These days construction sites are connected environments.
People carry connected things, equipment generates telematics data, materials can be RFID or BLE tagged, access control systems generate event logs, and even work teams use BIM, scheduling, ERP and other project-management tools.
As interesting as it is to collect what is happening on a project, the real challenge is connecting the heterogeneous physical world into a useful operational tool.
This is where AIoT, or AI plus IoT, can provide a fascinating architecture for commercial construction environments.
The basic architecture
A construction AIoT platform can be broken down into several layers:
Physical Jobsite
↓
Sensors & Devices
↓
Connectivity
↓
Edge / Data Processing
↓
Data Integration
↓
Analytics & AI
↓
Applications & Decision Support
Each layer addresses a specific set of needs.
1. Physical jobsite
The physical environment consists of workers, equipment, materials, work areas, access points, and installed components.
This differs from a factory environment in that the configuration changes constantly.
As such, construction lends itself to being an interesting IoT environment, as devices and assets are used in a changing and temporary infrastructure environment.
2. Sensors and identification technologies
There is no one-size-fits-all device or technology solution.
Depending on the application, a variety of RFID, BLE, UWB, GPS, Telematics, environmental sensors, and barcodes may be in use.
The key is to match an application's requirements for precision, environment, range, battery life cost, and type of information.
Connectivity matters
Once information has been collected, there needs to be a way to transport it somewhere.
Depending on the application, a construction environment might use cellular, WiFi, BLE gateways, LoRaWAN, or other means of communication.
For instance, a battery-powered sensor on a material asset may have very different requirements than a machine reporting telematics information.
As such, the architecture should be driven by the application requirements rather than the other way around.
Edge processing can reduce unnecessary data movement
Construction environments generate a lot of event data, and rather than sending everything that happens to an event-hub type of structure, edge processing can be used for some applications.
For instance:
Sensor
↓
Edge Gateway
↓
Filter / Normalize
↓
Relevant Event
↓
Cloud Platform
An edge layer can dramatically reduce bandwidth needs and processing requirements, while also providing some level of processing and decision-making capabilities on a potentially intermittent connection.
This can be particularly useful in large or changing construction environments.
The integration problem
Perhaps the biggest issue with construction AIoT is not the sensor.
It's the integration.
A commercial project may well have multiple systems already in place to manage various aspects of the project: BIM, scheduling, project management, ERP, procurement, equipment management, workforce management, quality control and document management.
If the information coming from IoT devices is not integrated into these systems, it is of very little use.
A better architecture would tie the events coming from the physical environment into the systems already used to manage the project.
For instance:
Equipment telemetry
+
Workforce activity
+
Material events
+
BIM information
+
Schedule data
↓
Integrated project data
↓
Analytics
With this richer model of the project, opportunities for valuable analytics emerge.
From events to context
Events are not necessarily information.
Let's say the system ingests the following event:
Equipment ID: EX-104
Location: Zone B
Status: Idle
Duration: 47 minutes
All the system knows is that something happened.
To extract value from the event, the system needs to understand the context.
Was the equipment supposed to be idle?
Is the crew waiting for materials?
Is the work area restricted?
Is there some other reason the equipment is waiting?
Is the equipment needed elsewhere?
This is where integration becomes critically important.
The more relevant information about the project that the system has, the more useful the analytics can be.
Where AI fits
AI can be used to help analyze the integrated data, spot patterns, and build correlations between events.
Some possible applications for AI in construction AIoT:
• Anomaly detection
• Equipment utilization analysis
• Progress analysis
• Predictive maintenance
• Resource optimization
• Schedule-risk analysis
• Pattern detection across jobsite events
That is not to say that AI is a substitute for good data engineering practices.
A good rule of thumb is:
Better AI starts with better data.
If timestamps are out of sync, asset IDs are wrong, location data is incorrect, and systems cannot be matched, an interesting model will still give questionable results.
Example: investigating a delayed work package
There is a mechanical installation package that is delayed.
In a traditional setup, the PM might have to consult various sources of information to determine the root cause.
Using an AIoT architecture, information could be integrated as follows:
Schedule
+
Worker presence
+
Equipment utilization
+
Material availability
+
Inspection status
+
Work-area location
Analytics could be used to find correlations worth pursuing.
In this example, the AI would determine that the materials for this work package arrived late, which in turn resulted in a lower presence of workers and delayed installation activities.
The AI might not have to actually take any action other than flagging this set of circumstances as worthy of further investigation.
In some cases, that might be sufficient.
The real value of such an AI might be in reducing the time needed to find relevant information.
Construction requires a different approach to IoT
One of the main differences between industrial IoT and construction IoT is the fact that a construction site is a changing environment.
Walls go up, floors become accessible, temporary structures are built and taken down, equipment is moved around, contractors come and go, work zones are redefined, and so on.
As such, the architecture must take this into account.
Ideally, location data should be contextual rather than absolute.
BIM, and similar spatial information, can be particularly useful in creating this kind of context.
Top comments (0)