From IoT Data To AIoT Intelligence: Why Context Matters
IoT systems have done a nice job with collecting information about the physical environment.
Sensors can gather all sorts of signals, from temperature to vibration and location, equipment state, environment, and other operational metrics. Connected devices carry those signals to applications where teams can inspect the data.
The challenge is, at some point a connection between all of the random bits of information needs to occur for anything interesting to happen.
That’s where the AIoT — Artificial Intelligence of Things comes in handy
IoT collects the data. The addition of AI helps to pull all of it together in a meaningful way.
IoT Architecture
Let’s take a look at a possible setup of an IoT system:
Physical Asset
Sensor
Connectivity
Data Platform
Dashboard
The architecture provides nice insights, such as a way to see how hot a particular machine is and if there are any abnormal vibrations.
But perhaps, it can go a bit further, right?
Let’s take a look at a possible AIoT architecture:
Physical Environment
Sensors / Cameras / RFID
Connectivity + Data
Contextual Data Layer
AI / Machine Learning
Insights / Predictions
Operational Action
There’s no magic bullet, but the goal is about data contextualization and making the most of the available information.
How Does It Work In Reality?
If a machine has recently started to emit stronger vibrations, what does it mean?
Well, what machine, how it was used,
what was the speed, when was the last service, were there any changes to the production process, the working environment, etc. All of that might be relevant to determine what exactly happened — and what needs to be done about it.
This is where the notion of AIoT makes a difference between simple IoT and something with more depth to it.
The information provided by a sensor is only a part of the puzzle. Other aspects contribute to an understanding of that information.
AIoT Is Not Simply an Extension of IoT
It’s easy to assume that AIoT is simply IoT + some machine learning, but there’s room for a much more complex architecture.
For example, one with the following components:
Physical sensing: Collecting information about machines, assets, the environment, or people.
Connectivity: Providing means for exchanging information between devices, applications, and systems.
Data infrastructure: Storing, managing, and processing data.
Context: Adding meaning to the data points in relation to the physical items they represent and external factors.
AI and ML: Finding patterns, connections, and other useful information.
Operational workflow: Consuming the insights and acting on them in some capacity.
This is an example of one architecture, so there’s no issue with modifying or replacing some components.
Notice how data engineers, software engineers, and people with domain expertise in the operational side of things — all of them are part of the process.
Let’s Apply It To A Use Case
For example, here’s a scenario with three data points:
Temperature Sensor ──┐
Vibration Sensor ────┼──→ Data Platform → AI Analysis
Machine Status ──────┘
The system could start with these three data points.
The AIoT architecture makes it possible to enrich the context of each data point.
It can relate them to a particular machine, its state, and any recent changes.
Here’s what the AI analysis might look like:
Machine A
Operating normally: No
Observed:
Increasing vibration
Higher-than-usual temperature
Observed in:
Load: Increasing
Recommended investigation:
Inspect machine condition and operating environment.
Nothing fancy really. Just a system that takes all of the pertinent information and makes it easier for a person or another system to consume.
It’s the addition of information and context that makes the architecture valuable in this case.
Opportunities And Limits Of This Approach
At the most basic level, it’s all about connecting various dots of information.
There’s probably a use case for this sort of architecture in most fields that involve some physical objects.
Manufacturing
It can make sense for production facilities that can use connected objects and related signals about the production process and the state of the machines.
Logistics
IoT devices provide the opportunity to collect data about the state and movement of objects.
The combination of data and the ability to process and analyze it in context could help logistics companies to better understand what’s happening in their operations.
Construction
Equipment, sensors, and other connected devices can provide data about the characteristics and state of a construction site.
The challenge is in organizing the available data and processing it into meaningful information that can be used to control costs, optimize the workflow, and ensure the safety of workers.
Infrastructure
The domain is vast, but the amount of data generated by the infrastructure systems is enormous. AIoT offers an opportunity to tie some of that data together with actionable insights.
But what about the limitations of such an approach?
It’s important to recognize that the quality of insights depends heavily on how much sense can be made of the data.
If the data is bad or incomplete, any additional processing is a waste of time.
Even worse, if no one knows how to make sense of it, the best mathematical models would not help.
It’s not about the sophistication of a single link in the chain, but about the chain as a whole.
Therefore, from a practical standpoint, the AIoT is not simply an AI problem, but a systems integration problem.
And the right question to ask is, “What data do I need, where can I get it, what context is required, and how do I intend to use the resulting data points?”
Aperture Venture Studio works at the intersection of AI and IoT technologies with a particular focus on industrial applications where these concepts form the basis for new technology ventures. Learn more about the AIoT approach in this overview by the Aperture Venture Studio: Aperture Venture Studio.
So, how do you think the AIoT can be applied in your context?
known more https://apertureventurestudio.com/
Top comments (0)