Closed loop refinery safety

Detect. Decide.
Act. In Under 500ms.

SPNG.ai turns existing CCTV into machine to machine safety intelligence. Cameras detect danger, site approved rules decide, and machines receive the right action through PLC, SCADA, or ESD controls.

See the safety loop
UNIT 04 · CONTROL LIVE12:48:09
427MS CLOSED LOOP

Machine stop acknowledged

Scroll to explore

Built for the systems already inside your refinery

EXISTING CCTVEDGE INFERENCEOISD RULESPLCSCADAESDOFFLINE MODEEXISTING CCTVEDGE INFERENCEOISD RULESPLCSCADAESDOFFLINE MODE

DETECTION IS NOT ENOUGH

An alert cannot stop
A Running Machine.

The danger continues while people read, call, and walk.

01

Most camera analytics end at a notification. The manual response that follows can take 5 to 15 minutes.

01

Alert arrivesA safety officer reviews the event.

02

Calls beginA supervisor is asked to reach the equipment.

03

Manual stopThe machine keeps running until someone acts.

Camera Event
Alert Inbox
Safety Officer
Supervisor
CONVENTIONAL AIAlert
OnlyDetect · Notify · Wait
Machine still runningWorker inside hazard zoneCritical risk · Human response required

THE CLOSED LOOP

One safety loop.
Less Than 500 Milliseconds.

Move from camera signal to approved control action without waiting for a manual call chain.

01
DetectUnder 100 ms

Edge models analyse existing CCTV for PPE, zone, fire, smoke, proximity, fall, and behaviour risks.

02
DecideOISD mapped

A site approved rule checks the event, location, confidence, equipment state, and required response.

03
ActUnder 500 ms

SPNG.ai alerts, requests confirmation, or sends a commissioned command through PLC, SCADA, or ESD.

01Camera feed
02Hazard found
03Rule matched
04Stop issued
Machine safe

CONTROL POLICY

Your plant decides how SPNG.ai responds.

Choose the action for every rule, zone, machine, shift, and operating state.

Alert onlyRecord the event and notify the right team.

Confirmed actionAsk an operator to approve the control command.

Automatic actionIssue a preapproved safe stop for a defined hazard.

Emergency pathTrigger existing ESD logic through a commissioned interface.

Plant controlled by design. Automatic actions require engineering review, fail safe testing, commissioning, and site sign off. SPNG.ai works with existing safety systems and does not replace a certified SIS or ESD.

REFINERY SAFETY MODELS

See the hazard.
Know The Required Action.

Purpose built detections connect directly to site rules and control policy.

01

PPE Detection

Check helmets, eyewear, gloves, and required workwear.

Live detection
02

Restricted Zones

Detect entry into exclusion and operating zones.

Site mapped
03

Fire and Smoke

Identify visible warning signs at the camera edge.

Early warning
04

Unsafe Behaviour

Flag unsafe movement and work practices near equipment.

Context aware
05

Vehicle Proximity

Measure dangerous separation between people and vehicles.

Distance aware
06

Fall and Man Down

Escalate a possible fall or immobile worker immediately.

Rapid response
07

OISD Rule Engine

Evaluate each event against site approved safety rules.

40 plus rules
08

Machine Control

Send approved commands through PLC, SCADA, or ESD paths.

Closed loop
09

Incident Evidence

Keep the clip, rule, decision, and action in one record.

Fully logged

MACHINE TO MACHINE INTELLIGENCE

From visual signal
to Machine Action.

Local detection, deterministic rules, and commissioned control interfaces let machines respond to machines as one system.

SITE INPUTS
Existing CCTV
Site Rules
Machine State
EDGE DECISION
Jetson EdgeUnder 100 ms detection
OISD RulesSite approved decision
SPNG.aiClosed Loop
SafetyDetect · Decide · Act
CONTROL ACTIONS
Alert Team
Stop Machine
Trigger ESD
Under 100 ms detection Under 500 ms loop Offline capable Every action logged

No direct bypass of plant safety logic. SPNG.ai sends commands through interfaces approved by plant engineering, with acknowledgement, timeout, and audit records for every action.

THE PRODUCT IN ACTION

See every decision.
See Every Control Action.

One interface shows the detection, matched rule, selected response, and machine acknowledgement.

SPNG Control · Refinery Unit 04 LIVE
CONTROL CAPABILITYVisual Risk Detection
Detection84 ms Edge inference
DecisionRule matched OISD mapped
Action427 ms Acknowledged
CLOSED LOOP LATENCYLIVE
00:0006:0012:0018:00NOW
CONTROL DECISIONThe hazard is detected locally.

Run refinery safety models on existing CCTV streams at the edge.

Process UnitArmed
Loading BayArmed
Tank FarmReview
PerimeterArmed

SPNG CONTROL COMMAND

Every camera.
One Control Loop.

See the detection, decision, action, and acknowledgement in one live view.

Control Command
All systems live
REFINERY UNIT 04Live control posture
Closed loop latency427 msWithin target
Detection latency84 msAt the edge
Protected cameras10All online
Active safety rules42OISD mapped
Response latencyLive · LIVE LOOP
00:0006:0012:0018:00Now
Control readiness12 ZONES
92%Armed
9 armed2 review1 manual
Live safety loop LIVE
Camera
Detect
Decide
Control
Safe
Recent control eventsAUDITED

Restricted zone breachLoading Rack 03 · 12 seconds ago

Missing face shieldProcess Unit 02 · 3 minutes ago

01Under 100 ms detection

02Under 500 ms action

0340 plus mapped rules

04Complete control audit

WHAT IS READY TODAY

Start with one unit.
Prove The Full Loop.

A focused pilot validates detection quality, rule logic, and safe machine response before plant expansion.

01

MVP COMPLETE

Detection engine live

PPE, restricted zone, fire, and smoke models run on existing CCTV feeds.
Existing camera ready
02

EDGE ARCHITECTURE

Fast local inference

NVIDIA Jetson processing keeps selected detections under 100 milliseconds and supports offline operation.
Data stays on site
03

OISD RULE ENGINE

Controls already mapped

More than 40 safety alerts are mapped to relevant controls in OISD standards 116, 118, and 145.
Site review required
04

PILOT READY

Unit pilot design

The hardware, deployment plan, rule calibration, and control integration sequence are ready for a critical unit.
Designed for refineries
PILOT PROGRAM

A clear path from site survey to commissioned action.

01Week 1Survey

Map cameras, zones, machines, and control paths.

02Week 2Connect

Deploy edge hardware and validate video quality.

03Week 3Calibrate

Tune models and review each safety rule.

04Week 4Prove

Test alerts, commands, acknowledgements, and audit logs.

START WITH REFINERIES

Built where people and
Machines Share Risk.

Refineries are the first focus, followed by other high consequence operations.

01

Refineries

View safety use cases

02

Petrochemical

View safety use cases

03

Mining

View safety use cases

04

Chemicals

View safety use cases

05

Heavy Manufacturing

View safety use cases

06

Energy and Utilities

View safety use cases

WHAT SPNG.AI CAN SEE

Visual context for the risks
Sensors Miss.

Each detection is tied to a rule, a response level, and a complete evidence record.

01

PPE and Workwear

Verify site required protection continuously.

02

Restricted Zone Protection

Protect mapped exclusion zones in real time.

03

Fire and Smoke

Surface visible warning signs at the edge.

04

Unsafe Behaviour

Flag unsafe work practices near active equipment.

05

Line of Fire

Detect people entering hazardous equipment paths.

06

Fall and Man Down

Escalate a potential fall without delay.

07

Vehicle and Pedestrian

Reduce dangerous proximity around mobile equipment.

08

Gas Alarm Correlation

Pair gas detector alarms with immediate visual context.

09

Safe Work Verification

Verify that required controls are present before work.

THE DECISION LAYER

Vision understands risk.
Rules Choose The Action.

Deterministic site policy sits between AI detection and machine control.

01Machine to Machine Intelligence02Computer Vision03Edge AI04OISD Rule Engine05Risk Scoring06PLC Integration07SCADA Integration08ESD Workflows09Event Audit Log10Offline Runtime
Review control security
SPNG.aiOISD Decision Engine
Machine to Machine Intelligence
Computer Vision
Edge AI
OISD Rule Engine
Risk Scoring
PLC Integration
SCADA Integration
ESD Workflows
Rule matched · Action issued in 427 ms

CCTV AND CONTROL INTEGRATION

Connect the systems
You Already Run.

Keep the camera network. Add an edge decision layer. Integrate only through approved plant interfaces.

CCTVONVIFRTSPVMSNVIDIA JETSONPLCSCADADCSESDGAS DETECTORSCCTVONVIFRTSPVMSNVIDIA JETSONPLCSCADADCSESDGAS DETECTORS
SCADADCSESDGAS DETECTORSACCESS CONTROLEHSEDGE GATEWAYREST APIsWEBHOOKSPRIVATE CLOUDON PREMISESSCADADCSESDGAS DETECTORSACCESS CONTROLEHSEDGE GATEWAYREST APIsWEBHOOKSPRIVATE CLOUDON PREMISES
Keep existing cameras Preserve OT segregation Commission every action

WHY THE RESPONSE GAP MATTERS

Faster detection only matters when the right action follows just as quickly.

010+

industrial fatalities reported in India over five years

020 to 15 min

human response window after a conventional alert

03$0.0B

chemical incident losses tracked from 2020 to 2025

04< 0 ms

edge detection target on selected refinery models

05< 0 ms

engineered target for the complete closed loop response

060+

safety alerts mapped to relevant OISD controls

Industrial figures reflect the SPNG.ai research period from 2020 to 2025. System latency varies with camera, model, network, control path, and site configuration.

THE LOOP IN PRACTICE

A response completes before
The Phone Chain Begins.

Every stage is timed, explainable, and available for review.

Risk detectedWorker 04 · 84 ms
Stop acknowledgedPLC 02 · 427 ms
01 / 04

Detect

Edge AI identifies the hazard on the existing camera feed.

SAFE CONTROL BY DESIGN

Control authority stays
With Your Plant.

SPNG.ai operates inside approved boundaries with segregation, permissions, and complete logs.

Read only camera ingestOT/IT segregationEdge processingSigned control commandsConfigurable control policyOn premises deploymentRole based authorityImmutable action logs
Review control architecture
CONTROL SECUREAuthority verified
Edge isolated Command signed Action audited

A CONTROLLED ROLLOUT

Pilot first.
Commission With Evidence.

Automatic action is enabled only after site testing and formal approval.

01

Site Survey

Map cameras, hazards, machines, and approved control paths.

02

Rule Validation

Review every model threshold and response with the safety team.

03

Control Integration

Connect PLC, SCADA, or ESD through plant approved interfaces.

04

Commissioning

Prove fail safe behaviour, acknowledgements, and audit records.

FOR SAFETY AND CONTROL TEAMS

Review the system
Before The Pilot.

Clear material for EHS, instrumentation, operations, IT, and plant leadership.

SPNG / 01
TECHNICAL NOTE

OISD mapped rule architecture

Control guide

SPNG / 02
PILOT GUIDE

Plan a refinery pilot

Deployment guide

SPNG / 03
ENGINEERING

PLC and SCADA action interfaces

Architecture note

SPNG / 04
DOCUMENTATION

Connect an existing CCTV stream

Setup guide

SPNG / 05
VIDEO

Inside the closed loop safety system

6 min watch

SPNG / 06
RESEARCH

Closing the detection to action gap

Safety brief

BUILT ACROSS PLANT AND SOFTWARE

Refinery knowledge meets
Real Time AI Engineering.

The founding team combines process safety context with computer vision and industrial control architecture.

MR
CEO and Co Founder

Meghna Ramaratnam

Mechanical Engineering, BITS Pilani

Meghna brings refinery operations, process safety, equipment, OISD, and PSU procurement into the product from the first rule review.

meghna@spng.in
DJ
CTO and Co Founder

Dhruv Jha

Computer Science, IIT Jodhpur

Dhruv builds the computer vision, edge runtime, real time decision engine, and PLC and SCADA integration architecture.

dhruv@spng.in

One shared goal. Build the safety system that closes the gap between seeing danger and stopping it.

CLOSE THE RESPONSE GAP

Do not stop at the alert.
Stop The Hazard.

Start with a focused refinery pilot. Prove detection, decision, and machine response inside one critical unit.

Talk to the founders