Building an Industrial IoT Data Layer for Automotive Manufacturing

Building an Industrial IoT Data Layer for Automotive Manufacturing

A modern factory for automotive production can be considered a highly distributed computing system. A number of systems operate simultaneously: PLCs, robotic cells, RFID readers, RTLS anchors, BLE gateways, AGVs, cameras, machine-vision equipment, mes applications, erp platforms, edge computers and many others. In this context the challenge is not generating the data in the factory, but building a manufacturing data layer from the individual sensor readings and the many different machine states.

The starting point is the Event Model

A robust industrial IoT architecture must focus on processing production events as structured pieces of data rather than individual sensor values. An event will carry at a minimum: the device or asset generating the data, a time stamp, the location, the type of event, the production context (for example, a specific job), and its status. An RFID tag, once related to the correct material, production zone and stage in the production process, suddenly becomes a much more valuable item of data.

This should extend to RTLS events.

Raw X-Y coordinate location data is less useful on its own than location information related to a specific AGV, production order or sequence.

MQTT will bring your data to services

MQTT (Message Queuing Telemetry Transport) is already used across IoT architecture; it's essentially a publish / subscribe service which does not require a many-to-many relationship between each industrial asset/device and each manufacturing service that requires its data. In an automotive plant a device can publish its data to a central MQTT broker and different manufacturing services can subscribe only to the relevant topics. The diagram above illustrates this:

Industrial Devices -> Edge Gateway -> MQTT Broker -> Services -> Applications

This dramatically reduces redundant point-to-point integration points and future-proofs the architecture for new consuming services to added into the system.

OPC UA complements MQTT.

MQTT is primarily a transport protocol, but for the connectivity across a whole factory an industrial-protocol based protocol may also be required especially from existing PLCs, SCADA systems, robot controllers. The Opc UA standard already is very successful at delivering a rich structured object-oriented view into a machine. Both OPC UA (via Opc UA gateways for example) and MQTT can and should operate side-by-side for full enterprise-level interoperability in a manufacturing environment; attempting to shoe-horn every piece of an industrial plant into a single, non-industrial, IoT standard simply does not work.

The role of edge computing in the plant

Sending all sensor data to a central cloud or server location isn't always a suitable strategy. An edge tier can add significant value by processing some data before sending it onwards. The kinds of jobs the edge layer can undertake include, but are not limited to; local aggregation/filtering of data, anomaly detection, temporal buffering, local AI, and correlation of local events. This is particularly beneficial where network connectivity is patchy, where very fast local reaction times are critical, and/or the sheer volume of raw event data needs pruning before it is sent onwards.

Correlation is King!

One of the most challenging engineering problems in a connected factory environment is correlation. An RFID reader will tag a part, RTLS will say the car is at a particular spot, a sequencing machine will report it has changed assembly. A well designed data layer will know these are events referring to the same production order and so link the information into one meaningful event for analysis or action.

Don't forget about device operations

Thousands of devices operating on a factory floor require maintenance. Any "data layer" for an industrial IOT system must consider its operational aspect; how will you provision devices, authenticate them, how will you update firmware, diagnose remote gateways and printers, administer readers, and monitor network connectivity to ensure your IoT deployment can function reliably as a production system.

Security first at every step

Designing any IOT system needs to consider security as it is often a secondary concern and "retro-fit" to systems originally designed for other purposes. The relevant considerations include: Network segregation, Device authentication, Identity management, Access controls to the system, secure data transmissions and protected firmware updates as well as monitoring of industrial communications and data transfer between the operational and enterprise worlds.

The Model Architecture for an industrial IoT Factory Data Layer

Putting all this together and in the context of an OEM car manufacturing factory the architecture may consist of the following layers:

Physical: The real-world devices such as, but not limited to, sensors, PLC/SCADA systems, robots, RFID, RTLS, BLE, AGVs. Connectivity: This layer encompasses all the communications protocols used: Ethernet/IP, Modbus, Profinet, etc at the automation level, OPC UA across the industrial hierarchy and potentially MQTT for event publication across the whole plant (from edge to servers). Edge: IoT gateways and/or mini computers are deployed throughout the plant performing a number of functions: telemetry filtering, aggregation and formatting of sensor/machine data, temporary storage (buffering) of data where network may be an issue and basic analysis or AI functions run on data locally.

Integration: This layer pulls together all the data for onward processing by the rest of the manufacturing and enterprise systems (MES, ERP, WMS, SCADA).

All the integration uses different protocols and APIs such as industry middleware and a variety of data exchange formats. Analytics: Production specific analytics applications that could analyze real-time data, predict future production failure, and provide immediate actionability for plant operators. Application: The end-user experience applications of all the analyzed data-including inventory, traceability, staffing requirements, dashboarding etc.

Of course in any actual factory configuration the detailed layering will vary on factory build, connectivity options available and manufacturing philosophy.

The essential engineering design lesson here.

Industrial IoT is more than simply connecting up more bits of hardware; it is about building efficient and resilient pipelines between physical reality and manufacturing decisions based on well structured, reliable and available data. Interoperability is essential, the data itself needs to be structured as meaningful events from machines as much as possible, edge devices bring many critical functions and the whole system needs management of device lifecycle and comprehensive security to support the productivity goals for the plant.

OEMNex AI's Industrial IOT Software Overview for OEM Vehicles provides OEMs an effective reference on what it takes and indeed what areas this new architecture touches for each specific factory in terms of relevant manufacturing domains. The final factory-floor-level-goal to this exercise is ensuring each factory's data can be readily understood, governed, and acted upon while ensuring the overall event generation simply increases proportionally for a gain in business metrics. For more info visit: oemnexai.com