By Alexander Damisch, Director IoT, Wind River
The Internet of Things (IoT) is not just a technology, or a system, or an architecture—it is mainly a business case, and requires a combination of all these pieces to fulfill its promise, which is a smarter way to do business. A major use case for IoT is predictive and preventive maintenance. The ability to accurately diagnose and prevent failures in real time is a major advantage for companies and is vital for critical infrastructure applications. A failure of high-tech machinery and equipment can prove to be highly expensive in terms of repair costs, in addition to lost productivity from the resulting downtime. Historically, technicians have been sent to carry out routine diagnostic inspections and preventive maintenance according to fixed schedules, which can be a costly and labor-intensive process with little assurance that failure will not occur between inspections.
One example in the renewable energy sector is a wind farm—and in the extreme case, an offshore farm. Wind turbine systems contain a great deal of technology, including a generator, a gearbox, and a multitude of electronics, including control systems to adjust blade pitch and many other parameters. If any element fails—due to dust buildup or cumulative vibrations, for example—the remote location means the repair cost will be extremely high. In addition, given that weather will always be a significant factor in this particular example, a turbine may not be producing electricity for some time.
However, onsite sensor-equipped systems can collect data from multiple turbines, not just a single turbine, enabling failure analysis to be performed to predict when a system or component is likely to malfunction due to stress or overheating, and thereby enabling better operator or autonomous decision-making for maintenance. For example, if there is a high likelihood of the gearbox breaking down within a turbine, then switching to a lower performance mode and a reduced mechanical load, while still delivering 80 percent efficiency, could mean continued operation and further electricity generation for several weeks. This would allow scheduled maintenance that combined the repair and maintenance of more than just one turbine. This clearly shows the value of the adaptive element of control and analytics is key for best possible performance.
A second important use case is adaptive analytics, which involves looking at an overall system or a system of systems. Based on much the same data already being collected for predictive maintenance, adaptive analytics enables equipment and devices to analyze enormous amounts of data and make real-time decisions to help refine and improve operational processes.
The adaptive analytics and predictive maintenance capabilities of IoT can also play a significant role in providing opportunities for new revenue streams, and not just reducing OPEX. In industrial markets, for example, historically the big players have had two main ways of generating revenue: the traditional way of selling devices such as control systems or motor drives or human–machine interfaces (HMIs); and complete hardware and software system solutions including maintenance service level agreements (SLAs). But both of these business models are under intense pressure from a cost point of view, with increasingly strong competition and significantly reduced margins—especially after the recent financial crisis, which has meant over-saturation in the market for manufacturing equipment.
In the case of major industrial and automotive equipment OEMs, they have a huge base of already installed equipment at customer premises—which can be a challenging situation for innovation. However, in addition to new products or assets, one way to increase revenue is to obtain new recurring revenue streams based on existing and already deployed devices. The ability to innovate and deploy simple solutions that allow the connection of equipment to IoT can provide customers with significantly reduced operational costs and additional value via predictive maintenance and adaptive analytics. A service fee could be based on production volume, number of deployed devices, or a certain amount of data. This could give rise to a subscription-based business model including the leasing of equipment with ownership retained by the device manufacturer.
However, while this model can work quickly and easily in some markets (for example, many smaller companies now moving from M2M to IoT have already transitioned from charging for devices to charging for data volume or a specific analytics service), many large industrial systems are well entrenched technologically, so there are many challenges to overcome. There has always been data collection in manufacturing and process automation markets via Supervisory Control and Data Acquisition (SCADA) industrial control systems. But data is collected in a very static way in a SCADA system, with no real-time access to information, and the existing OPC and OPC/UA protocols are not nearly enough. The reality is that over the past decade, much automation equipment has moved into a more network-like environment, where most of the parameters are being exchanged between devices via primarily IP-based industrial Ethernet protocols such as PROFINET, Ethernet/IP, EtherCAT, TSN, or Ethernet POWERLINK.
If a data-aggregation device or gateway is deployed, though, it can become an interface between devices, albeit devices that were designed for local operations technology (OT) communication only, without expectations of connecting to an IT system, and certainly not to send vast quantities of data up to an IoT cloud-based platform. This gateway therefore presents a huge challenge: the security perimeter requirement, essentially protecting against hackers and other threats. In the energy generation industry, there are Critical Infrastructure Protection (CIP) cyber security standards, created and administered by North American Electric Reliability Corporation (NERC) for the U.S.; and there are various International Society of Automation (ISA) standards for industrial control and automation markets.
Security is absolutely critical in the IoT environment, protecting equipment and assets from the outside during system boot time and run time, and preventing possible system shutdown or even potential threats to functional safety, while still enabling communication with the devices to obtain the data.
A second necessary component is system manageability and customization—there is limited value in the gateway if devices or platforms cannot be given orders because they object to a slight deviation from the process parameters. In most applications, there is no need to receive data on hundreds of parameters every other millisecond, so the system needs to be managed in a post-deployment phase—by connecting to a device and installing a new software application to filter the volume of information available, for example.
A third key requirement for IoT is obviously connectivity. In industrial automation in particular, there is a move away from the old-style static or cyclic data gathering to Ethernet-based collection. Protocols now becoming standard in IoT are Extensible Messaging and Presence Protocol (XMPP), which is primarily a one-way protocol and therefore largely secure. Another is MQ Telemetry Transport (MQTT), which is a lightweight publish/subscribe messaging transport protocol and is especially useful for communication with remote locations that require a small code footprint.
In Figure 1, a simplified end-to-end IoT architecture shows the combination of different layers requiring expertise from different market segments. At one end are sensors and devices. Sensors could be parking sensors, a traffic flow sensor in a smart city, or an actuator such as a valve in an industrial application. Through the well-known wires or bus system they would connect to a device, in most cases called a controller. Today, this is where most devices are connecting with a data aggregation or SCADA system. Currently these are mostly local supervisory systems connected to statically provisioned controls. While moderns systems can be reconfigured “on the fly” for additional data tags, they still require commissioning, and do not offer event based data aggregation supported by dynamic local intelligence, such as an algorithm that could be updated based on analytics.