By James R. Koelsch
, Automation World Contributing Writer, on August 7, 2015
Not just reserved for consumer gadgets, embedded systems make use of
advances in computing performance to extend the power of industrial
controls.
It's been 50 years since Gordon Moore first articulated his famous law about computing performance, yet his observation continues to withstand the test of time. The semiconductor industry’s ability to continually expand computing power into smaller spaces at lower prices has enabled incredible advances not only in consumer technologies, but industrial as well. Industrial automation is taking advantage of embedded technologies to consolidate electronics and extend the reach of processors to save money and boost productivity.
A case in point is the consolidation of the computing infrastructure inside the Glove Unique Reprocessing Unit (GURU) built by Pentamaster, an automation provider in Penang, Malaysia. The company has halved the number of industrial PCs driving the machine-vision modules in two of the unit’s seven workstations preparing used latex gloves for reuse.
The first vision module processes image data from four cameras in the automatic loading station to help position the gloves properly. The second vision module—in the fourth station, after chemical and thermal decontamination—also receives data from four cameras, but checks for cosmetic defects, reads the unique 2D code on each glove for traceability, and ensures that the glove is in the correct orientation for donning. The remaining three workstations find pinholes, decontaminate the gloves, and robotically package gloves that pass inspection and shred gloves that fail for recycling.
Pentamaster had found it necessary to upgrade the processors in the system to keep up with the evolving capabilities of machine vision. Executing complex algorithms on large data sets requires considerable processing power, and slow processing speeds restrict the number of high-resolution and high-frame-rate cameras that can be connected to an industrial PC.
To get the efficiencies that come with faster processing speeds, Pentamaster replaced the old processors in the vision modules with faster, multicore processors with integrated graphics processing. These third-generation Core i5 processors debuted Intel’s 3D tri-gate, 22 nm silicon architecture, running some important performance enhancements behind the scenes. Turbo Boost Technology 2.01, for example, adjusts processor speed automatically to match the required processing performance. Another example is the Hyper-Threading Technology that allows each processor core to work on two tasks simultaneously.
The enhancements reduced the inspection cycle to less than 2 seconds, boosting the inspection rate from 600 to 900 gloves/hr. The greater processing capability of the Core processor has also let Pentamaster consolidate the four industrial PCs supporting the vision modules to two. That, in turn, has simplified administration and maintenance, and has also reduced energy consumption and lowered the GURU’s operating costs.
Embedded remote diagnostics
The design team is already looking for other improvements, such as adding more testing routines to the set of vision processing algorithms already being used by GURU. Another avenue that the team is exploring is the Active Management Technology (AMT) designed into Intel’s vPro technology inside the Core family of microprocessors. This built-in remote management feature would allow Pentamaster to remotely manage, repair and protect the hardware in GURU’s computing infrastructure.
With this technology, operators or technicians can access outlying embedded systems remotely from their workstations and repair root-level software and firmware problems, including non-operational BIOS images and OS boot problems. To protect software and hardware that are known to be good, AMT allows supervisory users to install software remotely for making periodic updates, installing patches, and detecting and removing malware.
An important benefit of this technology is that a technician can diagnose problems remotely, even if the device containing the embedded system is not actually on, according to Shahram Mehraban, marketing director for Intel’s Industrial and Energy Solutions division. AMT can do this because it works below the operating system, a capability that differentiates it from conventional remote management software.
“Because conventional remote management typically relies on software that runs on an OS, the device must actually be running for someone to be able to remotely manage it,” Mehraban explains. “With AMT, even if the hard drive is corrupted and the device is turned off, you can still work below the OS to manage the device.”
This technology is useful in industrial PCs that run SCADA applications in manufacturing facilities and connect the factory floor to the company’s enterprise network and databases, Mehraban says. “We regularly have to do security and policy updates to various devices in the enterprise,” he says. “With this technology, we can extend this capability to perform them in our network of manufacturing facilities from a central location and administer all these patches at the same time.”
Although this remote technology is most commonly found on industrial PCs, it can be on any computing platform using Intel’s vPro processors. “A number of different applications on the factory floor today run on Intel silicon,” Mehraban notes. “Some high-end PLCs are based on Intel Core i5 and i7 platforms, and we have robotics and machine vision customers who are using our platforms.”
Supporting modular machine design
The ability to consolidate the computing systems within industrial equipment and add remote diagnostics under the operating system is not the only reason that embedded technology has become more attractive to industry in recent years. Another is that it supports a trend among machine builders to design and assemble their products from modular, off-the-shelf components. Because assembling equipment from pre-existing modules lowers design and assembly costs and shortens delivery times, modularity promotes customization and the building of increasingly more complex machines fitted with sophisticated technology.
“The complexity of machines is continuously increasing as end users push builders to increase productivity,” notes Sari Germanos, technology marketing manager for the Ethernet Powerlink Standardization Group. “Manufacturers are trying to do more with machines.”
Not only do they set more stringent internal timing and safety requirements for ever faster machines, but they also want these machines to consume less energy, support preemptive maintenance programs, and communicate with other machinery and inventory management systems, Germanos adds.
Embedded technology helps builders satisfy this demand by providing modular components like drives, inverters, sensors and HMIs with intelligence at relatively low cost. “The components are controlled and sequenced by a central PLC or industrial PC,” Germanos says. “They are connected together by one industrial Ethernet network, the architectural backbone of the machine.”
Intelligent components like drives rely on a mix of analog and digital application-specific integrated circuits (ASICs) and field-programmable gate arrays (FPGAs) to execute the motion profiles specified by the controlling PLC. In these cases, communications within the machine typically occur by means of industrial Ethernet hardware, advanced software protocols, and timing controlled by high-frequency FPGAs and ASICs.
The more notable advances in embedded technology tend to tighten integration within a machine. An important example is the integration of dual-core ARM processors with FPGAs from Altera and Xilinx. “These devices allow for very fast, low-latency computing within the FPGA hardware, and integrate complex software algorithms on the dual ARM core,” Germanos says.
This integration has allowed automation vendors to introduce highly integrated multi-axis drives at reasonable prices. “In turn, these drives allow machine builders to control several motors at once from one drive, making a machine more cost-efficient with better synchronization,” Germanos notes.
Tighter integration at the silicon level also allows component manufacturers to integrate electronic and mechanical functionality. Take for example an integrated motor like the AcoposMotor module from B&R Industrial Automation. “In this case, the traditional motor, drive, gearbox and encoder are integrated into one package, thus simplifying the logic and computational power required to control all of the components,” Germanos explains. “The PLC communicates with the bundle via a standard API [application programming interface] provided to the master controller in a standard XML format.”
Another example that Germanos offers of this tight integration is a mixed-signal processor from Analog Devices. “Here, an ARM-based processor can handle both analog and digital signals from the same piece of silicon,” he says. “This is ideal for motor control, and it provides a fast digital interface for the open source Powerlink Industrial Ethernet protocol.”
No FPGAs needed
Yet another form of integration rooted in silicon embeds real-time communication accelerators for standard network protocols like Ethernet Powerlink, EtherCAT and Profinet. A notable development here is the Industrial Communication Sub-System (ICSS) that Texas Instruments (TI) puts in its Sitara family of ARM Cortex-A series processors.
This embedded peripheral has helped developers eliminate either dedicated, fixed-function ASICs or FPGAs that they would otherwise have to use for embedding the protocols for linking to deterministic, extremely low-latency industrial networks. “Pairing an FPGA with any general-purpose processor is common for implementing some communication protocols or low-latency I/O expansion when those features aren’t available on the host processor,” notes Adrian Valenzuela, TI’s marketing director.
“Seeing this trend in industrial automation increasing, we’ve implemented on-chip, low-latency accelerators specifically designed for replacing FPGAs,” he continues. Eliminating this external device not only saves between $2 and $10, but it also reduces the complexity of the system and development time.
TI has put the ICSS into its ARM-based Sitara processors because of the popularity of the fast, low-power ARM processor. “A modern device can contain a dual-core 1.5 GHz Cortex-A15 yielding 10,500 DMIPS [Dhrystone million instructions per second],” Valenzuela says, noting that TI has a growing portfolio of ARM-based devices that will be released in the future in both 32- and 64-bit configurations.
Another benefit of having an embedded communications accelerator like ICSS is that it brings a measure of modularity to the implementation of communications protocols. Users and vendors of industrial automation strive to adhere to communications standards to promote safety, but often find that getting and maintaining certifications for these standards as they evolve can be costly and time-consuming. Even worse, the continuing effort can impede time to market.
ICSS solves this problem by encapsulating several protocols into a Lego-like module that can be pre-certified. “This allows developers to focus their development time on the application,” Valenzuela says.
Juggling operating systems
Embedded technology makes another contribution to integrating the intelligent components in a piece of equipment and consolidating computing resources. Real-time operating systems (RTOSs) running on multicore embedded processors can execute specialized graphical and textual machine-control languages based on the IEC 61131 standard for PLCs. These RTOSs should also comply with the IEC 61508 standards to offer the redundancy required for ensuring a safety integrity level (SIL) of 3, according to Germanos.
Here, virtualization seems to have found an industrial application in helping the limited computing resources typically found in embedded systems to run several concurrent instances of operating systems, including RTOSs. Vendors are developing virtualization schemes that address the concerns over latency and reliability that have made industrial users reluctant to embrace virtualization in the past.
“The first generation of embedded virtualization was very difficult to implement,” Intel’s Mehraban notes. “We’ve come a long way in the past two or three years though. More of our customers are building multicore-based solutions that use hardware-based virtualization acceleration.”
Emerson Process Management, for example, has based the virtualization in its DeltaV controllers on a Dell PowerEdge VRTX shared infrastructure platform. To avoid any latency problems, the system uses Intel’s fast Xenon processors on up to four computing nodes along with Intel Virtual Technology (VT). A software arbiter known as a hypervisor assigns hardware to each guest OS according to a default scheme or to the user-defined rules. Intel’s Advanced Programmable Interrupt Controller (APIC) technology also helps by offloading interrupt management from the hypervisor.
“Because virtualization can consolidate the workflows of multiple operating systems, it can consolidate things like an HMI, a soft PLC, and maybe a motion controller on a single platform,” Mehraban says. “Consolidation reduces the number of devices you have to maintain and service, and it increases the overall reliability and uptime.” In many cases, it can also reduce cabling and other installation costs.
Besides consolidating platforms, virtualization can contribute to network security by creating a partition to isolate the core of the system software. On the other side of the partition, the system software acts like a guest firewall that communicates with the outside world. Only specific information can cross the partition.
Source : - http://www.automationworld.com/embedded-control/embedded-control-reaching-more
