How will the Internet of Things (IoT) actually be implemented in an industrial enterprise? Many of the separate elements have been in place for a couple of decades, so the challenge is to understand what needs to be done to pull them together to function as a whole.

Part 1 of this four-part series defined IoT as a conceptual framework about communicating useful information between devices. Part 2 focused on specific practice and began with a definition from Tom Moore, IHS Analyst II, Industrial Automation, who says, “IoT is literally collecting sensor data from IP addressable sensors anywhere in the world for anything and transmitting the data.”

Part 3 examined three key challenges to adopting IIoT: technology, economics and people.

This final article addresses some of the nuts-and-bolts approaches for migrating to the Industrial Internet of Things (IIoT). This migration requires having an idea of what the IIoT will ultimately look like but equally important, a realistic assessment of where we are now.

For example, as suggested in Part 2, the IIoT networking protocol will eventually be based on Industrial Ethernet. In an article in the Industrial Ethernet Book issue 84/1, however, Tom Moore, IHS Analyst II, Industrial Automation, says, “While Ethernet certainly offers more modern technology, the usage of fieldbus is likely to continue long into the future.” In an interview for this series, Moore points out that industrial automation has a legacy of serial-based networking technologies and that some factories have lifecycles of 30–40 years.

Mark Watson, associate director for Discrete and Process Automation, IHS Technology, writing in IHS Quarterly says “smart manufacturing’s development has barely begun. Industrial manufacturing not only lags other fields in connectivity and automation, it also has substantial established assets that are hard to replace quickly, including older embedded networks, stand-alone sensors and dumb equipment—as opposed to smart machinery.”

Harvesting Information from Legacy Systems

The significance is that in order to move toward the IIoT, the emphasis has to be “getting connectivity and extracting information from existing processes,” says Mike Fahrion, director of product management at B&B Electronics. The largest challenge, therefore, is to harvest the information from existing siloed SCADA (supervisory control and data acquisition) systems and feed it into a network so it can be shared by multiple devices. Although the network will have to be Ethernet-based in order to build an IIoT, “the reality is that for the next pretty long period of time we’re going to have a gateway-centric architecture—an Internet of gateways,” says Fahrion.

Each gateway will have an IP address and all will be networked together using wire, fiber, wireless and cellular. The input side of each gateway will connect to whatever legacy protocol is being used: Modbus, DF1, proprietary and standards-based protocols over different physical layers, some serial-based and some Ethernet-based. The gateway can be used to aggregate the input data locally and decide which of it should be sent to the application.

The situation is complicated by the fact that when new automation equipment is added to a system it is likely to come with an Ethernet port built in. The engineer is one step closer, but even so, the device might only be designed to communicate with like devices. In that case, although a gateway still would be used, it wouldn’t have to translate between two different physical layers, for example, from Modbus to IP packets. The system could be programmed to directly grab significant data to send upstream. It could be used, for example, to connect a PLC with a VFD, both running Rockwell Ethernet/IP.

One method for integrating an existing system is to add an overlay network. This could be done in several ways. The main criteria for choosing the best approach are both technical and cost-based. Laying a network on top of an existing system rather than modifying it has distinct advantages. First of all, if the system was set up with significant expenditure of time and money and it has been running reliably for a number of years, the IT team is not going run the risk of upsetting a good thing.

One method to avoid that problem would be to add a redundant set of sensors to transmit the data back to the gateway, either over a wired or wireless network. The existing network in this case would not be aware of the overlay. Another nonintrusive approach would be to monitor the existing Modbus traffic and collect information from significant addresses. In another model, a master processor would be added to the existing Modbus network, tied to the system controller and then used to obtain the needed information. Since this method would impose an additional load, the controller would need to be sized correctly.

Real-World Networking

The first step toward implementing IIoT is to make information about all of the processes in a plant available in a single SCADA system. This can include one or more HMIs and one or more PCs.

What would a typical real-life system upgrade look like? Ken Johnson, an instrumentation and controls tech with Array Systems LLC, a systems integration company, says that his company uses PLCs for logical operations with 24 Vdc discrete I/O and 4-20mA for sensor data. Motor drives are often interfaced using Modbus. System data output from the variable frequency drive (VFD) is transmitted over a Modbus network through the SCADA system, which transmits information to and from the HMI.

Information is also sent to a PC, which stores operational data and uses it to produce trending information. Data trends can be useful for a predictive maintenance (PdM) system, in which maintenance can be scheduled during convenient down times before an actual failure has occurred. The system notifies appropriate personnel if devices such as motors, pumps or valves are starting to show signs that they may be heading toward failure. Symptoms might include higher current draw, elevated motor winding temperature or excessive vibration. Another important function for the PC is storing the data history required for environmental compliance reporting.

Johnson says that in a typical architecture for one of his installations he would install an Ethernet communication module into the system PLC. An Ethernet path to a remote panel in another room or across a campus would generally be run over copper, typically category 5e, or a combination of copper and fiber. The rack might contain remote I/O and an Ethernet communication module. The remote I/O will show up on the PC and HMI as part of the system architecture. The I/O on the remote panel can connect through another communications module to additional systems either inside or outside of the plant.

To address nodes from this panel, Johnson typically uses ControlNet, a fieldbus protocol administered by ODVA. The ControlNet physical layer is RG6 coaxial cable, which makes it particularly immune to electrical noise. It is a high speed protocol and (as opposed to Ethernet) it can be scheduled to communicate with different sensors in a predetermined order. ControlNet modules can be used in a redundant configuration in which two cables are run through separate conduits so that if one is severed, the system will fail over to the other. This feature is especially important for operations such as controlling a critical cooling system.

Some simple processes can be controlled with small PLCs such as Rockwell’s MicroLogix, which have both Modbus and Ethernet ports. The Micrologix processor can be used to control a VFD over Modbus and to interface with the main network over Ethernet.

Remote I/O important for the process is usually run on a dedicated Ethernet network separate from that used for data collection in order to ensure maximum speed (minimum latency).

The remote sensors typically use Highway Addressable Remote Transmitters (HART), which superimpose digital data on top of 4–20 mA analog signals. Scaling, configuration and calibration checks can thus be done at the “smart” sensor using a digital interface.

These same techniques are used to upgrade a legacy system or install a new one. Since these systems are all integrated via Ethernet, connection to office-level systems is possible, but not often used, says Johnson.


Security issues are a source of hesitation when considering installing networks that extend beyond a plant. A widespread effort exists to deal with improving security as a key element for supporting the growth of IIoT. Corporations that once paid little attention to cyber security are now setting up groups of engineers to focus full time on the problem.

Tom Moore of IHS describes a demonstration of one such solution he witnessed at a recent conference. A dummy system was secured through a “demilitarized zone” (DMZ) and a virtual private network (VPN) using an industrial PC. A screen at the conference displayed process data from a packaging line at a remote location. However, the engineer at the conference could not directly communicate with the processor. That could only be done at a control center located in a different city. In the example, a failure on the line was simulated and an operator at the center alerted the process engineer about the failure. The engineer analyzed the problem and told the operator at the remote center what corrective actions to take. Although information was transmitted over a long distance, there was only one secure path into the system and that was from a protected location.

IIoT Today

This series has presented an overview of where we are in what seems to be an inevitable progression to the IIoT. Engineers already have the theoretical means to implement it, but in real terms it will take time to solve the remaining problems and develop the will to make the required investment. A positive motivator is that despite the size of the initial investment and the difficulties involved, over the long term the investment will be paid back in reduced downtime, increased productivity and quality, and energy cost savings.