirements justify placement, selected publication where the evidence policy permits it, and manual control when the cell is ineligible. Hardware selection, model economics, network recovery and wider deployment governance need their own acceptance evidence. The cell earns operational trust when observation, interpretation and disposition can all be inspected, including the intervals in which the digital path could not perform its assigned task.
Check the record
Every knot is a source this chapter cites; select one to read its record, or filter the register by grade. Publication dates and inherited-source limits stay explicit. Read the paper's full evidence summary.
11 sources · 39 citation marks on this pageMost cited: S14, 5 times
Grade AGrade BGrade CKnot size: citations on this page
All 11 records shown
c06-S14 · Grade A
Web of Things Architecture 1.1
An industrial edge device collects selected data from various controllers and makes it available to a cloud backend service, e.g., for remote monitoring via a dashboard or analyzes it for preventive maintenance.
2023-12-05. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S16 · Grade A
ETSI GS MEC 003 V3.2.1: Multi-access Edge Computing, Framework and Reference Architecture
Multi-access Edge Computing enables the implementation of MEC applications as software-only entities that run on top of a Virtualisation infrastructure, which is located in or close to the network edge.
2024-04. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S17 · Grade A
NIST SP 800-82 Rev. 3: Guide to Operational Technology Security
This document provides guidance on how to secure operational technology (OT) while addressing their unique performance, reliability, and safety requirements.
2023-09. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S22 · Grade A
NIST SP 800-34 Rev. 1: Contingency Planning Guide for Federal Information Systems
Testing validates recovery capabilities, whereas training prepares recovery personnel for plan activation and exercising the plan identifies planning gaps; combined, the activities improve plan effectiveness and overall organization preparedness.
2010-11-11. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S24 · Grade A
ISO/IEC 27001:2022: Information security, cybersecurity and privacy protection, Information security management systems, Requirements
The ISO/IEC 27001 standard provides companies of any size and from all sectors of activity with guidance for establishing, implementing, maintaining and continually improving an information security management system.
2022-10. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S26 · Grade B
Textile Defect Detection Using Artificial Intelligence and Computer Vision: A Preliminary Deep Learning Approach
Data acquired in complex industrial settings are highly susceptible to external variables such as lighting variations and background noise, which can significantly degrade the detection accuracy of machine learning (ML) models [14].
2025-09-18. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S31 · Grade B
A Review of IoT Sensing Applications and Challenges Using RFID and Wireless Sensor Networks
Also, time synchronization between the nodes and spectrum sharing techniques are of the utmost importance to ensure the integrity of the data or avoid uncertainty in the sensed data, without consuming all the radio resources [37].
2020-04-28. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S40 · Grade A
Sparkplug Specification 3.0.0
The intent and purpose of the Sparkplug Specification is to define an MQTT Topic Namespace, payload, and session state management that can be applied generically to the overall IIoT market sector, but specifically meets the requirements of real-time SCADA/Control HMI solutions.
2022-11-16. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S41 · Grade A
ISO 8000-120:2016: Data quality, Part 120: Master data: Exchange of characteristic data: Provenance
ISO 8000-120:2016 specifies requirements for the representation and exchange of information about the provenance of master data that consists of characteristic data, and supplements the requirements of ISO 8000‑110.
2016-10. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S0601 · Grade A
MQTT Version 5.0, OASIS Standard
MQTT is a Client Server publish/subscribe messaging transport protocol.
2019-03-07. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c06-S0602 · Grade B
Industrial IoT-Based Energy Monitoring System: Using Data Processing at Edge
Once the edge receives the meter’s data, it stores the information in the database server, followed by the data processing method to find nine additional analytical parameters.
2024-09-28. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗S14 World Wide Web Consortium. 2023-12-05.
S16 European Telecommunications Standards Institute. 2024-04.
S26 MDPI, Electronics. 2025-09-18.
S17 National Institute of Standards and Technology. 2023-09.
S0602 MDPI, IoT (publisher-deposited abstract via Crossref). 2024-09-28.
S0601 OASIS Open. 2019-03-07.
S24 International Organization for Standardization and International Electrotechnical Commission. 2022-10.
S40 Eclipse Foundation. 2022-11-16.
S22 National Institute of Standards and Technology. 2010-11-11.
S31 MDPI, Sensors. 2020-04-28.
S41 International Organization for Standardization. 2016-10.
Keep the thread · chapter 06 of 14
Move the shuttle along the rack to preview any chapter: its sections, exhibits and reading time. Draft web edition. This research is not investment, legal or tax advice; illustrative models are not measured factory results.
The paper's rack · 14 boltsBolt length: reading time
01Part I: Decisions before devices
GFT starts with the production decisions and infrastructure constraints that a connected architecture must serve, because visualisation alone cannot create trustworthy operating evidence.
02Part I: Decisions before devices
Each connected-factory use case needs an event, an owner, a permitted delay, a fallback and a measurable consequence before any sensor or platform is selected.
03Part I: Decisions before devices
A useful reference architecture assigns each function to a responsibility boundary and information contract before it assigns that function to a product.
04Part II: Capture at the production edge
A value becomes production evidence only when its asset, occurrence time, uncertainty and transformation history remain inspectable.
05Part II: Capture at the production edge
The safest digital retrofit begins by observing equipment through its documented interfaces and proving that acquisition does not alter control timing, protective functions or maintainability.
06Part II: Capture at the production edge
Local inspection earns its place when a valid image becomes a traceable quality decision before the fabric leaves the decision point. Acquisition, persistence and an approved fallback matter as much as inference speed.
07Part II: Capture at the production edge
The connected factory must lose remote services before it loses local evidence, and lose local evidence before it ever loses safe machine operation.
08Part III: One governed data spine
Two systems can exchange packets and still misunderstand production, so the architecture must define information, state, events and asset meaning separately.
09Part III: One governed data spine
The factory's data store is an evidence system, not a growing pile of values, so raw observations, contextual events, KPI results and corrections remain distinguishable and replayable.
10Part III: One governed data spine
The dashboard is a role-specific control surface whose freshness, definitions and next action are visible, not a gallery of independently calculated charts.
11Part III: One governed data spine
Applications should arrive as bounded, observable and reversible changes to a working production architecture, with ownership and rollback proven before scale.
12Part IV: Prove value and scale safely
This chapter specifies evidence capture rather than accounting or compliance methods: metres, denominators, allocations and life-cycle boundaries must make every displayed result reconstructable.
13Part IV: Prove value and scale safely
A connected factory is not accepted when its screens work on a normal day, but when its controls, fallbacks and recovery perform under adversarial and degraded conditions.
14Part IV: Prove value and scale safely
GFT recommends a four-phase build that earns each expansion through evidence quality, operating value, safe failure and recovery rather than device count.