NetUnderstand the system.
Plant context and physics-informed models connect equipment, behaviour and constraints.
Explore the foundationTHE INTELLIGENCE LAYER FOR COOLING
In development & validationCooling intelligence.Clearer decisions.
We’re building AI that helps teams understand cooling systems, evaluate better operating choices and measure what changes, using the equipment they already have.
Read-only first. Operators stay in control.
THE CONNECTED COOLING PLANT · CONCEPTUAL VIEW
Cyan: cold supply · Amber: warm return
01 Understand the whole plant02 Review the operating choice03 Make the evidence visible
THE CHALLENGE
Your equipment is connected.
Its decisions should be, too.
Cooling is a system. But the information needed to run it often lives in separate places.
- 01Data without the full picture.Signals need equipment context, operating limits and quality checks.
- 02Trade-offs across the plant.A change at one chiller can affect pumps, towers and service elsewhere.
- 03Improvement needs evidence.Lower electricity use means little if the required cooling is not delivered.
FROM SIGNAL TO UNDERSTANDING
A clearer path from
plant data to decisions.
One connected workflow. Five jobs. Built around the people responsible for the plant.
01 / CONNECT & CONTEXT
Give every signal a place.
Bring equipment, meter readings, drawings and operating limits into one plant model. Identify what is known, what is missing and which data can support a decision.
A clearer picture of your plant.
PLANT CONTEXT
Equipment & connectionsMap
Meter readings & unitsCheck
Operating constraintsRecord
Illustrative workflow · In development
02 / DIAGNOSE
Understand the whole system.
Use plant context and physics-informed models to investigate behaviour across chillers, pumps and heat rejection. Separate a useful observation from an unreliable reading.
An observation with its evidence and uncertainty.
SYSTEM DIAGNOSIS
Plant behaviourObserve
Physical relationshipsEvaluate
Data confidenceExplain
Illustrative workflow · In development
03 / RECOMMEND
Put a better choice on the table.
TARA is the proposed operator-facing workflow: show the choice, the trade-offs and the constraints. Any future command must pass explicit policy and site permission checks.
A recommendation an operator can review.
DECISION WORKFLOW
TARA proposalExplain
Site limits & permissionsCheck
Authorised operatorReview
Illustrative workflow · In development
04 / VERIFY
Ask what actually changed.
Compare outcomes against an agreed baseline while accounting for load and operating conditions. Energy improvement only counts alongside the service the plant must maintain.
A traceable assessment, with its limitations.
OUTCOME EVIDENCE
Agreed comparisonDefine
Energy & serviceMeasure
Assumptions & uncertaintyDisclose
Illustrative workflow · In development
05 / LEARN
Carry the learning forward.
The long-term aim is to improve models and reduce repeated engineering effort. Reuse across sites depends on explicit data rights, suitable evidence and fresh validation.
Better future models, subject to evidence.
DEVELOPMENT GOAL
Data & training rightsAgree
Context, action, outcomeEvaluate
Unseen-site performanceValidate
Illustrative workflow · In development
Make it useful.
The proposed operator workflow turns observations into explanations and reviewable choices.
Meet the workflowShow what changed.
Trace a result to its inputs, assumptions, operating conditions and agreed comparison.
See the principlesFOCUSED FIRST. EXPAND WITH EVIDENCE.
A shared ambition.
A qualified scope.
Every application has a service to protect. New domains need their own validation.
Buildings &
central plants.
Water-cooled central chilled-water plants in noncritical offices, hotels and campuses.
Protect comfort, humidity and occupied-hour service.
Explore plant fitManufacturing &
process cooling.
Food, beverage and pharmaceutical cooling utilities, subject to process-specific assessment.
Protect temperature, throughput and product quality.
View the directionData centres &
critical loads.
A separately qualified scope for demanding cooling and reliability requirements.
Protect thermal limits, redundancy and uptime.
View the directionTRUST IS A DESIGN REQUIREMENT
Intelligence above.
Authority at the plant.
Our intended architecture keeps AI proposals separate from permission to act. Local controls remain authoritative.
Explore safety & evidenceStart by observing.
Qualify data and understand the plant before considering an operating change. Read-only access is the starting point.
Respect explicit boundaries.
Future commands require site-approved limits, valid inputs and commissioned interfaces. Existing BAS, PLC and equipment protections stay in place.
Make decisions traceable.
Record assumptions, permissions, applied states and outcomes. Show uncertainty rather than hiding it behind a confident answer.
THE DEVELOPMENT PATH
Each capability
has to earn its place.
A staged programme, with evidence gates before each expansion in scope.
Qualify & connect.
Map one plant archetype. Agree data rights. Establish measurement quality and useful questions.
FOUNDATION DEVELOPMENTForecast & recommend.
Test models against relevant baselines. Make observations and choices reviewable.
VALIDATION REQUIREDSupervise within limits.
Commission selected actions only after permission, safety and operating gates pass.
CONDITIONAL CAPABILITYLearn & transfer.
Evaluate reuse across different plants and qualify new applications on their own evidence.
LONG-TERM GOALWhat does Net0link do?
We are building a cooling-intelligence layer that connects plant data with engineering context. The goal is to help teams understand system behaviour, evaluate operating choices and verify outcomes through TARA workflows.
Does it replace our BMS or equipment?
The intended architecture works above existing building management systems (BMS), programmable logic controllers (PLC) and equipment controls. Site compatibility and data access must be assessed; existing protective logic remains authoritative.
Will AI control the plant on its own?
The development approach starts read-only. Recommendations are a separate capability. Any future supervised action requires site-approved limits, commissioning and a clear authorisation path. A language model does not directly command equipment.
Which sites are the first focus?
Noncritical buildings with water-cooled central chilled-water plants are the first validation domain. Offices, hotels and campuses may be candidates depending on their plant configuration, metering, documentation and operator participation.
How will savings be verified?
A site-specific evaluation must agree the measurement boundary, adjusted baseline, uncertainty and required service quality. This website does not present a verified Cooling Foundation savings percentage or guarantee.
What about manufacturing and data centres?
Food, beverage and pharmaceutical process cooling are future areas requiring separate qualification. Data centres are a later roadmap domain, with additional thermal, redundancy and uptime requirements.
Can plant data be used to train shared models?
Access to operational data does not automatically grant training rights. Permitted use, confidentiality and any shared learning must be explicitly agreed.
How can we get involved?
Start with a plant-fit conversation. Tell us the type of plant, the information you have and the operating question you want to investigate. The Cooling Foundation is in development and validation; a conversation does not commit you to deployment.
BUILD WITH US
The next chapter of cooling
starts with a real plant.
Help shape the intelligence your operations actually need. Let’s start with your system, your constraints and the decision you want to improve.