Machine tools not only think but readily communicate（沟通）. Those are the ambitious（有雄心的，野心勃勃的） goals of research initiatives（初衷） that not only promise a smarter factory, but are already delivering（交付，给与） smart maintenance.
Sept. 1, 2007 -- "Intelligent" and "smart" are words that machine tool makers increasingly use at product introduction time. These days, builders also seem to refer to(提到，谈及) "smart" features as routinely（例行公事） as their traditional claims（主张） of machine tool speed, accuracy and productivity.
So has the age of smart machines truly arrived? "Not yet," says John Kohls, executive vice president of Cincinnati''s TechSolve Inc., and one of the driving forces behind the industry consortium（社团，联盟） known as the Smart Machine Platform Initiative (SMPI). Instead he describes the industry''s piecemeal（粉碎） progress, smart characteristic by smart characteristic, towards that goal. As SMPI director his mission is to help coordinate（调整） and accelerate the evolution toward a new generation of machine tools. "While individual characteristics of the smart machine tool of the future are here now, the goal is to put the pieces together into equipment that can think," he says.
Kohls'' definition of machine thinking is more complex than simple sensor reaction to a machine condition. He adheres to(坚持) the original parameters outlined by the National Institute of Science and Technology (NIST), an early SMPI program participant.
In NIST''s vision, a smart machine:
·Knows its capabilities and condition with the ability to be interrogated（询问）,
·Knows how to machine a part in an optimal manner,
·Makes the first part (and every subsequent part) right,
·Monitors, diagnoses and optimizes itself,
·Knows the quality of its work, and
·Learns and improves over time.
Now in the second year of government funding, SMPI grew out of a series of workshops by NIST and the National Science Foundation -- a smart machine workshop in 2000 and a first-part-correct workshop in 2002. The targets: gains in productivity（生产力）, declines（下降） in inventory（存货，财产清单） requirements and manufacturing-related product improvements impacting（影响） price, quality and energy efficiency.
In outlining（轮廓） the SMPI''s goals, NIST reports that annual U.S. expenses（开支） on machining operations total more than $200 billion, about 2% of the U.S. Gross Domestic Product (GDP). NIST''s vision is for smart systems to complement and enhance the skills of machine operators, process planners and design engineers. By eliminating trial and error-based prototype（原型） development and reducing time to market, NIST sees U.S. builds a competitive(竞争的) proficiency（精通） in mass customization（专用华）.
Kohls says total funding for the four-year program is $12.5 million. The funding source: U.S. Army Research Laboratory with its Benét Laboratories in Watervliet, N.Y., serving as the program office.
The program is now in its technology evaluation phase. The test-bed: a four-axis horizontal machining center from Waconia, Minn.-based Milltronics Manufacturing Co. Kohls describes the evaluation platform as a standard machine tool with a GE Fanuc control. "What will make it smart are the technologies we are putting on it. Our intent is to use technologies that will enhance the ability to produce the first part correct."
Not Just Smart, But Also Good Listeners
Kohls expects some technologies to demonstrate synergistic benefits. "Combining feed and speed optimization with CAD/CAM, for instance, could result in a more beneficial depth of cut," he points out. "Machine intelligence could also help match tool life to the job. For example, machine operators could be forewarned if the remaining tool life would be shorter than the time needed on an upcoming cut. Machine intelligence can provide the ability to look ahead and prevent being caught with a failure-prone tool in the middle of a machining sequence."
Kohls says the goal is for machines that can predict and warn of conditions having the potential of affecting productivity, accuracy and quality. "Smart machine tools will be good listeners, too," adds Kohls. "They will react to abnormal process sound as experienced human operators would. The machine would communicate a warning and suggest ways of correcting the problem." For less critical situations detection and correction could be automatic, he notes. In addition to sound, intelligent machines could be made sensitive to force, vibration and temperature.
Thus far SMPI has not looked at vision from an in-process standpoint. "One difficulty with vision is the inability to see the cutter interface in action," Kohls says. Among the possibilities: inspecting tools and parts before and after a cut as well as visually inspecting chips.
"The biggest change that might occur would be the advent of model-based control where a model of the process drives the controller of the machine. We are planning a thorough investigation of that possibility," he adds.
Smarter and Communicative
The next step is leveraging smart machines with a communication standard that will enable a smarter factory. Kohls sees SMPI''s communication solution in MTConnect, the machine tool interoperability initiative announced in February by the Association for Manufacturing Technology (AMT), the machine tool trade group. Kohls already refers to AMT''s initiative as the foundation for SMPI.
MTConnect''s focus, says AMT''s Paul Warndorf, vice president, technology, is on providing a uniform, easy-to-implement way of confronting the typical production floor multi-vendor mix of equipment and software. "Instead of people spending all of their money on trying to figure out how to move the data, they can now pay attention to where their core competency is -- namely, how you use the data in the applications. What we want to do is move more toward the plug-and-play conventions of computer usage."
Adds Doug Wood, past AMT chairman and president of Parlec International Inc., "With computer equipment, multi-vendor implementations pose few if any connectivity or interoperability problems. However, anyone trying to install a multi-vendor implementation of production equipment and machine tools will find that it''s very difficult to get them to communicate and exchange information."
Wood says it''s time for the makers of machine tools and other production equipment to emulate the ability of computer equipment to be easily, quickly and seamlessly integrated. "The intent of MTConnect is to emulate that computer industry model. We want to remove the operator difficulties common to multi-vendor production environments." He notes that while the problem is minimized in single vendor environments, that approach has limited applicability.
"Unfortunately," notes Wood, "there is a tendency [for machine tool makers] to emphasize product differentiation in areas that jeopardize connectivity. Equipment users are then challenged to integrate multi-vendor environments." Even so, he does not see MTConnect removing the ability to differentiate via proprietary mechanical, electrical and hardware product approaches.
To resolve those difficulties, AMT''s board of directors authorized beginning MTConnect as a two-year $1 million effort. Initial work is being performed at the University of California at Berkeley. Warndorf says the goal is to develop a standard and demonstrate it at the International Manufacturing Technology Show (IMTS) in Chicago in September 2008.
Predict-And-Prevent
The concept of the smart factory may be most advanced in -- surprise -- maintenance. The evidence starts with the Center for Intelligent Maintenance Systems (IMS), a National Science Foundation, multi-campus university and industry initiative led by Jay Lee, a professor in Advanced Manufacturing at the University of Cincinnati. As director of the Center, Lee''s focus is on advanced prognostics and predictive maintenance technologies designed to achieve zero-breakdown productivity.
Lee says the challenge for reliability is in dealing with data from the past. "Failure is modeled, analyzed and -- to some extent -- predicted. Unfortunately, the prediction doesn''t take into account users or working environment-related constraints, and often the results aren''t that useful," says Lee.
"Condition-based maintenance (CBM) deals with online data. Machine conditions are constantly monitored and their signatures evaluated. However, this is done at the machine level -- one machine at a time. This is what we call the ‘fail-and-fix'' (FAF) mode." Lee says the challenge is to transfer to a "predict-and-prevent" (PAP) methodology.
"Today, CBM focuses on sensors and communications. All products and machines are networked by some means." Lee says that it is difficult to know, though, what to do with the data. "We need to turn data into information by using computational tools to process data locally," he adds.
The Center''s development of a Watchdog Agent is Lee''s solution for a more realistic, knowledge-based alternative to traditional practice of scheduled maintenance. Mounted on machine tools, the computer platform is a toolbox of more than 18 algorithms that translate sensor data, historical data and operating conditions into predictive data.
In addition to monitoring machine performance and health, the Watchdog Agent predicts the likelihood of process failure. Prognostics is the label Lee applies to this type of predictive maintenance. He describes prognostics as a paradigm shift to tomorrow''s predict-and-prevent maintenance thinking.
What distinguishes Lee''s characterization of prognostics is how maintenance intelligence is integrated at the machine, operations and enterprise levels. He notes that while the OnStar type of machine level data collection is now emerging in both production equipment and automobiles, degradation assessment using historical data is typically missing in both. That jeopardizes maintenance analysis. For example, Lee says the rate of change can be more significant than the degree of change.
Lee feels operations intelligence also needs further consideration. He''s referring to how the maintenance function should relate individual machine problems to overall production goals. "That involves how responses to problems are prioritized, optimized and scheduled."
The Center is also focusing attention on connecting the enterprise with synchronization intelligence. The idea, says Lee, is to facilitate the automatic conversion of data to information between the machine and business systems.
Lee says the Watchdog Agent has already been tested and installed at IMS partners General Motors, Harley-Davidson, TechSolve and Toyota''s Georgetown, Ky. plant. At Toyota, the Watchdog Agent was implemented in a compressor project, adds Lee. "The projects are demonstrating how the hardware and software platform can eliminate surprise shutdowns."
"The maintenance world of tomorrow is an information world for feature-based monitoring," says Lee. Intelligent sensor networks could let production equipment notice maintenance problems before they happen. "Information should represent a trend, not just a status. It should offer priorities, not just show ‘how much.'' If we do that, then our productivity can be focused on asset-level utilization, not just production rates."
The Center consists of research sites at the University of Cincinnati, the University of Michigan and the University of Missouri-Rolla, as well as two international sites and involves partnerships with over 40 global companies.
