Despite water covering 70% of the Earth, only 1% is drinkable freshwater. While one in ten people around the world don’t have access to safe drinking water, roughly 6 billion gallons of freshwater is lost per day leaking from pipes in the U.S. alone. Leak detection can provide early warning and the location of leaks so that water loss can be addressed immediately. Predicting leaks with AI is even better.
IoT + AI Blueprint:
Assemble the collection of components displayed below to address the use case of reducing freshwater leakage.
Sensor(s)
Pressure: Detect Excessive or Diminished Water Pressure in Pipes
Flow: Detect Water Flow Rate through Pipes to Diagnose Pump Failures
Moisture: Detect Water Leaking where it shouldn’t be
Single Board Computer (SBC): Raspberry Pi, NVIDIA Jetson, Orange Pi (If AC Power Available)
Power Options
Lithium-Ion Polymer (LiPo) Batteries
Solar
Power Over Ethernet (PoE)
AC Power (Utility | Mains | Wall Outlet)
Network(s)
Ethernet: If Present Within a Given Space for High Bandwidth Scenarios
Wi-Fi: If Available for Indoor, High Bandwidth Scenarios
Bluetooth: For Short Range, Low Bandwidth Scenarios
LoRa: Create Low Bandwidth Coverage Needed to Reach Internet Access
Cellular: If Coverage Available and Cost-Effective for High Bandwidth Scenarios
Satellite: When No Terrestrial Coverage Options are Available for Medium Bandwidth Scenarios
Digital Twin Modeling
Fresh Water Pipes and Pipeline Networks
Data Processing + Storage Location(s)
Edge: Near Water Pipes and Pipeline Systems
Cloud: Filtered Data Relayed from Edge to Monitor Broader Areas of Water Pipes
Streaming Analytics
Compare Sensor Data Values to Defined Setpoints, KPI Value Ranges and Thresholds
Filter out Duplicate Sensor Data Value Readings
Automation
Green KPI: No Action
Yellow KPI: Warn Notification for Awareness When Sensor Values are Trending Towards Leakage Conditions
Red KPI: Alert Notification to Appropriate Personnel to take Action or Automate the Closing of Water Valves When Leaks Detected
People
Deploy and Maintain Solution
SMEs Define KPIs and Actions
Building Maintenance Personnel, Plumbing, and Water Utility Organizations
Security
Uniquely Identify Each Device
TLS 1.3 Encryption for Data in Transit
Encrypt Data at Rest
Validate Device Messages to Ensure they use Expected Data Format
Rotate Security Tokens
Limit IP Address Ranges
AI Anomaly Detection
Zero Trust: Reauthenticate Device Messages Through Every Step of the System
Artificial Intelligence
Machine Learning Time Series Forecasting Model to Predict Future Occurrence of Leak Conditions
IoT Platform
Device SDK Captures Sensor Data from Physical Twin and Securely Sends it as a JSON Payload to the IoT Platform Along with a Unique Identifier and Security Token
IoT Platform Captures, Authenticates, and Saves Incoming Device + Sensor Data to a Message Queue
Background Process Takes Queued Data and Hydrates the Digital Twin by Saving it to a Database Table that Mimics the Structure of the Physical Twin
AI Model is Trained and Retrained with Digital Twin Dataset of Current and Historical Data Where Properties of Twin are Mapped to ML Features
Hot Path Data is Sent to Streaming Analytics to Facilitate Real Time Alerting and Automation
Same Hot Path Data is Also Sent to AI Model for Inferencing to Predict Future Occurrences
Rob is a writer, teacher, speaker, world traveller and undersea explorer. He's also a thought leader in the areas of enterprise mobility and the Internet of Things.
