
A New Architecture for Automatic Gauge Control Systems
David R. Wisti, Alfred F. Krzewki, Craig K. Godwin and Mark E. Zipf
This is an overview of the newly developed system architecture and the partitioning of the real-time control and signal processing responsibilities among the hardware and software components of the chosen commercially available equipment platform. Discussions concerning the developmental motivations and the evolution of the development process are provided.
1 – Introduction and Historic Perspective
In the cold rolling of flat strip products, the fundamental objective of an Automatic Gauge Control (AGC) system is to regulate the
rolled material thickness about a referenced set-point. The AGC performs this task by the coordinated actuation of the roll gap,
separating force, strip tension and/or mill speed. The theories and practices involved in AGC algorithms and roll force cylinder
control methods are well understood, and not the subject of this discussion. In this paper, the primary focus is on newly developed
technologies in the underlying hardware equipment and software techniques that allow AGC actions to be implemented and carried
out in real-time.
Over the past 25+ years, I2S, LLC, has supplied over 260 AGC systems worldwide, starting with one of the first microprocessor controlled AGC systems in 1980. Over the years, we have seen hardware/software technologies and their availabilities change radically, followed closely by the evolving needs of our customers; tighter tolerances, increased reporting, and expanded interconnections between the mill’s systems, including a wide range of plant-wide customer provided computer systems. Through the years, a great many hardware/software platforms and technologies have been employed to implement AGC systems, ranging from commercially available equipment to highly specialized, custom/proprietary arrangements. Early systems employed various combinations of discrete analog/digital components and early microprocessor/computer control technologies (e.g., Z80, PDP- 8, etc.). In that era, the lack of commercially available, sufficiently capable hardware was a formidable obstacle. This led some to implement their systems using a combination of available commercial equipment, special proprietary fabricated boards and in some cases, custom, hand-made, wire wrapped cards. These measures were a necessary fact-of-life, given that the available general purpose interface boards were not suited for this class of application, with their associated performance requirements and feature set capabilities.
AGC systems evolved with the prevailing technologies and soon embraced the emerging VME, Multi-Bus and IBM-PC compatible computer technologies, advanced multi-tasking operating systems, graphical user interfaces (GUIs) and a variety of more sophisticated general purpose interface and communication hardware offerings. Unfortunately, much of the commercial hardware still did not fully encompass the specific performance and feature requirements of a complete AGC system. Even so, it was very possible to utilizes these commercial building blocks to field AGC systems able to meet the broader tolerances of that era. To achieve higher levels of performance, suppliers of high-end systems had to, yet again, bridge the performance/capabilities gap with custom/proprietary boards, some employing embedded microprocessors to off-load the computer system’s real-time computational overhead.
Although an acceptable choice to performance minded end-users, custom/proprietary hardware required those same end users to invest in and willingly commit to a potentially perilous de facto marriage with the system supplier, who offered an inconvenient, single source pathway to spares and repairs. Yes, the resulting systems were highly optimized for AGC applications and tended to achieve the highest degree of performance. However and unfortunately, fears of the manufacture’s potential business failure loomed large in the end-user’s view of these seemingly precarious relationships. The system’s only spare/repair life-line hung tenuously on the thread of whether the equipment provider would be in existence (into the perceivable future) and would truly be able to support this custom hardware over the system’s lifetime.
General purpose, Commercial-Off-The-Shelf (COTS) technologies soon became a fixture and an ever present mantra in the market place, suggesting that end-users could free themselves from these de facto marriages, and openly purchase spares and repairs from other suppliers and vendors, thereby circumventing the OEM and liberating the end-user from concerns over the long term viability of the AGC system provider. This was an enticing proposition…but one that had to be weighed against the fundamental objective of the high end user (i.e., high performance AGC capabilities).
Early COTS equipment was developed for, and directed towards, a broad spectrum, high volume market, whose needs and capabilities may or may not have been suitably aligned with the requirements of high performance AGC applications. An AGC system’s requirement for specialized interfacing and signal handling capabilities were still not common among the needs of the general purpose market place. This led COTS manufactures to bypass these niche AGC requirements. AGC suppliers were left to decide whether their product lines would be fashioned to address the high performance needs of their narrowly intended application (and thereby consciously continue to employ certain amounts of proprietary equipment) or would consciously accept potential performance degradations to embrace a more “market attractive” COTS based system.
Interestingly enough, the rapid pace of commercial hardware development, and quick product end-of-life obsolescence, became a mainstay in the unstable computer and electronics industries. It was not uncommon for PC and VME interface boards to be obsolete and unsupported within 12-18 months of their initial unveiling. This caused many COTS based AGC system offerings to have the appearance of a patchwork quilt, of ever changing hardware components and supporting software drivers, with every new system incarnation being incompatible with effectively identical systems provided a year or two earlier. In addition, many qualified and seemingly stable hardware/software platforms were abruptly discontinued (e.g., Multi-Bus), leaving those utilizing these platforms in their AGC system with little hope for extended support.
So much for the alluring and enticing “Sirens’ Call” of the COTS philosophy… These computer industry facts-of-life led many equipment providers to still develop and employ dedicated, proprietary hardware in certain specific facets of their systems. We have been among them. This is not always a bad thing, time has shown that many proprietary systems have yielded a longer (and more stable) effective product lifetime and a higher potential to be supported for extended periods of time (i.e., greater than 5-10 years and upwards of 20 years, in some cases). However, the attractive philosophy of COTS based systems is still a mainstay in the marketing ploys of those suppliers choosing to accept and contend with the “ebb and flow” and uncontrolled whims of the commercial hardware manufactures. And, to be sure, COTS is not only a desire by the end user, there are advantages to the AGC supplier as well. COTS can aid in reducing cost and development time by utilizing the component manufactures expertise and capabilities.
Recent advances in COTS based controllers have now reached levels of capability to be suitably employed in high performance AGC systems. This has become a fashionable trend and a seemingly viable marketing strategy. Commercial motion/numerical control modules and high speed counting systems are often combined within commercial, general purpose Programmable Logic Controllers (PLCs). These open architecture systems utilize specialized application programming of the general purpose motion control module hardware to obtain the specialized control structures required in AGC applications. One downside of using PLC based controllers for AGC is that it locks the end user into the specific make/model of PLC (and the contained specialty modules) that the AGC supplier provides irrespective of the customer’s plantwide PLC requirements. Modern COTS based systems are constrained by the underlying hardware and software architectures developed by their commercial manufactures. These systems are focused on a broad, general purpose, application based market. The underlying issue in all current COTS based systems is that their general purpose hardware arrangements must be suitably reorganized, and their general purpose data flow architectures must be selectively programmed, to meet the fundamental requirements of the computational processing involved in AGC activities. This can lead to a “painted-into-a-corner” architecture that does not lead itself to future development and a great dependence on a limit number of components (which the COTS supplier may change or discontinue at any time).
2 – The Future, Here and Now
So, where does this leave us? Customers don’t like custom parts available from only a single source supplier. Suppliers are not fond
of inflexible, commercial system architectures that limit adaptability and constrain the design. What other options are there?
As noted previously, over the years, I2S has sold/installed/maintained hundreds of AGC systems in both domestic and international
settings. We owe it to our customers and ourselves to find and develop systems that are maintainable and can be systematically
upgraded to satisfy needs of the metals industry. The bottom line is that the chosen hardware platform must provide the necessary
capabilities and performance. The chosen software platform must provide the required real-time responsiveness, while also
supporting the chosen Human Machine Interface (HMI) and Internationally Standardized network interfaces.
In the past few years, a new COTS based alternative has presented itself. One that does not constrain the resulting system with a given
manufacture’s preconceived views of what hardware arrangements and data flow architectures will be provided or not provided. But
one that offers a general purpose, programmable hardware environment, through which software based control, signal processing and
data flows can be implemented with ease. This same COTS based alternative supports hard, real-time control and operating system
software to assure deterministic responses to real world events and conditions. Further, all standardized network interfaces are
supported, allowing a broad spectrum of commercial HMI software packages (often selected/directed by the customer) to be
employed.
This is very welcome news.
This new programmable hardware technology provides the ability to implement application optimized hardware with the flexibility of programming. This technology is Field Programmable Gate Arrays (FPGA). FPGA’s are programmable hardware. Integrated, single chip circuits containing millions of programmable gates that can be configured and interconnected for any need. When the code is compiled and downloaded to the target FPGA, the result is a true hardware implementation of the software application. The software configured / programmed hardware operates in a truly parallel nature which leads to extremely fast execution rates (parallel execution with 40 MHz clock speeds). Figure 1 provides an illustration of the internal components and architecture of a typical FPGA integrated circuit.
The approach presented here employs commercially available programmable hardware to specifically implement optimized signal processing and control architectures. This allows focus specifically on the issues and requirements inherent to AGC activities. There is no need to distort and contort general purpose commercial hardware (e.g., PLC, motion control equipment, etc.) to provide the required AGC capabilities. Formerly proprietary hardware and software are now replaced with programmed, general purpose commercial hardware, programmed via a commercial, open-architecture development language (i.e., National Instruments LabView). A great advantage to this approach is that the system is not restricted to an unyielding, single hardware system architecture. With very little setup time, this same system can be implemented as a PCI bus based computer solution or as a distributed, modular, selfcontained, compact remote I/O system (as shown in Figure 1).
The chosen programmable hardware platform coexists with a real-time controller, allowing dual/parallel processing activities to be
used. Using recently released multi-core CPUs (e.g., Intel® CoreTM2 Duo Processors) system overhead is greatly reduced. This means
sophisticated control algorithms can be used without worrying that the system will become overloaded.
Key features of the FPGA/Multi-core CPU based AGC are:
3 – System Architecture and Implementation
Figure 2 provides a block diagram illustration of the I2S FPGA AGC architecture. Figure 3 provides a hierarchical view of this same
system. The I2S AGC system is composed of two primary components:
I2S AGC Controller – This Intel® CoreTM2 Duo Processor based computer provides the real-time AGC functions, roll force
cylinder servo controls, and high frequency data acquisition for SPC analysis, reporting, and general data acquisition (often
coupled to IBA data logging systems). Real-time activities are performed by the National Instruments Real-Time LabView
ETS software system (executing on a PharLap Real-Time Operating System supporting symmetric multiprocessing for the
Core 2 Duo processor). This controller interfaces to the other control system computers (PLC, HMI, Drive System) via a
Gigabit class Ethernet network interface. Precision digital wrap counting is provided from the counting of the quadrature
pulse trains of the winder motors’ pulse tachometers / rotary encoders. The controller is housed in a hardened, industrial, 19
inch rack mounted chassis, incorporating a single board Core 2 Duo computer, passive PCI bus backplane and FPGA boards.
I2S AGC Human Machine Interface Computer (HMI) - This Windows-based PC compatible computer provides interactive,
graphical user interface screens that support the I2S AGC Controller and the HMI software operations. The user interface
typically consists of a high resolution, flat panel video monitor (often mounted on/in the Main Operator Desk), a keyboard
and pointing device (i.e., mouse, trackball, etc. – depending on customer preferences). The HMI is programmed with
commercially available (off-the-shelf) software packages (e.g., WonderWare, Intellution, Cimplicity, WinCC, RS View, etc.)
and is open to customer modification and adjustment. HMI data exchanges with the PLC and other control systems (AGC
Controller, Drive System, Gauging System, AFC, etc.) is provided through a networked interface via Internationally
Standardized OPC Server executing within the HMI computer. All AGC parameters, set-up, calibration, tuning, monitoring,
scheduling and utilities are orchestrated through the screens of this interface. All mill management (pass scheduling, coil
set-up, roll cluster/stack utilities, etc.) and quality recording, analysis and reporting (SPC/QC) are performed by this
computer and the screens supporting these functions. Fault and permissive annunciation are provided by screens supporting
these functions. Report and document printing is provided via the networked color printer. Depending on the customer
selected HMI package’s capabilities, multiple language support is typically offered.
5 – Conclusion
The history and prevailing issues associated with the supply and sustainable support of the hardware and software components of high
performance control systems have been discussed. Optimized, proprietary hardware may yield the best performance, but these
desirable technical advantages must be consciously weighed against the realities of a single source supply. The attraction of COTS
based technologies must not be allowed to conceal the very real possibility that these general purpose commercial products may not be
truly suited for the application at hand and may become unexpectedly modified or discontinued by manufactures focused on a broader
market place.
These issues form a classical paradox. The application required performance characteristics are best accommodated by specially designed, custom hardware and software systems, but their single sources of supply are unattractive. Commercially available equipment provides a desirable pathway of supply, spare and repair, but its general purpose designs do not necessarily align themselves with the application at hand.
End users and suppliers find themselves in an awkward position and must make a difficult decision. The key issue has been the lack of commercially available hardware that is appropriately optimized for high performance AGC applications. That is until now…
A new technology has emerged and offers the ability to implement the desired, optimized attributes of customized, proprietary hardware, in the form of commercial, openly available, programmable hardware. COTS based FPGA hardware systems are now commonplace and these programmable systems provide virtual hardware implementations that solve the issues of optimized hardware and commercial availability.
The new I2S FPGA AGC system, described in this paper, has been conceived and developed with the end user and these difficult
issues in mind. The first of these FPGA based AGC systems have been successfully installed and are now fully operational on the
following mill configurations: