Embedded Cloud powered by SUSiEtec: The IoT software framework SUSiEtec manages the data generated from its creation to its storage for long-term historical analyses in the cloud. SUSiEtec is the link between end devices, edge devices, control computers and clouds. © Kontron

In the case of remote production machines, the next time there is reliable contact between the edge computer and the cloud, the new data is exchanged between the real machine on the production line and the digital twin in the cloud and flows into the machine control system or into the cloud as new evaluations. Container technology makes it possible to exchange not only data, but also entire processes such as data pre-processing, by moving them from the cloud to the edge computer. Despite this, only small amounts of data are exchanged and latency times are significantly reduced. Another trend is the hybrid cloud, i.e. a local, on-premises cloud for time-critical and sensitive data and an off-premises/public cloud for downstream big data evaluations.

While increasingly powerful hardware is also becoming a matter of course in embedded computing, it is supported by the software required for edge computing. For pre-processing, filtering and containerization between cloud and machine, Kontron is expanding its portfolio of standard, modified standard and user-specific hardware to include software components based on the SUSiEtec IoT software framework for cloud connectivity.

More innovations thanks to new standards

Just a few years ago, it was unthinkable that the industry would one day be discussing whether standard IT Ethernet would also be suitable for deterministic machine communication. Thanks to corresponding standardization efforts such as Open Platform Communications Unified Architecture (IEC62541 OPC UA) and technical developments such as Time Sensitive Networking (IEEE 802.1 TSN), the rigid boundaries between OT and IT are being harmonized and made more permeable. In future, machine actuators and sensors can be connected directly to the edge computer and there will be continuous, deterministic communication right through to IT via TSN and OPC UA. Thanks to standardization, the user data can then be processed directly in the cloud.

The KBox A-15x family is available with CPUs of different performance classes. The cost-optimized Kontron Box PCs have a particularly compact design for flexible use on the DIN rail. The wide range of interfaces enables high interoperability and scalability. Industrial PCs of the KBox A-15x family are particularly suitable for use in fieldbus environments, for process control as well as for industrial firewalls and many embedded applications. © Kontron

Edge computers currently have to meet several requirements:

• Today, they must be the gateway that ensures a secure connection between the technologies at different levels of the automation pyramid, operating technology and information technology.
• They must be prepared to ensure communication in a cloud, whether embedded (on-premises/private) or public (from an external provider), today and even more so in the future.
• They must also have sufficient performance and storage capacity for current and future edge architectures to complete a wide range of tasks quickly on site at the machine. They not only take over tasks previously performed by the PLC (programmable logic controller), but can also be used to control several independent machines and processes through the virtualization of control computers.
• They must be prepared for current and future industry standards in order to integrate seamlessly into existing and future IT and OT infrastructures.

TSN-capable via plug-and-play with TSN network card

Kontron was one of the first suppliers to launch ready-to-use products for Time Sensitive Networking (TSN) with the OPC-UA (Open Platform Communications Unified Architecture) standard. A standard network card for TSN brings any expandable PC into a TSN network in which data packets are delivered in a time-sensitive and deterministic manner. Industrial computers in which several network cards can be inserted can therefore already be used for machine control. Expandable Kontron industrial computers pre-configured with the appropriate software can optionally be supplied with several TSN interfaces.

Kontron also offers a Computer-on-Module (COM) based on the SMARC 2.0 standard and an edge computer with the NXP Layerscape LS1028 multi-core processor with integrated TSN switch and up to five TSN-capable 1 Gb/s Ethernet ports for the implementation of TSN connections.

OPC UA and TSN conquer the field level

The KBox A-230-LS with the NXP Dual Cortex A72 LS1028 processor offers up to five TSN-capable 1GB Ethernet ports for applications in the IoT environment. © Kontron

For edge computers, this means that their task as a gateway is becoming increasingly less demanding, as TSN relies on standard computer and internet technology. The "intelligence at the edge of the network" can therefore be used for other, more complex tasks in machine control on site. Another task of the intelligent edge computer will be to pre-filter data so that only data required in the cloud reaches it, for example for evaluation in management systems such as business intelligence or enterprise resource planning applications. For these tasks, S&T Technologies, a sister company of Kontron, offers the IoT software framework SUSiEtec, in which these tasks can be realized in customer-specific projects.

AI applications at the intelligent edge

For special AI tasks, such as machine learning or image processing, edge computers must be expandable, for example with GPGPU (General Purpose Computation on Graphics Processing Unit), FPGA (Field Programmable Gate Array) or chips such as the Intel Movidius Myriad X VPU neural compute engine.

KBox C-102 is the high-end industrial computer family for control cabinet installation. The KBox C-102 family is thus available in four variants: from the compact KBox C-102-0 without expansion slots, through the KBox C 102 1 and KBox C-102-2, each with one or two PCIe slots, to the KBox C-102-4 with four PCIe slots. The industrial computers in the series are used, among other things, for controlling machines or visualizing and inspecting the production process. © Kontron

Neural network processors such as the Intel Movidius Vision Processing Unit (VPU) can be used for edge AI applications such as deep learning. They relieve the CPU and accelerate the inference process many times over. Intel Movidius VPUs are implemented by Kontron either via plug-in modules or as part of projects on carrier boards or SBCs.

S&T Technologies has developed an AI plugin for the SUSiEtec software framework specifically for image recognition using Edge AI and has created a special consulting offer. This includes a prepared training environment consisting of an embedded server together with corresponding pre-installed open source software, such as Keras and Tensorflow with a focus on visual machine learning. This package alone saves companies around six weeks of preparation and familiarization time.

Special, simplified software interfaces, which are also provided via SUSiEtec, ensure that .NET and Java developers can also use the corresponding OpenVINO middleware from Intel for the inference part of classic training results in a simplified manner. Unlike with native C programming, only a few lines of code are required in .NET and Java with SUSiEtec AI.

Performance "From Edge to Fog to Cloud" plus new applications

For Industrial IoT and Industry 4.0 concepts, intelligent edge computers bring computing and storage power directly to the machine. Due to both established and new technologies such as cloud computing, container software, OPC UA over TSN and artificial intelligence, the areas of responsibility are changing. The previous pure gateway computers are becoming powerful real-time computers on the machine. This ensures a seamless transition between OT and IT and also makes computing-intensive applications such as artificial intelligence efficient, secure and productive in the manufacturing industry without a direct cloud connection.

Reiner Grübmeyer, Head of Product Management Systems & Software at Kontron / ag