By Giles Peckham, Regional Marketing Director at Xilinx,
and Adam Taylor CEng FIET, Embedded Systems Consultant
One of the key enabling technologies of Industry 4.0 is embedded vision. For those not familiar with it, Industry 4.0 introduces automation, data collection and sharing into the manufacturing flow. Within Industry 4.0 solutions, embedded vision systems enable automation for positioning and guidance of piece parts, data collection and decision making. Of course, embedded vision is not limited to the visible element of the electromagnetic spectrum, it is also able to work with X-Ray and Infra-Red at different ends of the EM spectrum. This can provide additional information useful in the manufacturing flow. To further aid production many embedded vision systems embed machine learning inference at the edge. This allows more accurate detection and identification of not only piece part quality but also the correct assembly and positioning of objects upon the production line. Such approaches enable better yields as non-conformances are identified earlier in the manufacturing flow.
Designing a solution which can be deployed in the numbers typically needed in Industry 4.0 systems requires close consideration of the Size, Weight, Power and Cost, often called SWaP-C, as well as processing capacity and connectivity.
Within Industry 4.0 applications, embedded vision systems are typically deployed at the network edge and must be able to connect to other edge devices and the cloud via the Industrial Internet of Things (IIoT). They must be able to connect not only to the image sensor or camera but also to other modules in the system via both industrial standard interfaces and proprietary / bespoke interfaces. This allows support for illumination, actuation and positioning systems.
For Industry 4.0 applications, the IIoT converges both the Information Technology (IT) and Operational Technology (OT) networks. A converged network provides vertical integration across the layers of the automation pyramid, allowing the network to be expandable and enabling end-to-end communication from field devices like drives and sensors all the way up to workstations for the Enterprise Resource Planning (ERP) system. However, many current industry applications use separate networks, requiring gateways and bridges to connect the IT and OT networks.
This separation of networks occurs as OT networks need to support real-time deterministic communications and demonstrate a low packet delay variation, while IT networks are optimised for high bandwidth, flexible topologies and automated configuration. Unfortunately, separate networks are difficult to scale, require multiple protocols, need network engineering before installation and often have bandwidth restrictions in OT segments.
Converging the IT and OT networks addresses these issues, providing a network that is no longer strongly hierarchical nor limited in scalability or performance, along with achieving the desired vertical integration as previously identified.
One converged network solution used in Industry 4.0 applications is Time Sensitive Networking (TSN). TSN is a set of IEEE 802 sub-standards which enable deterministic communication over Ethernet networks. Embedded vision systems supporting Industry 4.0 and IIoT applications should be able to support TSN to enable ease of integration within the solution.
TSN enables different classes of network traffic to share the same link, providing both network management and a reserved path for scheduled traffic, ensuring deterministic communications. TSN therefore enables one common network to be implemented, which supports multiple communication standards. When TSN operates over an Ethernet link, there are several improvements to standard Ethernet. Standard Ethernet communications are not time aware; they distribute the data over the entire bandwidth of the link with packets queued in order for transmission. TSN implements a time awareness across the network, with scheduled traffic in time-defined slots, and supports cyclic data transmission while also providing pre-emption for higher priority packets. TSN is also better suited to manage the very high frame rates in optical monitoring processes due to its use of cyclic behaviour with high throughput.