October 7th, 2019, Published in Articles: EngineerIT

by Jack Lin, Moxa

The world today is witnessing a new dawn for digital transformation that will require manufacturers to rethink their existing business models and industrial automation infrastructure. To remain relevant and competitive in the age of “Industry 4.0”, manufacturers need to do more than simply adopt digital technologies and deploy predefined processes in isolated islands of automation as they have done in the past. The future of industry rests on understanding the factors driving this new wave of digitisation, the limitations of the current model of industrial automation to achieving digital transformation, and how to overcome these limitations to realise the full benefits of Industry 4.0.

 Current limits of industrial automation

Ever since the industrial revolution, manufacturers have been looking for ways to increase productivity. Following the mechanisation of production, manufacturers have embraced device connectivity as a means to improving efficiency and profits. Beginning in the 1980s, manufacturers began adopting digital devices, which led to the emergence of industrial automation as we know it today. A helpful way to visualise the current architecture of industrial automation is the often-referenced Purdue model.

In the current Purdue model, industrial automation forms a pyramid where isolated purpose built protocols occupy distinct layers. However, this model also gives rise to a number of infrastructure challenges for industrial networks today. Although independent purpose-built protocols may be very good at automating the original tasks for which they were developed, they are essentially speaking different “languages” which results in real-time communication difficulties. Traditional industrial networks in this model are also tuned for latency and control, unable to “share the wire”, and are often limited to 100 Mbps (or lower) transmission speeds that ultimately hinder scalability. Furthermore, using proprietary hardware and software for multiple applications obstruct interoperability and increases maintenance and operation costs. Consequently, systemic integration and visibility across layers becomes difficult to achieve, which negatively affects the overall value chain.

Evidently, manufacturing strategies also need to evolve for companies to remain globally competitive. Today, customer demands are becoming increasingly diversified and companies are looking for ways to satisfy these new and future needs while also increasing operational efficiency. Businesses strive to remain competitive globally by becoming as nimble, efficient, and responsive as possible. Gone are the days when manufacturers could scale their production based on sales forecasts alone. Instead, manufacturers may need to leverage relevant insights from big data analytics to help fulfil customer demands in real time and optimise production at lower costs. This is just one example of how manufacturers can deploy the latest technology to move towards “Industry 4.0” and come out ahead.

Equipment, devices, and people in complex and global operations are more connected than ever due to the fact that industries are continuing to digitise, automate, and innovate. Ultimately, industrial networks need to catch up with market and industry developments to ensure that businesses can turn the efficiency, flexibility, and availability made possible by a more reliable and scalable network into better performance, higher employee and customer satisfaction, as well as more growth.

The future of industrial automation

The traditional Purdue model, as represented by the “automation pyramid”, outlines different layers of network communication that remain fragmentary and potentially unreliable and difficult to maintain, particularly in the long term. Calls within the industry have been made to move towards an “autonomous pyramid” that is capable of responding to market and business conditions in real-time. In this newly envisioned architecture, isolated islands of automation and network data flows are able to communicate with each other through common semantics and a unified infrastructure.

As illustrated in Fig. 2, this new “autonomous pyramid” envisions the future of industrial automation as a seamlessly connected system where:

• Small-scale, static, and isolated control loops evolve into larger-scale, dynamic, and open control loop communications, also known as cyber-physical systems (CPS) that deeply intertwine software and physical components.
• Formerly closed loop data can openly communicate across a common foundation, enabling new bilateral data communication flows that can intelligently interact with each other.
• All business assets, from equipment to materials to personnel, are intelligently connected in a unified infrastructure that enables the fulfilment of diverse customer needs through end-to-end “autonomous” communications, collaboration, reaction, adaptation, and optimisation, all at the “right times”.

By using a unified network infrastructure for a multitude of disparate applications – including automation, maintenance, analytics, and more – manufacturers can achieve the following benefits:

• As different end devices are able to talk to each other in real time, configuring the system, devices, and applications to enable a real-time feedback loop becomes significantly easier. A unified, context-driven network structure also allows machine learning to take place, so in the long term, it is possible to leverage big data analytics and respond accordingly, further increasing made-to-order flexibility and efficiency.
• Improved access to data allows for production monitoring to be conducted in real time, so higher quality, more detail-oriented KPIs can be established under different scenarios.
• A more robust network infrastructure can support more applications for equipment on the shop floor, such as counting, sorting, quality control, and video surveillance. Machines no longer work in isolation but act in tandem with other machines to improve productivity, thanks to all the real-time data fed into the system. Combined with developing technologies in robotics and machine sensing, such as motion guidance, augmented reality, machine vision, and haptics, factory assets can deliver optimised performance at lower costs.
• Standardised technologies and a scalable structure, such as those based on Ethernet standards, afford much greater flexibility. Through standardisation of infrastructure technologies and protocols, topological differences—which used to present significant challenges for network configuration—can be managed in similar ways as other modular units or extensions. Building, maintaining, and removing layers are more cost-effective and less time-consuming.

Indeed, a unified infrastructure that breaks down the barriers separating the islands of automation in today’s Purdue model will create a system of connected, physical industrial objects and allow them to exchange and analyse data with the purpose of generating value-added information. In doing so, the IIoT enables the right decisions to be made at the right time and place, thereby transforming formerly predefined processes into truly dynamic processes.

Ultimately, the future of industrial automation and control systems is about the integration of information and internet technologies that continue to satisfy requirements for high availability and real-time communications, and also supports the development of new products and innovative solutions based on the optimal balancing of costs and benefits. More precisely, the unified network infrastructure of the future also requires deterministic communication capabilities that can ensure performance and QoS as well as, or better than, the purpose-built protocols that isolate our current islands of automation today. Thankfully, standards organisations and independent vendors have recognised the potential benefits of Industry 4.0 and are working together to develop a new, unified foundation for industrial networks: time-sensitive networking.

Setting the standard(s): Time-sensitive networking

Heeding the call for a unified yet deterministic infrastructure, time-sensitive networking (TSN) is a collection of standards that enables deterministic messaging over standard Ethernet networks. As defined by the Institute of Electrical and Electronics Engineers (IEEE), TSN involves a form of network traffic management to ensure non-negotiable time frames for end-to-end transmission latencies. Consequently, all TSN devices must synchronise their clocks with each other and use a common time reference to support real-time communications for industrial control applications. Although TSN standards were initially developed by the IEEE, it is important to recognise that TSN goes beyond the main IEEE standards and includes the hard work and collaborative efforts of many international organisations and companies.

Early standard Ethernet networks were incapable of guaranteeing data delivery and subject to high latency. As a result, industries that required high network reliability and availability developed their own specialised, proprietary networking solutions (e.g., modified Ethernet networks, fieldbuses) for industrial control systems and automation. To meet the high availability and low latency requirements of industrial applications for manufacturing, traditional best-effort Ethernet technologies have had to evolve to become more deterministic. TSN is essentially the next stage in the evolution of standard Ethernet technologies to satisfy the requirements of our IIoT future. Besides providing a set of standards for deterministic services over Ethernet, TSN is bringing together many different industry organisations and market leaders under a common goal to realise the full potential of Industry 4.0 and the promise of digitisation.

Evolving best-effort networks

Traditional Ethernet networking technologies generally include hubs and switches that employ best-effort packet delivery. Most of the time, data packets are successfully delivered in sequence, but there are no guarantees. Although best-effort networks may perform adequately for web browsing applications, industrial control applications require higher availability, zero packet loss, and lower latency. After all, if there are no guarantees for packet delivery, critical control data might not be delivered to the right place at the right time.

In the 1980s, when manufacturers started migrating to digital technologies from mechanical or analogue technologies, best-effort Ethernet networks were not considered a suitable infrastructure option for industrial control systems that required high precision, availability, and guaranteed real-time transmissions, despite offering higher bandwidths compared to traditional fieldbuses. Besides the prohibitive costs of Ethernet technologies at the time, Ethernet retransmission algorithms and collision detection could not satisfy the performance requirements for industrial control systems. Consequently, manufacturers had to develop purpose-built systems and protocols to enable digitisation through deterministic networking.

Unlike best-effort networking, deterministic networks support the following services:

• Time synchronisation
• Resource reservation
• Extremely low packet loss
• Guaranteed end-to-end latency and bandwidth

Since those early days of Ethernet and industrial automation, networking technology has evolved considerably. In fact, modern Ethernet technologies can even supply deterministic services that meet the needs of many industrial applications that formerly required proprietary systems and protocols. Due to the growing trend towards converged networks and the corresponding increase in demand for bandwidth, truly deterministic Ethernet networks could be more cost-effective and future-proof than purpose-built networks.

Determining TSN

To enable truly converged networks that can stream real-time controls, as well as audio/video in industrial facilities, the TSN task group of the IEEE 802.1 working group is defining a set of standards for the deterministic data transmission over Ethernet networks. As a collection of standards, TSN is more like a tool box than an all-in-one solution; you need to understand what “tools” are available and how each tool works in order to determine which tools are suitable for your application.

As the key protocols described in Table 1 indicate, TSN standards focus on the following main areas:

1. Time synchronisation
2. Latency
3. Reliability
4. Resource management

Protocol	Description	Focus area	Status
IEEE 802.1AS-Rev	Timing and Synchronisation	Time synchronisation	On-going
IEEE 802.1Qbv	Enhancement for Schedule Traffic	Latency	Completed
IEEE 802.1Qch	Cyclic Queuing and Forwarding	Latency	Completed
IEEE 802.1Qbu	Frame Preemption	Latency	Completed
IEEE 802.1Qca	Path Control and Reservation	Reliability	Completed
IEEE 802.1CB	Frame Replication and Elimination	Reliability	Completed
IEEE 802.1Qci	Per-Stream Filtering and Policing	Reliability	Completed
IEEE 802.1Qcc	Stream Reservation Protocol (SRP)	Resource management	Completed
IEEE 802.1Qcw	YANG Models for all TSN Queuing and Filtering Techniques	Resource management	On-going

TSN requires all network equipment to implement IEEE 802.1AS (IEEE 802.1AS-Rev in the future), which defines standards for timing and synchronisation. After all, a shared concept of time among all end devices and Ethernet switches is one of the key characteristics of deterministic networking. In addition, IEEE 802.1Qbv defines how devices must transmit time-critical frames according to a hard schedule, but also retain best-effort communications for other bulk traffic sharing the same “line”. Besides the network infrastructure itself, TSN also requires a new approach to handling data streams and corresponding requirements that require more complex computations.

Consequently, IEEE 802.1Qcc defines the management interfaces, mechanisms, and principles for enabling a new approach to network administration.

For illustrative purposes, TSN can be viewed as a railway system where trains are analogous to Ethernet frames of data. In this example, Ethernet switches and end devices are like railway stations. Imagine what would happen if each railway station kept a different local time without following a strict timetable for the entire system. If a train departs from Station A, how will passengers know when the train will arrive at Station B if the stations do not share a common reference for time? This problem was precisely why railroads began standardising time for railway passengers and trains, and also demonstrates why industrial networks require time synchronisation.

Managing network traffic

An integral part of TSN is a management model that manages and directs traffic streams on the network and allows the family of IEEE protocols to be configured for successful operation on the same network. In our railway system analogy, the network management model is akin to a railway signalling system that handles train (data) traffic, so the trains (payloads) arrive at their destinations without colliding with one another. As stated in the IEEE P802.1Qcc protocol, there are three possible management models, which include a fully centralised model, a fully decentralised model, and a partially centralised model.

• In the fully centralised model, end devices initiate communication with the centralised management entity regarding their stream requirements. The centralised management entity then uses these requests to work out the necessary schedule of streams on the network to satisfy those requirements and configures the switches and end devices (represented by railway stations in our analogy) accordingly.
• In the fully decentralised model, open streams are offered by talkers to listeners (usually end devices) and an application on an end device notifies the network elements along the way to reserve the resources required for a particular stream. No central management entity is necessary in this approach.
• Although there is a centralised management entity in the partially centralised model, the data from the end devices are passed to the closest bridge over a standardised protocol before being forwarded to the centralised management entity. In other words, the centralised management entity in the partially centralised model only manages individual network traffic streams and resources without addressing, at a global level, the stream requirements or payload data from every end device.

As defined by the IEEE 802.1CB protocol, “Standard for Local and metropolitan area networks – Frame Replication and Elimination for Reliability”, TSN stream identification offers several different ways to identify streams. These methods include destination MAC address and VLAN identifier, source MAC address and VLAN identifiers, and others. In addition, stream

Identification is used to compute the flow of data for a specific stream through the network as well as to handle redundant paths for fault tolerance.

A closer look at the fully centralised model

Although the fully centralised network management model is not the only approach for handling traffic in TSN, the model is the most straightforward of the three methods for illustrative purposes. As previously discussed, the fully centralised network management model has a centralised management entity that performs two critical roles. In the following diagram, these functions are represented by the Centralised User Configuration (CUC) and Centralised Network Configuration (CNC).

As depicted in Fig. 5, the fully centralised TSN model includes the following five components:

• End stations (talkers and listeners): These end devices run applications that require time-critical deterministic communications and function as the sources (talkers) and destinations (listeners) of the Ethernet frames transmitted through the TSN system.

• Bridges (ethernet switches): TSN bridges are Ethernet switches that send and receive the Ethernet frames comprising the time-critical communication streams. The hardware can be developed by any vendor but must be capable of transmitting messages according to a strictly synchronized schedule.

• Centralised user configuration (CUC): The Central User Configuration is a vendor-specific application that communicates with the CNC and the end devices. The CUC represents the control applications and the end stations, requesting deterministic communication with the CNC.

• Central network configuration (CNC): The Central Network Controller is a vendor-specific application that facilitates deterministic messaging for control applications on the network and defines the schedule according to which all the time-critical information streams are transmitted and then deployed to the TSN-enabled bridges (Ethernet switches).

• Time-critical information streams: The information transmitted between talkers and listeners in the TSN model comprise time-critical “streams”^[1]. Each time-critical information stream between talkers and listeners in the TSN model is uniquely identified by the end devices and has stringent time requirements that need to be honored for deterministic messaging.

In contrast to the fully distributed or partially centralised models that only handle individual requirements or network capabilities separately, the fully centralised TSN model uses centralised methods to represent both the “user requirements” and the “network capabilities” in order to automatically integrate all components throughout the entire system. Although the fully centralised model offers improved integration, more complex computations are required to ensure better network utilisation. In the end, whichever TSN model you choose depends on the particular requirements for your application and fall outside the scope of the IEEE standards developed by the TSN task group. But since the specific technologies and protocols deployed in each model and application can be provided by essentially any vendor, there is a clear need for independent vendors and other industry organisations to fill the gap.

Coming together for Industry 4.0

TSN technologies offer a scalable, predictable approach to deterministic networking over standard Ethernet. But since TSN is more of a toolbox than a single, comprehensive solution, system integrators must ultimately rely on independent vendors and multiple protocols to satisfy the specific requirements for each industrial application. This predicament is precisely why interoperability is the key to ensuring the success of TSN adoption. Ultimately, a unified infrastructure based on TSN fundamentally requires interoperability on two critical fronts:

• A common architecture that is TSN-compliant for Layer 2 networking and messaging.
• Common semantics for communication across multiple protocols throughout the network.

Recognising the benefits of Industry 4.0 and the future of smart manufacturing, global standards organisations, working groups, and independent vendors are putting their “best effort” into building a common infrastructure and enabling interoperability so that machine-to-machine collaboration, data access from cells, and more applications can be realised.

Today, the arrival of TSN means that standard Ethernet technologies are able to provide deterministic services, evolving beyond the traditional limitations of best-effort communications. With TSN, manufacturers no longer need to confine their applications to isolated islands of automation with purpose-built protocols and control systems. Instead, industrial applications can look forward to an integrated future with new bilateral communication flows that transcend the horizontal and vertical compartments of the traditional Purdue model.

Contact Leanne Meaney, RJ Connect, Tel 011 781-0777, leanne@rjconnect.co.za

Related Articles

South African Government COVID-19 Corona Virus Resource Portal

Now Media acquires EngineerIT and Energize from EE Publishers

Printed electronics: The defining trends in 2019

Charlie and the (fully-automated) Chocolate Factory

SANSA app calculates best HF communication channel