IT in Manufacturing

Building the industrial Internet of Things

May 2014 IT in Manufacturing

If the Internet of Things (IoT) is already big news among retail manufacturers, it is even bigger news in heavy industries like energy or shipping and transportation. For these companies, the possibility of rolling out cheap, massively distributed networks that can collect and communicate consumer and process data offers tantalising and easily understood business opportunities. The only challenge is finding the hardware that can do the work. IoT is the endpoint of the convergence of industrial automation with information technology currently underway. Moxa is already preparing for it with a full line of hardware solutions built for industrial cloud applications.

The platforms at the base of the Internet of Things

There is a pressing need amongst industrial automation suppliers for embedded computing platforms that are optimised for machine-to-machine (M2M) communications. 2013 demonstrated that the IoT is poised to become a material reality offering great possibilities to industrial providers. Yet until the lack of specialised platforms is overcome, the building of IoT clouds for industries like traffic management or the smart grid will continue to be delayed. The fundamental challenge in designing these systems involves negotiating the divide between IT and IA technologies. Industrial automation protocols are fundamental to the proper functioning of edge devices that make up the bulk of mass deployments; fieldbus, Modbus, I/O configuration, serial interfaces and the gateways that link all of these are common enough problems for industrial automation engineers, but for most IT engineers the entire field appears esoteric and mysterious.

On the other hand, securing networks with firewalls and VPNs, protecting against dropped packets or node failures, and the multifarious problems introduced by wireless communications are all things typical for IT, but which IA engineers would rather avoid.

The biggest challenge today when building an industrial M2M network is its massively distributed nature: there is very little value to be gained by a gradual transition that takes years. IoT deployments must be rapid enough to sustain investment value, but the massive number and scope of the cloud of devices employed makes completing the full installation speedily pretty much impossible. Specialised technicians are required for every individual station, and they must be competent electricians as well as familiar enough with network design to juggle the details of multiple protocols, interfaces and communications media. Unfortunately, few IT professionals have ever worked with the protocols and interfaces that are most common to industrial automation networks, and setting up I/O stations, or configuring sensors, are things in which IT people simply have no experience.

Thus, at the outset, there are two important reasons why effective software automation at the connectivity layer is critical: first, to facilitate the deployment of a cloud of devices that may ultimately include hundreds of thousands of nodes, and second, to flatten the learning curve as much as possible for the people who will be installing these devices. Software tools that automate the rollout of industrial automation devices and simplify overall network deployments are, therefore, a foremost consideration for anyone who will be managing the deployment of an IoT network.

Industrial grade stresses, enterprise class challenges

Yet another way in which IoT networks significantly differ from consumer networks are their strict availability and reliability requirements. Industrial M2M systems for intelligent transportation systems or the smart grid will, of necessity, operate 24/7, 365. At literally any moment the network must be able to call upon remote stations at the edge and command them to make adjustments, return data, or perform maintenance checks. Obviously, devices which are not capable of reliably maintaining network connectivity for years on end will not be very valuable to network administrators. Similarly, all devices along the network must be able to deliver key information for preventive maintenance and respond to a wide variety of common network challenges such as failed nodes, network congestion and wireless re-association. For the most reliable performance, network redundancy, automated connectivity checks, preventive maintenance routines and effective maintenance, monitoring, and control protocols must be integrated as deeply into the hardware level as possible.

Finally, there is the consideration of the related problems of data integrity and network security. Authorisation, access and accounting controls are as imperative for an M2M network as they are for any IT network. Authorisation and access are easily understood: illicit access to a massively distributed industrial network by a hostile party has clear potential for disastrous and potentially lethal consequences. For this reason, IoT networks (especially those for traffic systems, the smart grid, or other power applications) must support the strongest possible encryption and access controls. Similarly, accounting controls over the entire network are important not only for the monitoring and management of the network itself, but also to aid in preventive maintenance, as well as to perform forensic analysis on suspected intrusions or other security breaches.

Taken together, these imperatives amount to a lot of work that, until recently, had simply not yet reached a stage where machine-to-machine communication networks could be considered viable. Now, however, that has changed and the IT/IA convergence of recent years has arrived at a point where integrating these two technological realms in a secure, reliable, cost-effective manner has become a relatively easily achievable reality.

Converging solutions for converging technologies

The overall technical challenge that must be addressed when building an industrial IoT network may be broken down into three key aspects:

1. Network topology and how it relates to device deployment and engineering.

2. The deployment, setup and maintenance of network nodes and edge stations.

3. Overall monitoring and control.

First, let us consider how an M2M network can be envisaged in terms of its topology and how the technological implications of that structure affect the network’s design and implementation. At first glance, it might be tempting to simply divide the network into two layers of two dimensions: network/process, edge/core. That, however, would neglect some important opportunities for automation and optimisation. To begin with, the physical devices of the network are best broken up into three concentric layers, rather than two: the core, the connectivity layer and then the terminal edge stations where nearly all of the remote process data and events will be generated. Edge devices will necessarily include sensors, automatic metering infrastructure, embedded computers for control and monitoring and gateways to bind all of these devices together to allow effective communications between the various parts. The question then becomes: how can automation and effective device engineering speed up and simplify M2M deployments, monitoring, and management?

When viewed from this perspective, it is clear that when considering how best to optimise deployment efficiency (the second of our three points, above), the main areas where industrial automation and controls can give the most value is at the connectivity and edge layers. The setup and installation of devices that communicate over serial protocols and interfaces at the edge will need to be automated so that configuration tasks like addressing, tagging and logic programming can be reduced to processes that set up hundreds of devices in seconds, without any need for a technician experienced in low-level programming languages. Modbus and other fieldbus protocols will need to be supported, and to as great a degree as possible, these software optimisations should be centralised and administered from a central control station. In the ideal situation, a technician would only need to connect a device, test the interfaces to verify the communication links have been established, and then the device would be automatically detected, configured and initialised by the remote control centre.

Transparency for connectivity and communications

With effective engineering and design, computing platforms and network nodes in the connectivity layer may be reduced to a nearly transparent layer of automated deployment, monitoring and management. Embedded computing platforms should be engineered with as wide a variety of communications interfaces as possible, with software features that monitor and automatically respond to changes in the communications and connectivity layers. Switches and routers should be remotely configurable and support a robust implementation of SNMP that can trigger automated alarms, regular service logs and detailed diagnostic reports. Cellular stations should be capable of automatic re-association whenever a link goes down, and all other wireless links – ZigBee, Bluetooth, 802.11, or sub-gigahertz – should feature analogous reliability and availability assurances. With effective redundant ring backbones over NPL, BPL, or wired Ethernet, coupled with highly redundant, rapid-response wireless mesh topologies, a machine-to-machine network can virtually guarantee constant, highly available links with any associated edge station.

This is not to say that software optimisations at the core will not also be powerful aids. Yet in M2M networks, the central control, monitoring and management solutions will generally be more or less customised packages that are built and maintained either by the end user, or by a contracted system integrator who has built and is maintaining the system. Yet there are one-size-fits-all optimisations that are very useful here, as well, such as a master gateway interface that unifies the cloud into a single, centrally administered application. This master gateway would be much more than a simple scada or control HMI; it must incorporate software optimisations that optimise the rollout of devices, be customisable to the needs of any end-solution required, and should incorporate as wide a selection of IT and IA interfaces (serial, Ethernet, or wireless), standards (RS-485, 802.11, Bluetooth, ZigBee), protocols (Modbus, Profibus, TCP/IP) and legacy systems (including proprietary systems) as possible. If designed with suitably intuitive controls that have integrated the connectivity- and edge-layer optimisations discussed above, deployments of networking and edge devices could be reduced to a simple process: the remote technician plugs it in, the central control sees it light up, and then with a few quick clicks of a mouse the control centre can set up tagging, addressing, logic configuration, and a whole host of automated maintenance and management processes.

Flexibility that simplifies development and adaptation

To guarantee that the network remains as customisable and flexible as possible, open platforms should be utilised wherever they are prudent. Linux/GNU and other open source solutions provide an excellent platform for IoT integration, and may reliably power both RISC and x86 platforms. These proven software solutions offer strong security (for both data integrity and AAA protocols) while providing a wide-open system that allows customisation, optimisation, and feature development on any subsystem process, no matter how low- or high-level it may be. Linux/GNU systems also offer two additional advantages: strong security in the form of packet filtering, firewalls, VPNs (and the strongest RSA encryption available), along with the important benefit that end users can escape the danger of proprietary lock-in, whereby a device may become useless should the manufacturer one day disappear, or decide to cease support for that particular line of hardware. Thus, system integrators and end users alike benefit powerfully when using open source/free software solutions like Debian.

Software optimisations are not, however, the only consideration. The physical devices that make up the IoT network must also be specifically engineered for customisability, security, reliability and deployment flexibility. For embedded computers, features like a modular design are a critical feature that will allow end users to adapt devices to specific roles within the network, or even to repurpose a device that is being used in an obsolete role. A wide variety of communications modules must be available, as well: ZigBee, Ethernet/IP, 802.11, cellular and fibre are a must.

Five principles to guide the engineering of an IoT platform

Taking all of these observations into account, a clear vision emerges of what kinds of embedded computing platforms should be sought out when preparing to build an IoT solution.

1. IoT networking devices should conform to the strictest standards of flexibility, reliability, and security, starting with the physical hardware and then moving on up through every networking layer, right into user-space.

2. Software optimisations that automate configuration, setup, and the overall deployment of embedded computers and other edge devices are critical components of an effective IoT architecture.

3. All elements of an M2M networking platform should be able to be easily integrated into high-level, customised IoT customisations, to aid (rather than hinder) the optimal administration, maintenance, and management of the network as required by the particular vertical market it serves. At the highest central administrative layer, smart grid IoT solutions will share very little in common with intelligent traffic systems, while solar farm solutions will be distinct from both. IoT networking platforms must not intrude on the work of building the final solution envisaged by the customer, but should assist in achieving that goal in every possible way.

4. An IoT networking platform must make the connectivity layer as transparent as possible, effectively turning the intermediate portion of the network between the edge and the core into a black box, with which system integrators and application engineers never need concern themselves.

5. Communications between the edge and the core must reliably process all data, regardless of the health of the network as it is accumulated. This means asynchronous, encrypted transmissions between the edge and core, with strong fail-safes to guarantee the physical integrity of the data.

As these considerations reflect, any computing platform (or other networking device) intended for use in IoT deployments should be engineered from a system-wide perspective, where each device is viewed as part of a mutually-supportive interlocking whole, rather than as a single, one-off network tool. A recent solar power deployment for the smart grid is useful as an example, here. Consider a solar station where an embedded computer is used as a gateway and station controller for a residential solar generator system. In this system, the inverter, meter, and remote I/O system are all connected to an embedded RISC platform that manages local automation and data logging, while maintaining remote wireless communications with the central control station. This residential system must therefore maintain network communications over serial, Ethernet, and fieldbus interfaces, as well as either 802.11 or perhaps cellular wireless.

By way of example…

As IA engineers know all too well, when forced to work at the lowest programming layer the configuration of remote I/O gateways (for automated alarms and event-triggers) is a laborious task. Similarly, setup of fieldbus devices (using, say, Modbus) or wireless addressing and failovers can be equally toilsome. This is where pushing automated features out to the very edge of the network is most useful, and can deliver the most cost-effective solution. Engineers want all of these basic tasks reduced to simplified steps that may be bunched together or automatically initiated using a single, common process. In this way, masses of edge devices may be efficiently and rapidly configured with a dramatically reduced need for user input.

Moxa software is an excellent example of how platforms with integrated, automated setup and administration utilities put the work of configuration and maintenance well into the background of system design and deployment. MXconfig is a mass configuration tool that speeds up the configuration of 100 switches by a factor of 10. MXview uses SNMP to automatically discover, query, and configure edge devices (including the setup of OPC 2.0 tags), and then automatically assembles the results into a visualisation of the wired ring. At the same time, Synmap, the virtualised process monitoring and control interface, also utilises SNMP to serve as a universal control protocol that may be used to script any device that supports it, without any need for further compiling or low-level adjustments. Smart Recovery is a fully automated, BIOS-level system re-write utility that allows administrators to both trigger remote rewrites of the entire software system, or to configure remote devices for fully automated recoveries at either scheduled times or critical events. Because Smart Recovery operates at the BIOS level, it is even capable of restoring a system that has become so corrupt it can no longer boot up. Moxa’s DA-Center automates the setup and administration of databases, easing the conversion and display of field data and simplifying connectivity setup with edge I/O, while Active OPC Server delivers asynchronous, event-driven push communications – from the edge to the core – for remote I/O devices, while enabling DHCP addressing on remote ioPAC units. For cellular wireless connections, OnCell Central Manager and Guaranlink ensure that cellular nodes with hidden addresses may be set up for direct communications with the core, while also providing network association fallbacks in the case of station failures.

When combined with modular computing platforms like Moxa’s DA series of rackmount servers, or the UC series of universal computers for embedded and edge solutions, these software enhancements allow for secure, reliable, highly automated deployments of massively distributed networks like those now envisaged by smart grid solutions providers, traffic systems engineers, or residential solar power providers.

These are early, strong steps towards creating a virtualised connectivity layer specifically engineered for industrial cloud solutions. The automation involved vastly simplifies the deployment, setup, and management of industrial networking and edge devices, while consolidating and simplifying their management at the central core. By calling upon tailored software solutions carefully integrated with key hardware optimisations in networking, I/O, and embedded computing platforms, industrial cloud engineers will be able set aside the work of connectivity integration and low-level coding to concentrate on the work of developing the most effective system for their needs.


Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

Mitigate industrial network vulnerabilities
July 2021, RJ Connect , IT in Manufacturing
It must not be forgotten that ignoring common system vulnerabilities in today’s world could put your entire network at risk.

Digital transformation in mining
July 2021 , IT in Manufacturing
Rapid advances in technology have disrupted industries and businesses around the world, which are now being forced to accept this new reality rather quickly because the new technologies have tremendous potential to deliver value for those who adopt them.

Digital twins require accurate and reliable data to be effective
July 2021 , IT in Manufacturing
At the 25th ARC Industry Forum, hosted virtually, many industry experts gave their perspectives on accelerating digital transformation in the post-Covid world.

Frequency analysis without programming requirements
July 2021, Beckhoff Automation , IT in Manufacturing
Beckhoff expands TwinCAT Analytics with easy-to-configure condition monitoring functions.

Siemens adds AI to Simcenter
July 2021, Siemens Digital Industries , IT in Manufacturing
Siemens Digital Industries Software has announced the latest release of Simcenter Studio software, a web application dedicated to discovering better system architectures, faster. Simcenter Studio offers ...

ABB technology can help make SA steel industry competitive
June 2021, ABB South Africa , IT in Manufacturing
South Africa’s steel industry needs to invest in technology like automation and data analytics if it is to improve its productivity to the point where it is globally competitive.

Siemens expands CFD simulations
June 2021, Siemens Digital Industries , IT in Manufacturing
Siemens’ Simcenter portfolio expands capabilities for frontloading computational fluid dynamics (CFD) simulation and increased productivity.

Digital twin for refinery production
June 2021, Yokogawa South Africa , IT in Manufacturing
Within Repsol’s Industrial Business, the development of a refinery digital twin leads the digitalisation program. The digital twin maximises production while optimising energy consumption.

Building secure networks
June 2021, RJ Connect , IT in Manufacturing
This article explores how to build resilient industrial networks and deploy cybersecurity defences in order to sustain continuous industrial operations.

Improving the state of OT cybersecurity by sharing experiences
June 2021 , IT in Manufacturing
It is essential that we view improvements in cybersecurity as requiring improvements in multiple areas, including people and skills development, governance and process development and improved technology.