Distributed controls in the Internet of things create control engineering resources
Manufacturing companies will need to address support issues with the expanding coming Internet of things (IoT) as discussed last month, and the IoT also will provide for a new generation of automation systems that will have unparalleled capability and extensibility. There are a lot of spare CPU cycles available for control system applications when a system is made up of thousands of smart devices, all communicating using an IP Ethernet-based communications protocol. Even small nodes will have extra memory and computing power because of decreasing costs and increasing functionality as semiconductor electronics follows Moore’s Law. This extra memory and computing power can be effectively applied to distributed control strategies. While each node may only have a little spare capacity, each node only needs a little spare capacity when a control strategy is distributed. The smallest connected device can be a simple sensor, motor, or actuator. In essence, each connected device in the industrial Internet of Things is a single I/O point controller with a high-speed Ethernet interface that acts like a control system backplane.
Distributed control advantages
Automation systems have multiple models for distributed control, including the Fieldbus Foundation distributed control model defined in IEC 61158-2, the IEC 61804-2 function block model, the ISA 88 control modules and equipment modules model, and the emergent ISA 106 standard for procedure automation model. Function blocks, control modules, and equipment modules can be distributed to devices, each exposing specific functionality and accepting specific commands and parameters. These models also define a hierarchy of control strategies, with higher level strategies working in concert with low-level basic control. The modular approaches to automation systems, the growing functionality of connected devices, and the increasing speed of networks are the perfect storm for truly distributed control. We are fortunate in the automation industry that we don’t have to invent anything to take advantage of the coming industrial Internet of Things; all we need to do is to put all of the pieces together into a system.
Even today a DCS (distributed control system) is not really distributed. It is still a box that owns a collection of I/O points. Newer DCS systems are partially breaking the rule that I/O is exclusively owned by one controller, but you still need to add expensive controller boxes any time you just need to expand your I/O. A truly distributed control system would incrementally grow in capability as new sensors and actuators are added.
This automation capability isn’t just available in manufacturing applications; the IoT will be part of building automation systems, safety systems, ship control systems, utility control systems, and multiple other control applications. One new office campus currently under construction has over 500,000 Internet-connected devices in the building automation system. The devices include light switches, light sensors, door sensors, fire control, irrigation control, HVAC, parking space sensors, and security sensors.
Cisco calls the concept of a collection of low-level computing "Fog Computing," indicating that computing is not up in the cloud but closer to the ground. In the Fog Computing model the execution systems are using the spare cycles available in switches and routers and these low-level systems perform functions such as real-time, actionable analytics, processes, and filtering of data that is later pushed to the cloud. In the case of a truly distributed control system for automation, all of the computing capability is even closer to the ground and might be called "ground-level computing," where every I/O point has computing power and can participate in a truly distributed control environment.
A truly distributed control system may have "execute only" nodes that don’t have I/O for those control strategies that cannot be effectively distributed, such as multivariable control, model predictive control, or large-scale nonlinear optimization strategies. HMI systems could be "display only" nodes, which provide monitoring and operator control capability, but they don’t participate in control strategy execution, so they also become an easily expandable capability. Specialized nodes may be made for high-speed signal analysis from vibration sensors and vision systems.
Distributed across devices
When each device has sufficient memory, which may be a GByte or more, each may be able to hold all of your control strategy programs, executing only its part of the strategy. This information can be used when a device is replaced, by allowing the new device to discover who it is and what part of the control strategy it is supposed to execute by talking to its neighbors in its local segment. In a truly distributed control system, each node contains all the rules needed to reinitialize a new node and execute backup strategies in the case of device failure, essentially providing a holographic-like redundancy and backup model.
If you want to participate in this new functionality, it is important to ensure that your infrastructure and training are ready. Single I/O point controllers may be located at the device, in remote instrument enclosures, in junction boxes, or in termination rooms. Connections among devices in an enclosure may be wired, but wireless networks using the ISA100 standard can significantly reduce wiring costs and simplify adding new devices. If wired connections are used, then each enclosure will have an Ethernet switch. These should be managed switches, with rules to ensure that no node can "chatter" and monopolize the communication channel. Distributed control strategies also define communication paths between nodes, and these paths could be loaded into the switches to ensure that only authorized traffic is allowed on the control network. Make sure that the infrastructure is set up to support 100 MB wired networks everywhere, and that there is power or power of Ethernet available for single point controllers.
Be ready for distributed control
As automation vendors release new systems based on single point controllers, ensure your engineers are trained in the IEC 61804-2 function block models and the ISA88 control modules and equipment modules models. These models will allow you to take advantage of the capabilities of truly distributed control using single point controllers.
– Dennis Brandl is president of BR&L Consulting in Cary, N.C., www.brlconsulting.com. His firm focuses on manufacturing IT. Edited by Mark T. Hoske, content manager, CFE Media, Control Engineering, firstname.lastname@example.org.
See other Brandl articles linked at the bottom.
See other Manufacturing IT articles.