Division 03 is the core intelligence division of the AGL Innovation Hub. Its purpose is to investigate how Artificial Intelligence can be applied to AGL inspection, maintenance, diagnostics, engineering, testing, photometry, failure analysis, asset management and operational decision-making. The uploaded framework specifically identifies AI-based predictive maintenance, AGL failure prediction, AI fault diagnosis, image-based inspection, automated compliance checking and AI-assisted engineering/design as major areas. The overall objective: convert AGL data into engineering intelligence, predictions, diagnostics and actionable recommendations.
The question this division answers — How can AI turn AGL asset, inspection, maintenance, testing and photometry data into failure predictions, fault diagnoses, defect detections, compliance checks, design assistance and maintenance recommendations — while keeping every safety-critical decision under qualified human technical control?
Division 03 is a technology domain division. It sits downstream of Division 01's strategic direction, alongside the other technology domains, and feeds intelligence forward into automation, control, photometry and predictive-maintenance divisions.
Use historical and real-time AGL data to identify assets that are likely to experience degradation or failure before the failure occurs.
Predict the probability of AGL equipment or systems developing a failure.
Assist engineers in identifying the most probable cause of an AGL fault.
AI should function as engineering decision support, not as an uncontrolled replacement for qualified technical judgment.
Use computer vision and AI to analyze photographs or video captured during AGL inspections.
Develop AI models capable of identifying visible AGL fixture defects.
Investigate AI-assisted inspection of AGL primary and secondary cable infrastructure.
Visual AI cannot replace electrical testing where electrical integrity must be established. AI should complement, not replace, appropriate electrical test procedures.
Use AI and computer vision to assist inspection of airfield signs.
AI could compare inspection images against:
Investigate AI-assisted monitoring and inspection of PAPI systems.
AI image analysis should support engineering inspection; it should not independently certify PAPI compliance unless the complete system has been technically validated and approved for that purpose.
Develop AI-assisted systems that compare engineering information against applicable requirements.
AI should identify potential compliance issues. Final compliance determination should remain under the responsibility of the appropriately qualified technical reviewer.
Use AI to support AGL engineers during design development.
Use AI interfaces to assist engineers in performing or interpreting engineering calculations.
AI should explain, automate and assist calculations while the underlying calculation engine remains deterministic and technically validated.
This aligns with the source framework's broader engineering-innovation objective of automated engineering calculations and engineering decision-support systems.
Use AI to analyze large volumes of AGL testing data.
This should become one of the strongest specialist AI applications within the AGL Innovation Hub. The source explicitly identifies AI photometry analysis within Division 03.
This can later connect directly with Division 14 — Advanced Photometry & Optical Innovation.
Automatically classify AGL failures into standardized failure categories.
Historical failure records can be used to identify patterns and assign a probable category to new failure events.
This should feed directly into the AGL Failure Database.
Assist engineers in identifying probable root causes of recurring AGL failures.
AI identifies similar asset failures, similar environmental conditions, similar installation characteristics, similar operating conditions, and similar maintenance history.
Convert asset and failure intelligence into actionable maintenance recommendations.
The engineer should retain authority to accept, reject or modify the recommendation.
Develop a dynamic health score for AGL assets.
| Score | Classification |
|---|---|
| 90–100 | Excellent |
| 75–89 | Good |
| 60–74 | Watch |
| 40–59 | Degraded |
| Below 40 | Critical |
The actual scoring thresholds should be validated against the organization's asset data rather than assumed universally.
This is the highest-level application of Division 03. The objective is to bring multiple AI outputs together into an engineering decision-support platform.
Division 03 is recommended to be structured around five layers, with human technical validation as the final and non-negotiable layer for safety-critical AGL applications.
Every AI-for-AGL project should follow the same disciplined lifecycle, ending in continuous monitoring once implemented.
For safety-related AGL applications, the Innovation Hub should record and maintain the following for every deployed model.
To make Division 03 operational rather than theoretical, these registers should be maintained and kept current.
KPIs are grouped across the seven functional areas AI touches within AGL — prediction, diagnosis, inspection, maintenance, photometry, engineering and overall program health.
The strongest feature of Division 03 is that it should not operate independently — it converts data supplied by other divisions into intelligence consumed by other divisions.
The long-term objective is an AI-enabled AGL Engineering Intelligence Platform: AGL field data (sensors, fixtures, CCRs, transformers, cables, signs, PAPI, photometry) feeds into asset, inspection, test, failure, maintenance and photometry data, which AI detects, classifies, diagnoses, predicts, analyzes and turns into recommendations. That engineering intelligence — asset health, failure risk, root cause, maintenance priority, engineering recommendation — goes to the engineer to review, verify, decide and approve, before any action (inspect, test, repair, replace, monitor, optimize) is taken.
That makes Division 03 — AI for AGL the intelligence engine connecting the AGL Failure Database, Smart AGL, IoT, Digital Engineering, Photometry, Predictive Maintenance and Asset Lifecycle systems — while keeping final safety-critical engineering decisions under qualified human technical control.
Quick answers about AI for AGL in the AGL Innovation Hub.
Division 03 is the core intelligence division of the AGL Innovation Hub. Its purpose is to investigate how Artificial Intelligence can be applied to AGL inspection, maintenance, diagnostics, engineering, testing, photometry, failure analysis, asset management and operational decision-making. It covers 18 topics (AID-01 to AID-18).
How can AI turn AGL asset, inspection, maintenance, testing and photometry data into failure predictions, fault diagnoses, defect detections, compliance checks, design assistance and maintenance recommendations — while keeping every safety-critical decision under qualified human technical control?
The 18 topics are AI-Based Predictive Maintenance, AGL Failure Prediction, AI Fault Diagnosis, Image-Based Inspection, AI Fixture Defect Detection, AI Cable Inspection, AI Sign Inspection, AI PAPI Inspection, Automated Compliance Checking, AI-Assisted AGL Design, AI Engineering Calculations, AI Test-Result Analysis, AI Photometry Analysis, AI Failure Classification, AI Root-Cause Analysis, AI Maintenance Recommendations, AI Asset-Health Prediction and AI Decision-Support Systems.
Division 03 is a technology domain division. It sits downstream of Division 01's strategic direction, alongside the other technology domains, and feeds intelligence forward into automation, control, photometry and predictive-maintenance divisions.
Division 03 is recommended to be structured around five layers, with human technical validation as the final and non-negotiable layer for safety-critical AGL applications. The steps or layers are: Layer 1 — AGL Data: Asset + Failure + Maintenance + Inspection + Testing + Photometry + IoT; Layer 2 — AI Processing: Machine Learning + Computer Vision + Pattern Recognition + Predictive Analytics; Layer 3 — Engineering Intelligence: Prediction + Diagnosis + Classification + Anomaly Detection + Risk Assessment; Layer 4 — Decision Support: Recommendation + Prioritization + Maintenance Planning + Engineering Assistance; Layer 5 — Human Technical Validation: Engineer Review → Technical Decision → Approved Action.
Every AI-for-AGL project should follow the same disciplined lifecycle, ending in continuous monitoring once implemented. The steps or layers are: Problem Definition; Data Collection; Data Cleaning; Data Labelling; Feature / Data Preparation; AI Model Development; Training; Validation; Performance Testing; Engineering Review; Pilot Deployment; Operational Validation; Controlled Implementation; Continuous Monitoring.
For safety-related AGL applications, the Innovation Hub should record and maintain the following for every deployed model. It includes Dataset source, Dataset size, Data quality, Training methodology, Validation methodology, Model version, Accuracy, False-positive rate, False-negative rate, Confidence level, Known limitations, Applicable operating conditions, Human-review requirements, Model update history and Approval status.
To make Division 03 operational rather than theoretical, these registers should be maintained and kept current. It includes AGL AI Project Register, AI Dataset Register, AI Model Register, AI Training Dataset Register, AI Validation Register, AI Failure Prediction Register, AI Fault Diagnosis Register, AI Inspection Register, AI Photometry Analysis Register, AI Compliance Review Register, AI Engineering Calculation Register, AI Asset Health Register, AI Maintenance Recommendation Register, AI Model Performance Register and AI Decision-Support Register.
KPIs are grouped across the seven functional areas AI touches within AGL — prediction, diagnosis, inspection, maintenance, photometry, engineering and overall program health. Prediction; Failure-prediction accuracy; Early-warning time; False-positive rate; False-negative rate; Diagnosis; Fault-diagnosis accuracy; Correct root-cause identification; Diagnostic time reduction; Inspection; Defect-detection accuracy; Inspection-time reduction; Image-processing success rate; Maintenance; Preventive-failure reduction; MTBF improvement; MTTR reduction; Maintenance-cost reduction; Photometry; Photometric anomaly-detection accuracy; Analysis-time reduction; Historical-performance detection; Engineering; Calculation-time reduction; Engineering-review efficiency; Design-error detection; Overall; Number of AI models validated; Number of AI pilots; Number of AI solutions implemented; Number of AI-assisted decisions; Measurable operational benefit.
The strongest feature of Division 03 is that it should not operate independently — it converts data supplied by other divisions into intelligence consumed by other divisions. It includes Division 02 — Smart Airfield Lighting generates connected lighting data, Division 03 — AI for AGL converts data into intelligence, Division 04 — IoT & Connected AGL provides sensors and connectivity, Division 05 — Digital Engineering provides digital models and digital twins, Division 12 — AGL Automation converts intelligence into automated processes, Division 13 — Digital AGL Control & Monitoring provides control and monitoring infrastructure, Division 14 — Advanced Photometry & Optical Innovation provides advanced photometric intelligence, Division 15 — Intelligent Maintenance & Predictive AGL converts predictions into maintenance strategies and Division 25 — Innovation Database & Project Intelligence stores the resulting innovation intelligence.
Use historical and real-time AGL data to identify assets that are likely to experience degradation or failure before the failure occurs. Potential Data Sources: Asset history, Failure history, Maintenance records, Inspection records, CCR operating data, Transformer data, Cable-test results, Fixture operating history, Photometric results and Environmental data.
Use AI to support AGL engineers during design development. Potential Applications: Design-option comparison, Circuit-layout assistance, Equipment selection, Design optimization, Design-document review, Design-risk identification, Design-data generation and Design consistency checking.
Predict the probability of AGL equipment or systems developing a failure. Failure Categories: Fixture failure, CCR failure, Isolation-transformer failure, Cable failure, Connector failure, Sign failure, PAPI failure, Control-system failure, Communication failure and Circuit failure.
Artificial intelligence in AGL starts with understanding the engineering problem — predictive analytics, computer vision, fault detection and machine-learning models applied to airfield lighting. The next step is building the technical foundation required to evaluate, develop and apply these technologies.
Develop structured knowledge across:
Understand how engineering problems become opportunities, technologies, concepts, validation activities and measurable improvements.
Understand today's AGL. Build tomorrow's capability.
Please read before relying on any content on this page or on any record downloaded from it.