Many-core technology fills in the blanks for advanced machine control
When many-core technology became available for industrial machinery, engineers had questions. “What can you do with all of that processing power?” some wanted to know. “When would you even need it?” others asked. At the time, running a programmable logic controller (PLC) program in PC-based automation software required one core. Even with HMI and a few extra programs, industrial servers with dual 16-core processors, for example, appeared excessive. There were blanks strategically built in for additional future programs, but engineers weren’t sure how to fill them in and with what. However, production machinery is no longer what it once was.
Constant advances in automation technology (AT), combined with the greater convergence of AT, operations technology (OT) and information technology (IT), have created more efficient, reliable and complex machines. The data acquisition and responsiveness necessary for smart factory and Industry 4.0 concepts have also led to significant changes.
Systems that once used a few PLCs, stepper motors and a basic fieldbus, for example, have received major updates in motion control with robotics and mechatronic linear transport systems (LTSs), EtherCAT communication, machine vision systems, operator interfaces (OIs) with voice commands, mobile human-machine interfaces (HMIs) and, recently, machine learning (ML). These, along with many emerging technologies, continue to fill in the blanks and justify the integration of many-core technology today.
PLCs and programmable automation controllers (PACs) have not kept pace with the resulting massive influx of data. Multi-vendor, distributed control architectures have not always proven effective due to the “handshakes” required to make the systems work together. Advanced machines require advanced controls. PC-based control has proven its capabilities for many years, but these abilities have grown through the introduction many-core CPUs.
Today, multi-core industrial PCs (IPCs) still meet the majority of machine control requirements, but the rapid increase in requirements and the opportunity to gain competitive advantages make a convincing case to explore many-core options for upgrades and future machine designs.
What are many-core IPCs?
The key difference between many-core and multi-core control is not so much the number of processor cores as it is the actual processor structure. Many-core builds on high-performance computing (HPC) principles by using embedded processors optimized for greater parallelism and throughput. Parallel data stream processing on a large scale means lower power consumption for concurrent completion of tasks due to the tasks’ spatial layout. Many-core also relies on enhanced thread synchronization to resolve data bottleneck issues seen in most low-range CPUs.
In most applications, multi-core technology can simultaneously carry out numerous complex tasks with ease when paired with suitable automation software for standard machine control logic and advanced functions. Many-core CPUs are engineered to extend this ability to the most taxing applications with the same high scalability and flexibility. As a result, many-core control principles could extend to a range of devices from DIN rail-mountable embedded PCs with quad-core processors as easily as they do to industrial servers with dual 20-core Intel Xeon boards and beyond. No matter the size, a key strength of the technology is the use of PC-based automation software for core isolation.
Advanced control applications: How many cores?
IPC software with core isolation allows engineers to dedicate specific tasks to individual cores or clusters in software. The processor’s memory affinity leads to faster processing times, with task data cached in specific locations for higher performance. Demanding programs, such as integrated ML or real-time simulation with Matlab/Simulink from The MathWorks, can take up multiple cores located near each other and run concurrently with similar tasks.
This also is true for advanced motion control architectures, such as LTSs and planar motor systems with levitating movers, which require dedicated neural networks. Multiple cores also could be required for sophisticated analytics and oscilloscope software, especially with the quantity of data available through Gigabit Ethernet and 10 Gbit/s communication speeds.
IPC selection also depends on the number of tasks and systems supported as well as the cores available, rather than the highest clock speed. Durability also is a concern for production environments. For production environments, durability also is a concern. It is important, then, to choose vendors that provide scalable products in rugged form factors.
On the lower range of many-core controllers, some vendors supply PC-based controllers in standard, DIN rail-mountable form factors. Some embedded PCs, for example, offer four to 12 2.2 GHz processors, 8 to 64 GB DDR4 RAM and operating temperature ranges of -25 to 50°C. On the higher end, a few industrial servers boast dual processors featuring six to 20 cores, with clock rates varying based on core count. These can offer hard drive capacities from 240 GB SSD up to 4 TB, 1,024 GB of DDR4 RAM and a working range of 0 to 50°C. Scalability is very important in these cases; not every application calls for 40 cores of processing power, of course, but a reasonable number could require more than four cores.
Workload consolidation for smart factories
The benefits of centralized control systems are within reach with many-core IPCs. Many-core machine controllers create a multitasking device by consolidating all tasks while limiting hardware, minimizing footprint and increasing overall performance. This is a large improvement from previous systems that divided processes between various PLCs, motion controllers and network PCs, which created communication delays.
While the IPCs can also connect to the cloud, their storage capacity and ability to run numerous programs on the device make the controllers more self-sufficient, benefitting manufacturers and machine builder original equipment manufacturers (OEMs) across many industries. Some OEMs may choose to develop their intellectual property to handle advanced machine learning and artificial intelligence (AI) by running their own proprietary software on the many-core device. Manufacturers may also be wary of the cloud if, for example, their machines process volatile compounds. Even without an internet connection, engineers have access to a more efficient platform for implementing Industry 4.0 and smart factory concepts.
The automation software these controllers use has a crucial impact on overall performance gains and capabilities. With multi- and many-core architectures, OEMs and manufacturers can face many new challenges because the PC-based controllers evolved and expanded capabilities over time. Software that is as advanced should be tested and proven to adapt to these challenges. When selecting systems, engineers should ensure software and hardware show the results of years of experience in the field and preparation specifically for use in many-core controls.
The processing power contemporary machine architectures require today has caught some vendors off guard. However, many advanced machines and systems show the value of many-core technology.
Keywords: industrial PCs, many-core technology, Industry 4.0
Many-core technology options can help companies keep up with upgrades and future machine designs for industrial PCs (IPCs).
Many-core CPUs are engineered to extend this ability to the most taxing applications with the same high scalability and flexibility
Many-core PCs can consolidate operations and streamline data management.
What benefits could your company receive from many-core technology and where would it help the most?
For more information: