Modern automotive factories are populated with a variety of connected and semi-connected systems.
Machines generate telemetry, RFID systems can identify components, and BLE or UWB can provide localization information. PLCs manage industrial processes and MES and ERP platforms track production and business data.
The issue is that these systems do not always tie together to provide an actionable understanding of what is happening on the factory floor.
This is where AIoT can come in: AIoT (Artificial Intelligence of Things) is a concept which combines connected physical assets with data processing, analytics, and artificial intelligence to generate insights into the physical world.
A simplified architecture could look something like the following:
Physical Assets → Sensors → Connectivity → Data → AI/Analytics → Insight → Action
The physical layer could include machines, vehicles, tools, materials, and other assets; sensors and identification technologies generate and/or transmit data and information, and connectivity transports that information for applications or processing systems.
AI and analytics then add value in another layer.
The important point is that AIoT is not just using AI on an IoT deployment: the value is in the connection between the physical-world events, and the software that can interpret it.
Different technologies can generate different data: automotive manufacturing environments can use a variety of technologies, from RFID for identification of vehicles, parts, tools, or materials, BLE for proximity and location applications, and UWB for more precise indoor positioning in appropriate environments.
GPS/GNSS can be helpful for certain outdoor or mobile assets.
Industrial sensors can provide temperature, vibrational, motion, or status information.
These technologies generate different data, but an application can become more valuable when multiple are combined: the identity of an asset can be used to find out more about its location or related production event; instead of "someone scanned a part", a system can potentially tie that event to a location event and production stage.
An AIoT architecture must also connect to other factory systems.
Automotive plants likely use a number of manufacturing related systems, including:
PLCs
SCADA
MES
ERP
Industrial gateways
Databases
APIs
Edge platforms
The goal is not to throw these systems away, but to use appropriate connections between them and to ensure that the functions already provided by these systems are maintained.
Protocols and interfaces such as MQTT, OPC UA, APIs, and gateways can be important, based on a specific architecture or equipment.
The details will matter based on the environment, legacy equipment, latency tolerances, security model, and type of data being transported.
The value of AI is that it can analyze connected manufacturing data for specific insight into operational events.
Examples could include:
Finding unusual equipment behavior
Identifying patterns in machine telemetry
Predictive maintenance
Analysis of production events
Asset utilization
Visibility into vehicle movements
Correlation of production and location data
AI is not magic, and not every data set will automatically become valuable: data quality, context, integration, and the domain knowledge all matter. Poorly connected data in an unrefined format can be a poor basis for any analytics or AI model.
Edge processing can be important for certain manufacturing events.
Not every manufacturing event requires the data to flow somewhere else before an action is taken.
Certain applications might require faster or more localized processing, and such an architecture could use an edge computing model.
A system could then potentially look something like the following:
Sensors → Edge/Gateway → Data Platform → Analytics/AI → Manufacturing Application
This model could also help separate time-sensitive, on-the-ground processing from other analytics and reporting functions.
The true industrial AIoT challenge is not connecting more things, but rather combining different data sets in a way that better understands a particular operational reality.
A vehicle location event, machine condition signal, production event, and inventory update may look straightforward on a data level, but combining these into the appropriate processing context is crucial in gaining operational insights.
To learn more about the potential of AIoT in automotive OEM environments, OEMNex AI's automotive manufacturing overview can help.
The general approach is simple: Connect → Collect → Analyze → Act
For automotive manufacturers, the engineering challenge is taking this sequence and making it a reliable and safe part of everyday operations.
Top comments (0)