Transnet Rail AI — Dashboard

Datasets Ingested

—

Models Active

—

Anomalies Flagged

Fleet Availability

—%

Activity Timeline

Anomaly events by domain · live feed

Domain Breakdown

Fault distribution

Traction

Braking

Bogies

Couplers

Energy

Maintenance

Active Alerts

Braking model drift detected — retrain recommended 12 min ago

Coupler fatigue threshold exceeded on corridor 4 43 min ago

Energy regen efficiency below baseline on unit TFR-1042 1 hr ago

Analytics

Triage trends and model performance metrics

Total Records Processed

—

Avg Model Accuracy

—

False Alarm Rate

—

Drift Events Detected

—

Category Distribution

Classification breakdown by severity

Critical

Urgent

Routine

Noise

Model Performance

Accuracy by domain

Fault Triage Feed

AI-classified engineering events awaiting engineer review

Traction Monitoring

Wheel-slip detection · motor torque anomaly pending

Awaiting data

Braking Awaiting Data

Pressure decay deviation · RUL model initialising

Awaiting data

Bogies Monitoring

Vibration RMS threshold · hunting flag inactive

Awaiting data

Couplers Awaiting Data

Shock-logger force signature · fatigue index pending

Awaiting data

Energy Monitoring

Regeneration efficiency drop · braking loss nominal

Awaiting data

Maintenance Awaiting Data

MTBF recalculation · OEE baseline pending

Awaiting data

The Challenge

TFR's Data Is Underutilised.

4,000+

estimated manual dataset review hours per year

The problem is not data collection. It is interpretation. TFR generates vast operational datasets across locomotive, rolling stock, and infrastructure value chains that conventional tools cannot process at scale.

Data Overload

Locomotive, telemetry, GPS, condition monitoring, and maintenance streams generate millions of records daily with no unified analytical layer.

Missed Fault Signals

Critical failure patterns buried in routine operational noise. Conventional tools lack the sensitivity to detect emerging faults before they cause downtime.

Zero Predictive Capability

Reactive maintenance dominates. Without AI/ML models, TFR cannot forecast RUL, MTBF degradation, or anomaly trajectories across its fleet.

Our Research Programme

Structured postgraduate talent deployed against TFR's priority engineering domains

Honours

Braking Systems · Bogies & Running Gear

Anomaly detection and RUL classification using on-board recorder datasets

Open for Recruitment

Duration: 12 months

Masters

Traction & Propulsion · Energy Management

Predictive modelling of motor torque degradation and regenerative braking efficiency optimisation

Open for Recruitment

Duration: 24 months

PhD

Couplers & Drawgear · Maintenance Planning

Digital twin development for coupler fatigue lifecycle modelling linked to SAP asset records

Pipeline

Duration: 36 months

Research Brief & Proposal

All programme documentation lives inside this platform

Transnet Freight Rail · AI Research Partnership Brief

Strategic overview, postgraduate programme structure, ISO/IEC 42001:2023 governance framework, IP assignment terms, and domain research roadmap. For internal circulation and academic partner engagement.

ISO/IEC 42001:2023 ISO/IEC 27001 POPIA

PDF · Updated: March 2026

Research Ecosystem

Institutional affiliations supporting this research programme

Member Organisation

South African AI Association

4,000+ AI practitioners across commercial, government, academic and NGO sectors. SA's largest responsible AI community.

Potential Academic Partner

Tshwane University of Technology

Gauteng-based institution with recognised engineering and AI/ML research faculty. SAAIA advisory board member. Potential engagement for postgraduate domain research alignment.

Under Consideration

Potential Research Partner

Council for Scientific & Industrial Research

South Africa's premier research and technology organisation. Gauteng-based, government-affiliated, and SAAIA advisory board member. Potential engagement as research and technology organisation partner per TFR URS requirements.

Under Consideration

AI Governance Framework

ISO/IEC 42001:2023 alignment — AI Management System

✓ AI Policy & Objectives Defined

✓ Risk Assessment for AI Systems In Progress

✓ Data Governance & Lineage Defined

✓ Model Transparency & Explainability Defined

✓ Human Oversight & Intervention Defined

✓ Continuous Improvement & Audit Trail Defined

Alignment demonstrated. Full certification pathway in progress under Global Apex Pty Ltd (Reg. 2025/623140/07)

ISO/IEC

42001:2023

AI Management System

Alignment Demonstrated

Certification pathway active

Domain Models

Configured ML models across all engineering domains

Traction & Propulsion

Wheel-slip detection, motor torque analysis, power output degradation

F1: — AUROC: —

Metrics

Chart placeholder

Anomaly Detection

Braking Systems

Pressure decay deviation, brake pad RUL estimation, ECPB analysis

MAE: — F1: —

Metrics

Chart placeholder

Regression

Bogies & Running Gear

Vibration RMS classification, hunting oscillation, wheel polygonisation index

F1: — RMSE: —

Metrics

Chart placeholder

Classification

Couplers & Drawgear

Shock-logger force signature analysis, coupler fatigue cycle estimation

MAE: — F1: —

Metrics

Chart placeholder

Regression

Energy Management

Regenerative braking efficiency, energy consumption per gross-tonne-km

MAPE: — RMSE: —

Metrics

Chart placeholder

Regression

Maintenance Planning

MTTF, MTBF, MTTR forecasting, OEE scoring, spares demand prediction

MAPE: — F1: —

Metrics

Chart placeholder

Regression

Data Ingestion

Pipeline status and dataset health across all TFR data streams

Source	Format	Domain	Records	Last Sync	Status
On-board Recorders	CSV	Traction	—	Awaiting	Pending
Telemetry & GPS	JSON	Bogies	—	Awaiting	Pending
Condition Monitoring	Parquet	Braking	—	Awaiting	Pending
Maintenance Records	CSV	Maintenance	—	Awaiting	Pending
SAP/Oracle Extracts	CSV	Maintenance	—	Awaiting	Pending
QuantumX Measurement	Parquet	Couplers	—	Awaiting	Pending
ECPB Brake Logs	CSV	Braking	—	Awaiting	Pending
Incident Event Logs	JSON	All Domains	—	Awaiting	Pending

Total Sources

Formats Supported

CSV · Parquet · JSON · OPC UA · MQTT

POPIA Status

Compliant

Completeness by Domain

Data coverage across ingested streams

Traction

Braking

Bogies

Couplers

Energy

Maintenance

Validation Checks

Automated data quality gates

✓ Schema Validation Enabled

✓ Null Detection Enabled

✓ Drift Monitoring Enabled

✓ Synchronisation Check Enabled

✓ POPIA Compliance Scan Enabled

Research Reports

Generated artefacts, model cards, and audit trail

PDF

Model Performance Report

Validation metrics across all domain models. F1, AUROC, MAE, RMSE, MAPE benchmarks.

Last generated: —

PDF

Anomaly Detection Summary

Flagged events by domain with confidence scores and false-alarm analysis.

Last generated: —

CSV

Data Pipeline Audit

Full ingestion log. Source integrity, completeness scores, drift residuals.

Last generated: —

PDF

Maintenance KPI Digest

MTTF, MTBF, MTTR, OEE scores with spares forecasting summary.

Last generated: —

Audit Trail

Model versioning and experiment tracking

Version	Domain	Action	Timestamp	User
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—

ISO/IEC 42001:2023 — AI Governance ISO/IEC 27001 — Information Security POPIA — Data Protection

Version	Domain	Action	Timestamp	User
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—

Version	Domain	Action	Timestamp	User
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—

Version	Domain	Action	Timestamp	User
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—
—	—	—	—	—