SBCs provide the key to Industry 4.0

The popular image of the industrial shopfloor is of conveyors passing components and subsystems down a fixed production line. Since the days of Henry Ford, it has been the metaphor for high productivity in manufacturing. However, it is one that is notoriously inflexible. The static nature of the fixed production line makes it hard to customise products and to reschedule operations to produce different products as demand changes. Industry 4.0 breaks down the production line, providing the opportunity to make manufacturing a far more flexible, cost-effective and sustainable operation than has ever been possible before.

There are several key components to Industry 4.0. One is to replace the single fixed production line with cells that can be dynamically reconfigured down to the level of individual orders. It may use automated guided vehicles (AGVs), robots, conveyor segments and automated manipulators to move components and subsystems around the factory to where they are next needed. To facilitate this smarter approach, each product is tagged. Staff, machine tools and robots use that information associated with the tag to determine what needs to happen next on the product’s journey through the facility. Even the cells themselves, which could consist of several machine tools and robots, may be virtual in that they can be dynamically assigned to other cells based on demand.

Secondly, machine tools interact with each other using the ubiquitous connectivity associated with an Industry 4.0 plant. The systems employ a combination of wired and wireless networking technologies to exchange information in real time with each other and with distant cloud servers. Those remote servers can use their powerful computation capabilities and Artificial Intelligence (AI) to schedule the shopfloor systems in smarter ways.

Finally, sensors track the status of everything on the shopfloor. RFID readers pick up information from each product using its associated tag. Vision and chemical sensors check the progress and quality of the product at each step so that repairs can be made before it is too late as well as to guide maintenance. If the surface finish of a product is found to be moving out of specification, for example, it indicates a potential problem with the upstream machine tools or the raw materials. By recognising these situations quickly, the Industry 4.0 factory avoids waste and the costs it incurs. The result is a manufacturing environment where products can be customised down to the per-unit level based on customer demand and one that can quickly switch schedules if a key source material is in short supply or if the customer-ordering systems indicate a change in ordering behaviour.

A new generation of single-board computers are now available that support industrial use; developed for and increasing adopted to smarten manufacturing facilities and support the move to Industry 4.0. Although Industry 4.0 introduces novel ways to restructure the factory, manufacturers do not have to replace everything, and SBCs can be adopted to drive new benefits in a range of ways.

Couple SBC-based computing with industrial networking

Manufacturers can leverage existing investments in the machine tools and production automation they already have to deliver significant benefits. Many of the machine tools already in use can be adapted for the Industry 4.0 environment as long as they are augmented with additional levels of communication and intelligence which can be done economically, and effectively with SBCs. Indeed, there may be no need to replace the programmable logic controllers (PLCs) that provide the instructions needed to carry out the operations that each machine tool provides.

Many fieldbus protocols have been adapted to work over industrial Ethernet to ensure commands can be relayed to and from a nearby compute module. One possibility for integrating this functionality into both existing and new machine-tool panels lies in the Kunbus range of DIN-rail compatible Revolution Pi modules, an open-source industrial PC based on Raspberry Pi. These combine the compute power of an Arm Cortex-A processor with Ethernet connectivity plus expansion for sensor feedback through a range of I/O modules and fieldbus interfaces.

Image processing and machine learning for quality control

Industrial Shields, a leading European manufacturer of industrial automation devices, uses the Raspberry Pi platform to provide another option for implementing smarter, more flexible PLCs. One application for this generation of higher-performance PLC lies in the coordination of movements of subsystems and materials around the factory. A USB connection can provide the interface to a barcode reader or RFID scanner that picks up the tag on an incoming pallet or product carrier. A display connected using HDMI can be used to provide confirmation to an operator assigned to check its operation. When the package is confirmed, the Raspberry Pi-based PLC uses industrial networking and I/O connections to activate motors to move the package through a series of conveyors to its destination. Alternatively, it may communicate a route to an AGV that picks up the product and delivers it. When the first cell has processed the product, the PLC can then act to guide it to its next destination or pass control to a motion-control PLC that is closer.

A key advantage of using hardware such as the Raspberry Pi is its future upgrade path. Many of the existing industrial-control solutions are based on the third-generation module but products are now being built around the latest iteration of the hardware: the Computer Module 4. The increased processing power of the module, which is based on a quad-core Arm Cortex-A72 processor attached to up to 8GB of high-speed DRAM and 32GB of eMMC non-volatile storage.

The level of performance available in the Computer Module 4 can support intensive machine-learning and vision-processing applications. In addition, as the Computer Module 4 runs Linux, the many tools and development environments on that platform (such as Tensorflow, PyTorch and OpenCV) provide easy access to highly sophisticated techniques for analysing components and subsystems to check they meet quality standards. Subtle changes in colour or surface composition that AI can identify means alerts can be sent to upstream supervisory systems to take corrective action.

The supervisory systems can also harness the processing power of Intel’s ecosystem. Intel’s NUC family includes models that scale up in cost-effective performance to processors such as the i7-8665U, a quad-core device that can run at several gigahertz. The NUC boards and systems are highly suited for use as supervisory systems. Multichannel video connections provide the ability to run several displays at once. NUC-based computers can therefore provide a high degree of local intelligence, reacting to alerts generated by PLCs and other SBCs on the shopfloor and sharing graphical updates with shopfloor staff so they can see if problems are building up that require their attention.

Equipment monitoring and information analysis

At the other end of the scale, flexible processes need responsive, easily programmable low-level control. This can be delivered, for example, by the Arduino platform, a combination of microcontroller-based hardware and an optimised software-development environment that supports rapid prototyping and algorithm evaluation. The Arduino Pro Portenta provides a low-cost but powerful option by using a dual-core pairing of the Arm Cortex-M7F and M4F processor cores, both of which support integer and floating-point arithmetic. This makes the Arduino Pro Portenta suitable for the execution of mathematical models and closed-loop control algorithms.

For greater performance in a compact package, the DFRobot LattePanda couples an Arduino-compatible microcontroller with an Intel quad-core 1.8GHz processor able to run Windows 10. Using this combination, the SBC can perform tasks such as AI-assisted equipment monitoring as well as image processing and computer numerical control (CNC), making it highly suited to building customised machine tools.

BeagleBone AI provides a further option for adding support for machine learning, smart sensor and image processing in real time. By using a variety of sensor modalities, it can provide access to non-destructive testing in real time coupled with equipment monitoring. The onboard dual-core Arm Cortex-A15 running at 1.5 GHz works with a pair of TI C66 digital signal processors and four Embedded Vision Engines with support for TI’s deep-learning software.

Conclusion

The path to Industry 4.0 for systems integrators, machine-tool builders and factory owners is becoming easier to navigate, supported by a rich and growing collection of SBCs. These cost-effective performance SBCs, and the development environments that accompany them, enable existing tools to be upgraded and incorporated into a network alongside new robotics and manufacturing systems in a seamless way. The result will be a far smarter factory.

Romain Soreau is head of single board computing at Farnell

https://uk.farnell.com/