Automotive manufacturing is increasingly a data intensive space.
Modern manufacturing plants can see information being created from production equipment, vehicles, sensors, inventory, tools, logistics and enterprise systems. The question is not of a lack of data. The question is connecting the right data and providing context to enable operational insight.
This is where the concepts of AIoT (Artificial Intelligence + Internet of Things) come into play.
By considering physical-world data and analytics holistically, we can see how IoT data connects to decision-making.
Starting With the Physical Side
An AIoT architecture begins with the physical manufacturing environment.
Machines, vehicles, tools, inventory and other assets can emit meaningful signals. Sensors and identification technologies can capture information on conditions, movements, location and activity.
Depending on the specific use-case, this layer can incorporate a range of technologies including:
RFID
UWB
BLE
IoT sensors
PLCs
Industrial wireless networks
Additional connected systems
The value comes from pairing the right technology to the question being asked.
For example:
RFID: What asset / item is this?
RTLS: Where is it?
IoT telemetry: What's happening with the equipment or process?
Manufacturing systems: What production context is the data part of?
The ability to combine these signals is how we begin to create meaningful information about our manufacturing operations.
The Importance of a Pipeline
Connecting sensors is a starting point, but a valuable AIoT architecture will have a way to get from sensors to systems and to insights.
A logical (but not complete) pipeline would be:
Physical Assets → Sensors → Connectivity → Data Pipeline → Analytics/AI → Decision → Action
In an automotive plant, this may require bringing information from disparate operational and enterprise systems (MES, ERP, SCADA, PLCs, etc) into a single view. Context becomes one of the most critical engineering challenges. A location event by itself may not tell us much, and a production event without asset context may be equally limited.
By combining data from different sources, we can achieve an enhanced view of operations on the factory floor.
What Role Does AI Play?
A critical realization about AI is that it is not a replacement for the underlying data infrastructure.
Sensor information that is inaccurate or incomplete, system integrations that miss key information, and the absence of critical operational context will all limit the value of analytics. AI is most valuable when it has relevant and reliable information to work with.
For manufacturing teams, this can mean pattern recognition, anomaly detection, relationship discovery or finding information that warrants closer examination. A good rule of thumb is to not think of these as AI use-cases, but rather as operational questions to answer.
The wrong question to ask about AI in automotive manufacturing is "How do I apply AI to my factory data?". The right question is "What operational problem can connected data and analytics help me understand?". That approach is vital to designing industrial AIoT systems.
Traceability Is a Data Integration Challenge
Automotive traceability is a good example of a data integration challenge. Vehicles move through different manufacturing phases and information can exist in a variety of systems. VIN-level traceability and genealogy can provide a mechanism to associate relevant information with a specific vehicle. From an engineering perspective, the ability to integrate these data points is more important than storing isolated records. The relationships between events, assets, components and production processes can unlock additional analytics opportunities.
The IT/OT Divide and Interoperability
Perhaps one of the largest challenges in industrial IoT is the IT/OT divide. Manufacturing environments can have legacy industrial systems co-existing with modern IoT, analytics and enterprise applications. Full replacement of existing systems is not always feasible or desirable.
Integration can allow these systems to work together while enabling additional data to be leveraged for wider operational insight.
This is where interfaces, industrial protocols, edge compute, data pipelines and cybersecurity come into play.
A More Practical Approach
Organizations looking to leverage AIoT can begin with the operational question. For example:
What do we need to see?
What information already exists?
What are the gaps?
How should different data sources be connected?
What decisions can the information enable?
How should the system be secured, maintained?
This enables organizations to avoid the trap of thinking about more connections = better information. OEMNex AI focuses on the intersection of AIoT, industrial connectivity, manufacturing visibility and automotive OEM operations. OEMNex AI
The Larger Context
AIoT in automotive manufacturing is ultimately about creating an effective connection between the physical factory and its digital twin.
The path can be summarized as:
Connect → Collect → Contextualize → Analyze → Decide → Act
IoT provides much of the underlying connectivity and information infrastructure. AI and analytics offer additional approaches to analyzing data. The engineering challenge (and opportunity) is creating the layers that enable information to be useful to the people running the factory. That is how we transform a set of connected devices into a truly smart manufacturing environment.
Top comments (0)