Purdue, industry partners creating ’intelligent’ grinding process

Researchers at Purdue University are working with industry to develop an “intelligent” system that could save U.S. companies $1 billion annually in manufacturing costs by improving precision-grinding processes for parts production.

“Precision grinding is currently an art that relies heavily on the experience and knowledge of employees who have been in the business for years,” said Yung Shin, a professor of mechanical engineering who is leading the Purdue portion of the research. “The problem is that many factories don’t have enough of these very experienced people, so a lot of grinding processes are run under suboptimal conditions.

“Our system strives to enable relatively inexperienced employees to operate grinding machinery with the same precision as these rare, highly experienced workers.”

The “intelligent optimization and control grinding processes” use artificial-intelligence software, which mimics how people think, in order to learn and adapt to changing conditions.

Shin has been working on the method for 15 years. He will present an overview of his work on May 12, during the Automation & Assembly Summit, May 10-12, at the American Airlines Training Facility in Fort Worth, Texas. The conference was organized by the Society of Manufacturing Engineers.

“We estimate that if this method is fully implemented in the United States, we could save about 10 percent of the cost of current grinding practices, which is a really conservative estimate,” Shin said. “That adds up to about $1 billion per year in the U.S.”

TechSolve Inc., in Cincinnati, is leading the team of industrial partners in a three-year, $6 million project funded through the National Institute of Standards and Technology’s Advanced Technology Program.

“Precision grinding is becoming increasingly important for the automotive, aerospace, medical-device and electronics industries,” said Anil Srivastava, manager of manufacturing technology at TechSolve. “Grinding is often the final machining process for creating parts that require smooth surfaces and extremely fine tolerances.”

Other industrial members of the team are Delphi Energy & Chassis Systems in Dayton, Ohio; Applied Grinding Technologies in Wixom, Mich.; and Landis Gardner in Waynesboro, Pa.

If successful, the process would save Delphi millions of dollars annually by increasing productivity, saving energy, reducing the number of grinding wheels needed, reducing scrap and improving the overall quality of parts, said David Yen, manager of advanced manufacturing engineering at Delphi.

“Going from the lab to real-world applications won’t be easy and will require a lot of hard work and diligence,” Yen said. “By the end of the three-year time span, we will identify several pilot applications, all in automotive areas, and validate the methodology, and then we will extend the technology to other grinding applications.”

The intelligent system will use a wealth of data collected by various sensors, as a given part is being ground. Then the method will apply advanced software, such as neural networks and genetic algorithms, to operate specialized “computer numerical control” grinding machines that cost up to $1 million apiece.

The machines, commonly referred to as CNC machines, are widely used in industry and are increasingly being equipped with sensors that provide information about the grinding process in real time. The machines use grinding wheels containing ceramic or diamond particles to apply a fine-finish surface to precision parts, and sometimes they are used to create a part from scratch.

“Ceramic parts, for example, cannot be machined, so they are created entirely with grinding,” Shin said.

The sensors collect information about such details as forces exerted on bearings, speed, vibration and temperatures during various parts of the process.

“A lot of machines are now coming out with these sensors,” Shin said. “The question is, ’what do you do with all of that information?’

“We capture that information in the software to establish a database that will be used to set the machine to optimal operating conditions.”

Shin has demonstrated that his method works in small-scale applications, but he said it will be a challenge for the team to apply it on a large-scale industrial basis.

“It is high risk because we are going from the lab to full-scale industrial systems,” he said. “That sort of endeavor is always difficult because the magnitude of complexity in industry is much greater than in the lab.”

Writer: Emil Venere, (765) 494-4709, venere@purdue.edu
Sources: Yung C. Shin, (765) 494-9775, shin@ecn.purdue.edu
Anil Srivastava (513)948-2004, srivastava@techsolve.org
David Yen, (937)455-9259, david.w.yen@delphi.com
NIST Contact, Michael Baum, (301) 975-2763, michael.baum@nist.gov
Purdue News Service: (765) 494-2096; purduenews@purdue.edu