Three-year AI models for rail data analytics collaboration

Transnet SOC Ltd registered office: 49th Floor, Carlton Centre, 150 Commissioner Street, Johannesburg 2001.

The NDA itself does not specify technical or financial eligibility criteria for the AI service provider. It establishes a legal prerequisite: any company wishing to submit a bid must be willing and able to enter into this NDA. The company must act as a principal (not an agent) and must have appropriate technical and organizational measures for data security.

The bidder must supply a certificate signed by a director confirming full compliance with the requirement to expunge or destroy Confidential Information upon written demand.

Bidders must comply with confidentiality obligations under the Non-Disclosure Agreement. Key requirements include:

• Treat all Confidential Information received from Transnet as secret and confidential.
• Use Confidential Information only for technical and commercial discussions related to the bid or subsequent contract performance.
• Ensure agents who need to know Confidential Information are aware of confidentiality duties.
• Notify Transnet promptly if required by law to disclose Confidential Information.
• Maintain proper and secure storage of all Information.
• Keep written records of Confidential Information provided and its location.
• Return or destroy Confidential Information within seven days of a written demand from Transnet.
• Comply with South Africa's privacy protection under Section 14 of the Bill of Rights.
• Implement technical and organisational measures against unauthorised data processing, loss, or damage.

The Non-Disclosure Agreement includes the following key terms:

• Confidential Information must be kept secret and used only for bid-related purposes.
• Confidential Information may be disclosed to agents who need to know, provided they are bound by confidentiality.
• Disclosure is permitted if required by law, with prompt notification to Transnet.
• Unauthorised copying, disclosure, or use must be reported and remedied.
• Confidential Information remains the property of the disclosing party.
• Written records of Confidential Information must be kept and provided upon request.
• Upon Transnet's written demand, all Confidential Information must be returned or destroyed within seven days, with a director's certificate of compliance.
• No announcements about the bid may be made without prior written consent.
• Neither party may use the other's name or information for publicity without consent.
• Confidentiality obligations survive discussions and continue for five years.
• Each party acts as principal and bears its own costs.
• Transnet may assign the agreement to any Transnet Group member.
• The agreement is governed by South African law, with exclusive jurisdiction in South African courts.

The service provider must supply personnel for the following roles:

• AI Specialist
• Senior AI Researcher
• Data Scientist
• Data Engineer

Specific eligibility criteria are not provided in the given document excerpt. Applicants should refer to the full tender document for detailed eligibility requirements, which typically include relevant experience, financial stability, B-BBEE certification, and technical capabilities.

The service provider must supply personnel for the following roles:

• AI Specialist
• Senior AI Researcher
• Data Scientist
• Data Engineer

No specific requirements found

The project involves AI models for predictive and prescriptive analytics to improve asset/fleet reliability, maintenance planning, failure prevention, and operational decision-making in rail or similar transport sectors.

Returnable documents include: SAQA-registered qualifications for staff, statute/mandate and research QA policy for RTOs, and CHE recognition for HEIs. Failure to submit required evidence will lead to a score of zero. When calculating delivery periods, one week is taken as 5 business days; Saturdays and Sundays are not counted.

Bidders must be either Higher Education Institutions (HEI) with CHE recognition and SAQA-registered qualifications, or Research and Technology Organizations (RTO) with founding acts/mandates and research quality policies. All bidders must submit signed IP ownership declarations, background IP registers, and POPIA compliance documentation.

The scope requires AI models for data analytics collaboration focused on predictive and prescriptive analytics to improve asset/fleet reliability, maintenance planning, failure prevention, and operational decision-making in rail or similar transport sectors.

Key deliverables and requirements:

• Provide model architecture and approach explanation (e.g., machine learning, deep learning, time-series forecasting).
• Demonstrate live deployment in rail or similar transport sectors showing improvements through predictive analytics.
• Show integration capability with existing railway management or similar transport/logistics systems, including supported data formats.
• Provide evidence of model performance, accuracy, and precision.
• Demonstrate little to no-coding capabilities of the models.
• Submit training manual and implementation plans, with provision for training material, classroom training, and practical training.
• Submit three reference letters for minimum of 3 projects, each containing customer company name, contact details, project scope, and start/end dates.
• Transnet reserves the right to contact clients to verify information.

Methodology: Demonstrate little to no-coding capabilities of the model(s).

Experience: Letter from other companies confirming successful implementation of similar project.

Quality management requirements:

• For HEIs: Provide CHE recognition letter/URL and SAQA registration of qualifications.
• For RTOs: Provide founding act/mandate, governance charter, research quality policy, and external quality marks (e.g., ISO 9001, ISO 27001, NRF ratings).

Scoring for institutional legitimacy & quality assurance:

• No evidence: 0%.
• Partial (one document): 25%.
• Full for chosen track (HEI: CHE/SAQA documents; RTO: governance, statute/mandate + research QA policy): 50%.
• As above plus external quality marks: 100%.

Bidders must submit cost models from previous similar AI model projects, including a cost breakdown structure related to the three reference letters submitted.

Scoring for cost model evidence:

• No cost breakdown structure submitted: 0%.
• One relevant cost breakdown structure submitted: 25%.
• Two relevant cost breakdown structures submitted: 50%.
• Three or more relevant cost breakdown structures submitted: 100%.

Compliance requirements include:

• IP and Data Governance: Submit signed IP ownership declaration, background IP register, POPIA-aligned data-handling, secure environment description, publication approval process, and written undertaking confirming background IP licensing does not restrict TFR foreground IP use.
• Evidence of compliance with data privacy (POPIA) and cybersecurity requirements.
• Professional registrations for key personnel (e.g., ECSA, PMI).
• Institutional legitimacy: For HEIs, provide CHE recognition and SAQA-registered qualifications; for RTOs, provide statute/mandate and research QA policy.

Contractual terms include IP and data governance requirements: bidders must submit signed IP ownership declaration, background IP register, POPIA-aligned data-handling, secure environment description, publication approval process, and acceptance of TFR publication approval requirements. Background IP licensing must not restrict TFR foreground IP use.

Special condition: When calculating the delivery period, 1 week is taken as 5 business days; Saturdays and Sundays are not counted.

Delivery period scoring: Project plan showing delivery within the shortest specified timeframe (within 36 months) and confirmation letter from other companies or Transnet Divisions = 100%.

• Appointment of a service provider to provide artificial intelligence (AI) models for data analytics collaboration.
• Contract duration: Three (3) years for Transnet Freight Rail.
• Services delivered via Purchase Orders and governed by a Master Agreement.

Returnable Documents: No specific submission guidelines found in the extracted text. Refer to the main tender documents for submission instructions.

Service Provider must: Be a registered legal entity; Have a valid B-BBEE Verification Certificate (renewals must be maintained); Not appear on National Treasury's Register of Tender Defaulters or List of Restricted Suppliers; Have tax clearance compliance; Possess necessary insurances; Have the capacity to enter into a 3-year contract; Ensure personnel have right to work in South Africa.

• Scope: Provide artificial intelligence (AI) models for data analytics collaboration for Transnet Freight Rail.
• Duration: Three (3) years.
• Service Level Agreement (SLA) required, detailing processes, deliverables, key performance indicators, and performance standards.
• Services to be delivered via Purchase Orders issued by Transnet.
• The Service Provider must comply with railway safety requirements per the National Railway Safety Regulator Act, 16 of 2002.
• Subcontracting is permitted with Transnet approval, but the Service Provider remains liable for performance.

• Pricing detailed in Schedule of Requirements/Work Orders.
• Prices subject to review and adjustment negotiations; market comparisons may be used.
• Payment terms: Transnet pays within 30 calendar days after receipt of valid, undisputed Tax Invoice and supporting documentation.
• Interest may be charged on late payments at Standard Bank's prime rate.
• Transnet may withhold payments if irregular expenditure is suspected.
• VAT is applicable and must be indicated separately on invoices.

• B-BBEE: Service Provider must submit B-BBEE status details annually and notify Transnet of any negative changes. Non-compliance may lead to termination.
• Tax Compliance: Service Provider must ensure Tax Clearance Compliance for the duration of the agreement and comply with all tax legislation.
• Labour Laws: Must observe all South African labour legislation, including the Occupational Health and Safety Act, 85 of 1993.
• Environmental Compliance: Must comply with environmental legislation and have an environmental management policy.
• Supplier Integrity Pact: Must adhere to Transnet's Supplier Integrity Pact to avoid dishonesty, fraud, and corruption.
• Subcontracting Restrictions: Subcontracting limited to 25% of contract value to enterprises with equal or higher B-BBEE status, unless subcontracting to an Exempted Micro Enterprise (EME).
• National Treasury Restrictions: Bidders on the Register of Tender Defaulters or List of Restricted Suppliers are excluded.

• Project: Appointment of a service provider to provide Artificial Intelligence (AI) models for data analytics collaboration for Transnet Freight Rail.
• Duration: Three (3) years.
• Background: TFR seeks to enhance operational efficiency and data-driven decision-making using its large volumes of operational data from locomotives, rolling stock, and infrastructure.
• Collaboration Goal: Partner with a Gauteng-based Academic Institution or research technology organisation to develop AI/ML models or a unified analytics platform.
• Focus Areas: Predictive maintenance, asset reliability, incident investigation, resource optimisation, and strategic planning insights.

• Closing Date: 12 August 2026.
• Closing Time: 10:00 AM.
• Non-Compulsory Briefing Session: 24 July 2026 at 11:00 AM, conducted on Microsoft Teams.
• Clarification Deadline: Submit clarification requests by 31 July 2026.
• Bid Validity Period: 180 days after the closing date.

• A non-compulsory briefing session will be held on 24 July 2026 at 11:00 AM via Microsoft Teams.
• Confirm attendance and send attendee details to [email protected].

• Contact Person: Gladys Mtambo.
• Telephone: 011 584 0597.
• Email: [email protected] (for technical enquiries and briefing confirmations).
• Complaints Email: [email protected].
• Submission Portal: Transnet e-Tender Submission Portal (https://transnetetenders.azurewebsites.net).

• Submit via Transnet e-Tender Submission Portal: https://transnetetenders.azurewebsites.net.
• Closing date: 12 August 2026 at 10:00 AM. No late submissions accepted.
• Upload at least one day before closing to avoid technical issues.
• Each upload limited to 30MB; multiple uploads permitted.
• Each company must register its own profile and submit under its own name.
• Joint Ventures may submit using one party's registered profile.
• Sign, stamp, and date every page of your submission.
• Include all returnable documents listed in Section 5.
• Bid validity period: 180 days after closing date; may be requested to extend.
• Clarifications deadline: Submit RFP Clarification Request Form to [email protected] by 31 July 2026.
• Complaints: Email [email protected].

• Refer to Section 5 for the complete list of returnable documents that must be submitted with your proposal.

Gauteng-based academic institutions or research organizations preferred; must provide proof of institutional legitimacy (CHE recognition for HEIs or statute for RTOs); demonstrate AI capability with rail/transport sector experience; key personnel must have relevant qualifications and professional registration (e.g., ECSA, PMI); comply with B-BBEE requirements and tax obligations; submit valid B-BBEE certificate or sworn affidavit.

• Scope: Provide Artificial Intelligence (AI) models for data analytics collaboration for Transnet Freight Rail over three years.
• Objective: Develop analytical models or a unified digital platform to enhance data analytics capabilities using TFR's operational data.
• Preferred Suppliers: Accredited Gauteng-based Academic Institutions or research and technology organisations with AI/ML expertise. Open-market suppliers are less preferred.
• Key Aims:

• Enable predictive and prescriptive analytics for asset/fleet reliability, maintenance planning, and failure prevention.
• Develop domain-specific AI models for traction, brakes, bogies, couplers, and energy management.
• Enhance incident investigation via data correlation and anomaly detection.
• Optimise resource scheduling and deployment.
• Assist with predictive maintenance and provide visual insights for decision-making.
• Encourage innovation and capacity building through student involvement and joint R&D.

• Technical Evaluation Requirements:

• Evidence of previous AI model deployments in rail or similar transport sectors.
• CVs of key personnel with engineering and project management qualifications, registered with professional bodies (e.g., ECSA, PMI).
• Training manuals and implementation plans.
• Clear project plans and evidence of previous AI model applications.

• The supplier must be fully responsible for acts/omissions of its personnel.
• Must comply with all RFP requirements.
• Refer to evaluation methodology for structured approach.

• Key Personnel: Submit CVs for key personnel, showing qualifications and registration with professional bodies (e.g., ECSA, PMI).
• Experience: Provide evidence of previous AI model deployments in rail or similar transport sectors.
• Project Plans: Submit clear project plans and confirmation letters from previous clients.

• Bidders must demonstrate quality assurance: For Research and Technology Organisations (RTOs), provide statute/mandate and research QA policy.
• External quality marks (e.g., ISO 9001, ISO 27001, NRF ratings) are advantageous.
• IP & Data Governance: Submit signed IP ownership declaration, POPIA-aligned data-handling procedures, and breach-response protocol.

• Detailed pricing must be submitted as per Annexure C.
• Prices are estimates; orders will be 'as and when required'.
• Rates must align with DPSA consultant fee guidelines.
• Prices quoted must be kept firm for the contract duration if awarded.

• Pricing Format: Submit pricing strictly as per Annexure C (Pricing Schedule). Deviations may render bid non-responsive.
• Currency: All prices must be in South African Rand, inclusive of VAT.
• Basis: Prices quoted on a delivered basis to Halfweg.
• Rate Alignment: Proposed rates must align with the DPSA 'Guide on Hourly Fee Rates for Consultants'.
• Discounts: Any discounts must be unconditional to be considered.
• Imported Goods: If applicable, use SARB exchange rate from date of tender advertisement.
• Price Negotiations: Transnet may engage in post-tender negotiations if initial prices are not market-related.
• Contract Duration: If awarded, prices must be kept firm and fixed for the contract duration.

• Tax Compliance: Bidders must submit SARS Tax Compliance Status (TCS) PIN or printed TCS certificate. Alternatively, provide Central Supplier Database (CSD) number if registered.
• Central Supplier Database (CSD): All respondents (except foreign suppliers with no local entity) must self-register on the National Treasury CSD.
• B-BBEE: Submit a valid B-BBEE Status Level Verification Certificate or Sworn Affidavit (for EMEs & QSEs).
• Joint Ventures: For evaluation, a JV requires one consolidated B-BBEE scorecard.
• Employment Equity Act: Compliance with the Act, including Section 53, is required.
• General Compliance: The successful supplier must comply with all applicable laws and regulations.
• Foreign Suppliers: Must complete a questionnaire regarding RSA residency, branch, establishment, and income.

• Submit a B-BBEE Status Level Verification Certificate or Sworn Affidavit (for EMEs & QSEs).
• Certificate must be issued by an Accounting Officer, SANAS-accredited verification agency, or Registered Auditor.

• The supplier must comply with the Occupational Health and Safety Act.
• Compliance with the National Railway Safety Regulator Act is required.
• Bidders should describe control measures for OHS compliance in their proposal.

• Transnet requests details of your entity's environmental policies regarding waste disposal, recycling, and energy conservation.
• Submit information on your environmental commitments and carbon footprint as part of your proposal.

• Contract governed by Transnet's General Bid Conditions (Annexure F) and Supplier Integrity Pact (Annexure G).
• Master Agreement provided in Annexure E.
• Non-Disclosure Agreement in Annexure H.
• Legal review of proposals by Transnet's counsel; material deviations may cause disqualification.
• Security clearance may be required for successful respondent's personnel to CONFIDENTIAL/SECRET/TOP SECRET level.
• Transnet reserves rights to modify scope, split awards, cancel the process, and validate submitted information.

• Tax Compliance: Ensure tax matters are in order with SARS. Submit TCS PIN or certificate.
• CSD Registration: Self-register on the National Treasury Central Supplier Database (https://secure.csd.gov.za/).
• Foreign Suppliers: Complete questionnaire regarding RSA presence and income.

• See main evaluationCriteria section for detailed stages and weightings. Refer to Annexure D for the Scoring Matrix.

• The service involves providing AI models for data analytics collaboration, to be delivered in conformity with defined specifications.

Bidders must not be on National Treasury's Database of Restricted Suppliers; must have no serious breaches of law (corruption, fraud, etc.) in the last five years; must not have conflicts of interest with Transnet employees; and must comply with all anti-corruption and ethical standards outlined in the pact.

• The service provider must deliver AI models for data analytics collaboration in conformity with defined specifications.
• Transnet may require the bidder to furnish a copy of its code of conduct, which must include a compliance programme and reject bribery and unethical conduct.
• The bidder must uphold the ten principles of the United Nations Global Compact (UNGC) in Human Rights, Labour, Environment, and Anti-Corruption.
• A 'competitor' is defined as any individual or organisation, whether affiliated or not, who has been requested or could potentially submit a bid, or provides the same goods/services.

• A 'competitor' includes any individual or organisation who has been requested to bid, could potentially bid based on qualifications/abilities/experience, or provides the same goods/services.

• The Integrity Pact aims to secure the contract at a reasonable and competitive price.
• Bidders must avoid under-pricing and anti-competitive practices.
• The objective is to enable Transnet to obtain the desired contract at a competitive price in conformity with specifications.

• Bidders must comply with all relevant laws, regulations, and ethical standards.
• Making an incorrect statement regarding Local Content or B-BBEE status can lead to restriction, unless proven to be made in good faith with reasonable steps taken to verify correctness.
• Submitting false information under the Preferential Procurement Regulations, 2017 is a ground for restriction.
• Bidders must declare no serious legal transgressions (e.g., corruption, fraud) in the last five years with any public or private sector entity.
• Bidders found on the National Treasury Register of Tender Defaulters or List of Restricted Suppliers will not be awarded the contract.
• Transnet's 'Tip-Off Anonymous' hotline (0800 003 056) is available to report dishonest behaviour.

• Bidders must uphold the UN Global Compact environmental principles: a precautionary approach to environmental challenges, initiatives promoting greater environmental responsibility, and encouraging environmentally friendly technologies.

• All bidders must read and certify agreement with the Integrity Pact in the RFX Declaration Form. It will be automatically incorporated into the final contract.
• Transnet and the bidder agree to avoid dishonesty, fraud, corruption, and anti-competitive practices.
• Transnet commits to prevent corruption, treat bidders fairly, provide equal information, and exclude employees with personal interests.
• Bidders must adhere to a 'Zero Gifts' policy, prevent corrupt practices, avoid collusion, not misrepresent facts, and uphold the UN Global Compact principles.
• Bids must be arrived at independently without collusion on prices, markets, formulas, or submission intentions.
• Suspicious bids may be reported to the Competition Commission and/or National Prosecuting Authority.
• Transgressions can lead to disqualification, removal from the vendor database, or exclusion from future bids for up to 10 years or more.
• Bidders on the National Treasury Register of Tender Defaulters or List of Restricted Suppliers will not be awarded the contract.
• Grounds for restriction include bid withdrawal in bad faith, failure to sign a contract, unsatisfactory performance, bribery, fraud, false statements on Local Content/B-BBEE, and litigation in bad faith.
• Conflicts of interest must be disclosed in the RFX declaration form or notified immediately in writing.
• Disputes should be resolved amicably; litigation in bad faith (vexatious, perjury, scurrilous allegations, abuse of process) is prohibited.
• The Integrity Pact is governed by South African law, valid indefinitely, and includes a confidential hotline (0800 003 056) for reporting misconduct.

• All bidders must sign and comply with the Integrity Pact to be registered as a Transnet vendor.
• Foreign bidders must disclose agents/representatives in South Africa; South African bidders must disclose foreign principals involved in the bidding process.

• Bids will be evaluated fairly and transparently. Transnet commits to treat all bidders with equity and fairness.
• Any attempt to gain an advantage in the bidding process, bid evaluation, contracting, or implementation through improper means is prohibited.
• Submitting false information regarding subcontracting arrangements or other matters required by the Preferential Procurement Regulations, 2017 will affect the evaluation of a bid.
• Bidders must not collude regarding the quality, quantity, specifications, conditions, or delivery particulars of goods or services.

{"closingDate":"00 February 2026"}

{"name":null,"email":null,"phone":null,"department":"HPC High-Performance Computing","address":null}

Gauteng-based South African academic institution or research organization with recognized AI/ML research expertise. Must have secure research infrastructure, POPIA compliance, and ability to handle sensitive TFR data. Collaboration focuses on research and prototyping, not production implementation.

Transnet Freight Rail (TFR), through the Technology Management function in the Resource Management Division,

seeks to collaborate with a Gauteng-based Academic Institution or a research and technology organization with

advanced Artificial Intelligence (AI) and Machine Learning (ML) expertise, to research, design, and prototype

Artificial Intelligence (AI) and Machine Learning (ML) analytical models that enhance the value, utilisation, and

interpretation of TFR’s operational and engineering datasets that will be used to provide a single, reliable source

of information for the organization.

TFR generates large and diverse volumes of data across its locomotive, rolling stock, and infrastructure value

chains. These include:

• on-board operational and event-recording systems
• telemetry, GPS and location-based feeds
• asset identification and tracking data
• condition-monitoring and diagnostic outputs
• maintenance, inspection and operational activity records
• specialised measurement, test-run and instrumentation datasets
• incident, exception and event-based operational logs

Extracting actionable insights from these complex, high-volume datasets using conventional tools has become

increasingly challenging. To strengthen data-driven reliability, maintenance optimisation, and operational decisionmaking, TFR intends to establish a research collaboration with universities that possess advanced AI/ML

capabilities.

The collaboration will focus on applied research and the development of analytical models, targeting priority

engineering and operational domains such as traction performance, braking systems, bogie and wheel behaviour,

coupler dynamics, energy management and incident analytics. Activities will include data preparation, modelling,

feature engineering, analysis, visualisation, and knowledge transfer.

This initiative aims to:

• enable predictive, prescriptive, and diagnostic analytics to improve asset reliability;
• provide data-driven insights that support maintenance planning, operational optimisation and risk

mitigation;

• promote innovation, skills development and student involvement; and
• establish a sustainable industry–academia partnership that supports TFR’s operational excellence and

contributes to South Africa’s digital research ecosystem.

Overall, the collaboration reflects TFR’s commitment to leveraging modern AI methodologies to advance nextgeneration rail analytics and support long-term technology transformation.

1.1 Scope of Collaboration

This specification defines the scope of a collaborative research partnership between Transnet Freight Rail (TFR)

and a South African university with recognised Artificial Intelligence (AI) and Machine Learning (ML) research

capability. The collaboration focuses on research, development, and prototyping analytical models that support

TFR’s operational, engineering, and asset-management objectives.

The scope covers research activities only; production deployment, enterprise integration, and software

implementation remain out of scope unless explicitly contracted by TFR.

1.1.1 AI/ML Research and Prototype Development

of 11

Bbh----

Design, experimentation, and prototyping of AI/ML analytical models using TFR’s operational and engineering

datasets. Model outputs may include predictive, prescriptive, diagnostic, optimisation, classification, or anomalydetection algorithms aligned to TFR’s operational context.

1.1.2 Data Preparation and Enrichment

Preparation of datasets required for modelling, including:

• data cleaning, synchronisation, and validation
• transformation, harmonisation, and feature engineering
• enrichment with contextual or environmental variables where appropriate and approved by TFR

Any field instrumentation, new sensors, or hardware trials must be pre-approved in writing by TFR and, if required,

separately contracted.

1.1.3 Analytics, KPIs, and Visual Insight Development

Creation of research-level analytics, including:

• exploratory data analyses
• model-based diagnostic indicators
• prototype KPIs and visualisations
• conceptual dashboards or insight views

These artefacts are intended to support engineering and operational decision-making.

1.1.4 Operational and Maintenance Workflow Concepts

Development of conceptual workflow models to illustrate the potential operational application of AI/ML outputs

(e.g., incident triage, maintenance prioritisation, anomaly alerts, or reporting automation).

1.1.5 Functional Domains of Application

The collaboration may span multiple domains, including but not limited to:

• traction and braking systems
• bogies, running gear, and wheel–rail interaction
• couplers and drawgear
• energy and auxiliary systems
• condition monitoring and diagnostic subsystems
• operational performance and fleet reliability
• maintenance planning and asset-health modelling

Specific research topics will be agreed with TFR for each work package.

1.1.6 Integration and Architecture Research Concepts

Where relevant, the university may propose integration-ready research artefacts, such as:

• Conceptual API designs
• open-format data-exchange schema
• reference architectures or data pipelines suitable for future implementation

1.1.7 Governance, Security, and Model Lifecycle Practices

The university must apply research-level model-governance principles, including:

• experiment tracking and reproducibility
• versioning of models, datasets, and code
• documentation of data lineage
• explainability and transparency of model behaviour

of 11

Bbh----

• secure storage and handling of sensitive TFR datasets
• compliance with POPIA and TFR ICT security requirements

Where applicable, the university’s practices should align with recognised standards such as ISO/IEC 27001 for

information security.

1.1.8 Collaboration Deliverables and Responsibilities

The university must:

• produce prototype models and research artefacts as defined with TFR
• provide reproducible methodologies and clear documentation
• participate in technical review sessions with TFR
• support knowledge transfer, including postgraduate student involvement
• deliver all outputs in an open, transparent, and implementation-ready format

2 functional and technical requirements

The selected university must demonstrate the capability to undertake research-grade AI/ML modelling, advanced

data analysis, and prototype development using diverse operational and engineering datasets from Transnet

Freight Rail (TFR). All work shall be performed in a secure, POPIA-compliant research environment and must

contribute directly to improving TFR’s operational insight, reliability, maintenance planning, and engineering

decision-making.

2.1 Functional Requirements

The university shall deliver research projects that achieve the following outcomes:

2.1.1 Core Analytical Capabilities

• Ingest, process, and analyse TFR datasets originating from on-board recorders, asset-tracking

systems, telemetry feeds, diagnostic platforms, maintenance systems, test-runs, and incident logs.

• Develop predictive, diagnostic, prescriptive, optimisation, classification, or anomaly-

detection models relevant to TFR’s assets and operating environment.

• Implement and evaluate automated anomaly-detection and early-warning methods for

latent faults or emerging failure patterns.

• Provide prototype KPI and analytical visualisations, including reliability indices, utilisation

measures, and energy-performance views.

• Develop integration-ready research artefacts, such as API specifications, data schemas, or open-

format data-exchange models suitable for future incorporation into TFR enterprise systems.

• Produce insight-generation outputs such as comparative analyses, fault clusters, trend

evaluations, and investigative summaries for engineering teams.

• Support AI-assisted reporting and simulation concepts for maintenance, energy optimisation,

and operational decision support.

• Use modular, extensible modelling approaches that support future expansion, new datasets,

and iterative improvements.

• Ensure secure dataset handling, controlled access, and compliance with POPIA and TFR

data-governance policies.

2.2 Data and Analytical Capability Requirements

2.2.1 Data Sources and Acquisition

The bidders should be capable of processing data from:

• on-board operational data recording systems
• asset identification and tracking datasets

of 11

Bbh----

• telemetry, GPS, and location-based data streams
• condition-monitoring, diagnostic, and telemeter systems
• maintenance and inspection records (SAP/Oracle/DB2 extracts)
• specialised measurement systems (QuantumX, accelerometers, strain/vibration loggers, rope

transducers)

• production-train datasets (air-brake, ECP, pneumatic logs)
• incident, exception, and event-based operational logs

The bidder shall perform cleaning, validation, synchronisation, harmonisation, and feature engineering necessary

to prepare data for modelling.

2.2.2 Analytical and Model Development Capability

Research outputs may include:

• supervised, unsupervised, and reinforcement learning models
• fault-pattern discovery and causal inference (where applicable)
• context-aware models (weather, gradient, load, traffic density)
• experiments using historical batch data or simulated real-time streams
• feedback-loop retraining and drift-assessment experiments

2.3 Functional Domains of Application

Models may be applied across multiple locomotive and rolling-stock domains, including:

• Traction and Propulsion Systems

o analysis of traction-converter and inverter parameters

o motor torque-temperature correlation, wheel-slip detection

o energy-consumption modelling and regeneration efficiency

o DC-link stability, surge-arrestor behaviour, cooling performance

• Brakes Technology

o wear-rate estimation, brake-actuator diagnostics

o pneumatic leak detection, pressure-decay deviations

o ECPB log analysis, valve timing, communication errors

o Remaining Useful Life (RUL) models

• Bogies and Running Gear

o detection of bogie vibration, hunting, wheel polygonisation

o fusion of accelerometer/axle-load/track-geometry data

o reinforcement-learning-based overhaul optimisation

• Couplers and Drawgear

o shock-logger force-signature analysis

o lifecycle modelling and fatigue prediction

o digital-twin concepts linked to SAP

• Electrical, Power and Auxiliary Systems

o transformer, rectifier, inverter health monitoring

o current/voltage imbalance, overheating, insulation tracking

o compressor, motor, and auxiliary load monitoring

• Pantograph, OHTE and ADD Events

o ADD/pantograph-strike detection and localisation

of 11

Bbh----

o line-voltage correlation, environmental context analytics

o Diesel Engine and Fuel Systems (where applicable)

o engine-health modelling, fuel-consumption optimisation

o degradation modelling for turbochargers, injectors, filters

o integration with SARS Diesel Rebate compliance datasets

• HVAC, Cab and Crew Environment

o cab temperature/humidity performance

o HVAC subsystem fault prediction

• Energy Management and Regeneration

o regeneration efficiency analysis

o braking-energy loss assessment

• Protection, Safety and Control Systems

o VCB/circuit-breaker fault-log analytics

o vigilance-overspeed-penalty event correlation

• Simulation and In-Service Testing

o train-performance analysis and black-box event reconstruction

o THG and driver-performance assessments

• Maintenance Planning and Scheduling

o predictive maintenance KPIs (MTTF, MTBF, MTTR, OEE, availability)

o spares forecasting

o tool and calibration management concepts

• Administrative and Process Optimisation

o AI-assisted routing, claims processing, documentation drafting

o semantic search over historical records

2.4 Research Tooling, Infrastructure, and Integration Readiness

The bidder shall provide:

• GPU-enabled or HPC research computing environments
• secure, POPIA-aligned research data-storage environments
• model versioning, experiment tracking, reproducibility, and audit trails
• open-format integration artefacts (CSV, Parquet, JSON, REST, MQTT, OPC UA—where applicable)
• conceptual API designs or data schemas suitable for LDS/SAP/Telemetry/VIS
• automated data-quality scoring and inference-logging pipelines at research level
• low-/no-code prototypes where beneficial to engineering teams

Production system integration is not part of this scope unless expressly contracted.

2.5 Performance and Validation Requirements (Research-Level)

Each research project must define appropriate validation criteria, but the following baseline applies:

2.5.1 Data Quality & Processing Metrics

• ingestion and processing times
• completeness, error rates, and integrity checks
• multi-stream synchronisation accuracy and drift residuals
• scalability trials on HPC or equivalent environments

of 11

Bbh----

2.5.2 Model Accuracy & Reliability Metrics

• precision, recall, F1-score, AUROC (classification)
• MAE, RMSE, MAPE (regression)
• confusion-matrix analysis
• false-alarm analysis and drift monitoring
• benchmarking against baseline and peer models

2.5.3 Prototype Responsiveness & Visualisation

• insight refresh rates suitable for demonstration
• query turnaround time
• alert/event latency in research harnesses

2.5.4 Interoperability & Security

• conformance to data schemas and interface definitions
• test-integration success rates
• adherence to POPIA and TFR ICT security requirements

2.5.5 Operational Value Demonstration

• case studies showing potential improvements to reliability, availability, MTBF/MTTR, energy efficiency, or

maintenance effectiveness

• engineering feedback from TFR review workshops

These are research metrics, not production SLAs.

2.6 Project Delivery, Governance, and Knowledge Transfer

The bidder shall demonstrate structured research-project governance, including:

• research plans, milestones, risk registers, and ethics/data-access approvals
• reproducible documentation (data schemas, feature lists, modelling assumptions)
• code notebooks, workflows, model cards, and reproducibility packs
• technical walkthroughs and knowledge-transfer sessions
• postgraduate student involvement where appropriate
• continuous improvement and research governance aligned with ISO/IEC 42001:2023
• periodic retraining or recalibration of models using updated datasets

2.7 Constraints and Compliance Requirements

The bidder shall operate within the following constraints:

• Data access: NDAs, POPIA compliance, and ethics clearance where necessary
• Security: all modelling and analysis must occur within secure, approved research environments
• Data availability: some datasets may be intermittent or time-bound; TFR will define access per project
• Publication: any academic output requires prior written TFR approval
• Intellectual Property: all foreground IP developed under this collaboration remains the property of TFR

This specification outlines the essential requirements for a research partnership that will enable Transnet Freight

Rail to advance its AI and data-analytics capabilities. By providing a structured framework for secure, innovative,

of 11

Bbh----

and high-value model development, the collaboration aims to strengthen TFR’s operational insight, support

informed engineering decisions, and contribute to long-term digital transformation across the organisation

of 11

Design, experimentation, and prototyping of AI/ML analytical models using TFR’s operational and engineering

datasets. Model outputs may include predictive, prescriptive, diagnostic, optimisation, classification, or anomaly-

detection algorithms aligned to TFR’s operational context.

1.1.7 Governance, Security, and Model Lifecycle Practices

• experiment tracking and reproducibility
• versioning of models, datasets, and code
• documentation of data lineage
• explainability and transparency of model behaviour

of 11

• secure storage and handling of sensitive TFR datasets
• compliance with POPIA and TFR ICT security requirements

Where applicable, the university’s practices should align with recognised standards such as ISO/IEC 27001 for

information security.

Author: Thenjiwe Mtsheku

Train Design Technologies 03/02/2026

Rev 00 February 2026 Thenjiwe Mtsheku

of 11

• on-board operational data recording systems
• asset identification and tracking datasets

of 11

to prepare data for modelling.

evaluations, and investigative summaries for engineering teams.

• automated data-quality scoring and inference-logging pipelines at research level

2.5.1 Data Quality & Processing Metrics

• precision, recall, F1-score, AUROC (classification)

• Contract Quantities: Transnet does not guarantee a specific purchase volume; orders are issued based on operational requirements.
• Delivery: Lead times must include all non-working days and holidays.

{

"briefingSession": {

"date": null,

"time": null,

"venue": "Refer to RFX document for site visit/briefing details. Attendance may be compulsory.",

"is_compulsory": false

}

}

• Site Visits: Respondents may be requested to attend briefings or site visits. If marked as compulsory in the RFX, failure to attend will result in disqualification.

• Clarifications: Direct queries to the contact person listed in the RFX Document via the Transnet e-Tender Submission Portal.
• Post-Closing Communication: Only communication with the contact person listed in the RFX Document is permitted after the closing date.

• Submission Method: Bids must be uploaded to the Transnet e-Tender Submission Portal (transnetetenders.azurewebsites.net).
• Deadline: Bids must be submitted no later than the specified closing date and time. Late bids will not be considered.
• Completeness: All Bid Documents must be completed in their entirety. Failure to submit mandatory returnable schedules/documents will result in disqualification.
• Form Usage: Respondents must use official Transnet forms/formats. Using own terms and conditions may result in rejection.
• Price Alterations: Any changes to bid prices must be made by deleting incorrect figures and inserting correct ones, and must be initialled by the signatory. Failure to do so may result in exclusion of the item from the award.
• Communication: All clarifications must be uploaded to the Transnet e-Tender Submission Portal before the closing date. Unauthorised communication regarding the bid during the evaluation period may lead to disqualification.

• Mandatory Documents: All returnable schedules listed in the RFX must be submitted; failure to do so results in disqualification.

Not listed on National Treasury's Register of Tender Defaulters or List of Restricted Suppliers; Must comply with all mandatory submission requirements; Must attend compulsory site visits/briefings if specified; Must submit all required samples by deadline if applicable; Foreign respondents must have authorized South African representatives with proper Power of Attorney.

• Scope of Supply: Transnet is not bound to purchase a definitive quantity; orders will be placed as required against the contract.
• Delivery Basis: Unless otherwise specified, prices must be on a Delivered Duty Paid (latest ICC Incoterms) basis to the specified destination.
• Quality: Unless otherwise stipulated, all goods must be new (unused, not second-hand or reconditioned).
• Samples: If required, samples must be marked with the Respondent's name, address, Bid number, and item number, and delivered before the closing date.
• Drawings/Specifications: Bids must strictly adhere to Transnet's drawings and specifications. Foreign specifications (other than British/American standards) require translated copies and a covering letter detailing any variations.

23 respondent's samples

23.1 If samples are required from Respondents, such samples shall be suitably marked with the Respondent's name and address, the Bid number and the Bid item number and must be despatched in time to reach the addressee as stipulated in the Bid Documents on or before the closing date of the Bid. Failure to submit samples by the due date may result in the rejection of a Bid.

23.2 Transnet reserves the right to retain samples furnished by Respondents in compliance with Bid conditions.

23.3 Payment will not be made for a successful Respondent’s samples that may be retained by Transnet for the purpose of checking the quality and workmanship of Goods/Services delivered in execution of a contract.

23.4 If Transnet does not wish to retain unsuccessful Respondents’ samples and the Respondents require their return, such samples may be collected by the Respondents at their own risk and cost.

27 quality of material

Unless otherwise stipulated, the Goods offered shall be NEW i.e. in unused condition, neither second-hand nor reconditioned.

• Local Supplies: Quoted on a Delivered RSA named destination basis.
• Imported Supplies: Quoted on a Delivered Duty Paid (latest ICC Incoterms) basis to the end destination in South Africa.

• Currency: All monetary amounts must be in South African Rand (ZAR) unless specifically permitted otherwise.
• Pricing: Prices subject to confirmation will not be considered.
• VAT: Local supply prices must be inclusive of VAT (shown separately on the Tax Invoice). For foreign services, the local portion (commission) must show VAT separately.
• Payment: Transnet prefers Rand-based agreements. Alternative financing proposals may be evaluated if full details are provided.
• Exchange/Remittance: Transnet can remit a portion of the contract value directly to a foreign principal/manufacturer if requested at the time of bidding. South African Reserve Bank approval is required for foreign currency payments.

• Registration: Bidders must not be on the National Treasury Register of Tender Defaulters or the List of Restricted Suppliers.
• Identification: Full names of directors (companies), members (CCs), or partners (partnerships/individuals) must be stated in the Bid.
• Securities: Successful bidders may be required to provide a Deed of Suretyship from an approved South African bank, building society, or insurance corporation within 30 days of acceptance.
• Foreign Respondents: Must provide written proof of authorisation from their South African representative/agent. A legal Power of Attorney (complying with Rule 63 of the Uniform Rules of Court) is required for agents signing on behalf of principals.

• Identification: Full names of directors, members, or partners must be provided in the bid.
• Samples: Transnet may retain samples for quality checking; unsuccessful bidders must collect their samples at their own risk and cost.

• Defaults: Failure to enter a contract, accept an order, furnish security, or comply with conditions may allow Transnet to cancel the bid and recover additional expenses from the Respondent.
• Acceptance: A successful email communication is considered proof of delivery 1 day after submission.
• Governing Law: The contract is governed by the laws of the Republic of South Africa.
• Securities: A Deed of Suretyship must be provided by an approved South African financial institution within 30 days of acceptance. Failure to do so may result in contract cancellation.
• Contract Documents: The contract comprises the General Bid Conditions, Terms and Conditions of Contract, and any Special Conditions.

22 identification

If the Respondent is a company, the full names of the directors shall be stated in the Bid. If the Respondent is a close corporation, the full names of the members shall be stated in the Bid. If the Respondent is a partnership or an individual trading under a trade name, the full names of the partners or of such individual, as the case may be, shall be furnished.

33 bids by or on behalf of foreign respondents

33.1 Bids submitted by foreign principals may be forwarded directly by the principals or by its South African representative or agent to the designated official of Transnet according to whichever officer is specified in the Bid Documents.

33.2 In the case of a representative or agent, written proof must be submitted to the effect that such representative or agent has been duly authorised to act in that capacity by the principal. Failure to submit such authorisation by the representative or agent shall disqualify the Bid.

33.3 When legally authorised to prepare and submit Bids on behalf of their principals not domiciled in the Republic of South Africa, representatives or agents must compile the Bids in the names of such principals and sign them on behalf of the latter.

33.4 South African representatives or agents of a successful foreign Respondent must when so required enter into a formal contract in the name of their principals and must sign such contract on behalf of the latter. In every such case a legal Power of Attorney from their principals must be furnished to Transnet by the South African representative or agents authorising them to enter into and sign such contract.

a) Such Power of Attorney must comply with Rule 63 (Authentication of documents executed outside the Republic for use within the Republic) of the Uniform Rules of Court: Rules regulating the conduct of the proceedings of the several provincial and local divisions of the Supreme Court of South Africa.

b) The Power of Attorney must be signed by the principal under the same title as used in the Bid Documents.

c) If a Power of Attorney held by the South African representative or agent includes matters of a general nature besides provision for the entering into and signing of a contract with Transnet, a certified copy thereof should be furnished.

d) The Power of Attorney must authorise the South African representative or agent to choose the domiciliumcitandietexecutandi.

33.5 If payment is to be made in South Africa, the foreign Supplier/Service Provider [i.e. the principal, or its South African agent or representative], must notify Transnet in writing whether, for payment by electronic funds transfer [EFT]:

a) funds are to be transferred to the credit of the foreign Supplier/Service Provider's account at a bank in South Africa, in which case the name and branch of such bank shall be furnished; or

b) funds are to be transferred to the credit of its South African agent or representative, in which case the name and branch of such bank shall be furnished.

33.6 The attention of the Respondent is directed to clause 24 above [Securities] regarding the provision of security for the fulfilment of contracts and orders and the manner and form in which such security is to be furnished.

34 database of restricted suppliers

The process of restriction is used to exclude a company/person from conducting future business with Transnet and other organs of state for a specified period. No Bid shall be awarded to a Bidder whose name (or any of its members, directors, partners or trustees) appear on the Register of Tender Defaulters kept by National Treasury, or who have been placed on National Treasury’s List of Restricted Suppliers. Transnet reserves the right to withdraw an award, or cancel a contract concluded with a Bidder should it be established, at any time, that a bidder has been restricted with National Treasury by another government institution.

• Communication: During the evaluation period, communication is restricted to the designated contact person.