Cloud, Edge and 5G: The Backbone of a Fully Connected Digital Factory

Bridging the Technology Gap Towards the Digital Factory

In manufacturing, cloud computing refers to the utilization of cloud-based services for storing, processing, analyzing, and managing data pertaining to various facets of production and supply chain operations. The cloud provides access to scalable and elastic resources, which are provided to manufacturing companies as a service, eliminating the need for them to plan and install their infrastructure based on capacity requirements.

Cloud computing encounters certain limitations

However, the cloud can only partially accommodate the needs of a digital twin for a production line: traditional cloud computing is technically ill-suited for fully representing a digital twin of an entire factory, which encompasses all assets, products, and personnel within a factory. This is primarily because “data gravity” is prohibitive of frequent or real-time synchronization between the physical and digital twin in the cloud, i.e. the amount of data and refresh cycles are too high to transfer the data back and forth to maintain and analyse a meaningful digital twin of the factory scene.

While planning processes and their results can be effectively represented using the digital model in the cloud, mass data generated on the shop floor cannot be processed in the cloud to support timely decision making for the production line. This limitation is attributed to significant transmission latency and the sheer volume of data involved. To illustrate, consider the video surveillance of a critical process on the assembly line, involving human-robotic collaboration. It necessitates uncompressed 4K video quality to discern crucial details, a minimum image frequency of 50-60 frames per second (fps) for rapid identification of safety risks, and an end-to-end processing latency of just 50 milliseconds (ms) – considered the “gold standard” – to facilitate swift decision-making for quality management and human safety warnings.

In the future, a digital factory will encompass historical, real-time, and predictive insights into the production process. Self-regulating process control loops will empower assets on the assembly line to access data from prior processes in real-time, allowing them to enhance their current process parameters and reduce the occurrence of quality defects. For instance, factors such as room temperature, air pressure, and humidity can significantly impact product quality in Carbon Fibers and Carbon Fiber-Reinforced plastic production. However, such influences can be mitigated by adjusting process parameters.

Edge computing enables latency-critical decision making

Edge computing is a decentralized computing paradigm that brings data processing closer to the data source. This approach delivers low latency, increased bandwidth and the capability for autonomous (offline) decision making. Consequently, it enables low latency digital factory use cases that demand real-time or near-real-time responses. Edge cloud infrastructure can be established within the factory or campus, managed by an internal or external dedicated team, or sourced as a cloud-like service provided and maintained by cloud or other service providers.

5G allows for high-quality communication

5G stands as the sole radio technology that facilitates controlled quality of service for latency-critical communications. It enables the deployment of private, dedicated 5G radio networks within campuses, as well as the utilization of public 5G networks with allocated network slices for specialized applications. This presents an appealing alternative for mobile use cases, such as nationwide logistics across warehouses and manufacturing plants, and for small- and medium-sized enterprises.

Within the factory environment, 5G complements various existing technologies, including WiFi which is often used for devices from the IT domain (e.g., mobiles, handhelds, printers, terminals) and ultra-wideband (UWB) which is used to derive positioning information. 5G can be employed for communication with controlled quality of service, for location triangulation via upcoming protocol standards, and the aggregation of other network technologies needed on the shop floor to obviate the need for labor-intensive wired network upgrades.

Edge cloud and 5G provide a high-performance and scalable infrastructure

5G and edge cloud unlock possibilities that are presently unattainable due to existing WiFi latencies or practical constraints related to wired network infrastructure, such as limitations on the number of installation and network extension cycles allowed per year when production must be paused.

Exemplary use cases

Factory asset positioning

Allows real-time location system of materials, equipment and vehicles in production halls and warehouses to support output, quality and logistics applications

• This allows for:
  • Optimized material flow
  • Reduced standstill/downtime
  • Real-time ERP connection
  • Enhanced SC transparency
  • Improved planning foresight
  • Enablement of digital production processes
  • Improved root cause analysis
• Tech requirements:
  • Precise positioning (~cm range) with high reliability (≥99.999%) and low latency connectivity (sub 1 – 500ms)
  • Real-time edge computing and analytics, sensorics (e.g. motion, health)

AI-based video control

AI-based computer vision systems to automate visual monitoring and inspection throughout the entire production process to support safety, productivity and quality.

• Benefits:
  • Early detection of safety issues and defective equipment
  • Optimized traffic (e.g. AGV navigation)
  • Assured product quality (e.g. visual inspection)
  • Optimized supply chain (e.g. inventory monitoring)
• Tech requirements:
  • Fusion of static and mobile computer vision
  • Sensorics (e.g. motion, collision, video)
  • Precise 5G positioning (~cm range)
  • Real-time edge computing and analytics (incl. AI for video analysis, trajectory prediction)

Safety improvement for mixed human/mobile robot areas

Real-time camera system detecting safety risks (e.g. collision) in mixed human/mobile robot areas with automatic asset emergency stop and deployed safety zone my mobile robot

• Benefits:
  • Early detection of safety risks & reaction
  • Increased production uptime
• Tech requirements:
  • Fusion mobile & static computer vision, sensorics (e.g. motion, collision, video), precise 5G positioning (~cm range)
  • Real-time powerful analytics (incl. AI for video analysis, trajectory prediction), navigation steering & control of mobile robots (sub 20ms, 99.9% transmission reliability)
  • Interfaces to mobile robots/ fork lifts incl. emergency stop

These use cases are particularly data- and compute-intensive and may rely on artificial intelligence for their realization. For instance, in the context of video control, AI serves as an intelligent observer, utilizing advanced algorithms to process video data, identify patterns, and offer real-time insights that enhance safety, quality, and efficiency in the manufacturing process. With time, AI continually refines its accuracy by learning from new data through fine-tuning, incremental learning, and generalization, while adjusting model parameters to minimize disparities between predictions and actual outcomes through loss optimization.

5G and edge cloud enable connectivity, authentication, and over-the-air updates as managed services, bringing the vision of a “plug and play” factory within reach for manufacturing companies. Examples include physical positioning, network configuration, and the automatic linking of “neighboring” machines.

These examples show that the next step of digital factory transformation requires new infrastructure paradigms that allow the establishment of data rich digital twins infused with live data from the shop floor, giving rise to new insights about live production and other benefits.