Innovative Collaboration Between LTU and Eaton Uses AI For Prescriptive Maintenance

There is no idea more fundamental to the study of engineering than that of maintenance, which has aptly been called “the price we pay for living in a dynamic world.” LTU, in collaboration with Eaton Corp. and supporting the U.S. Army Corps' Engineer Research and Development Center (ERDC) under Contract No. W912HZ22C0022, is exploring a new approach to prescriptive maintenance: a quality control based artificial intelligence algorithm that can predict when machines are going to fail. This cutting-edge research keeps LTU at the forefront of learning institutions committed to supplying industry partners with bold and innovative solutions to real-world challenges.

To appreciate its significance and potential, it is necessary to understand the evolution of maintenance as a concept and a practice. Reactive maintenance—“You wait until it breaks and then you fix it”—was the first commonly accepted maintenance paradigm. This gave way to planned maintenance, changing something after a specific period of time according to a preconceived timetable. Predictive maintenance was next. It demanded that specialists examine data to predict failure and anticipate needs. The last step in this conceptual evolution is prescriptive maintenance: trying to prevent failure, recommending steps to forestall decline, and extending the life of the part.

Truly effective prescriptive maintenance requires having parts already sourced and personnel ready so that maintenance does not impede production. Traditionally, prescriptive maintenance requires a significant amount of data, gathered by expensive sensors and subjected to the analysis of experts with extensive maintenance records at their disposal. It is time and labor-intensive.

Dr. Giscard Kfoury, associate professor in the A. Leon Linton Department of Mechanical, Robotics, and Industrial Engineering, explains the bold new approach being undertaken by LTU and Eaton: “What we’re trying to do is reduce that amount of work and data and ideally use what we call quality control data to perform prescriptive maintenance.”

On the cutting edge of quality control, cameras armed with AI software are examining parts for potential defects, eliminating the need for human involvement in that part of the process and precluding any duplication of effort. “Using that data for prescriptive maintenance, to see if we can correlate if a part that we’re producing is starting to have defects, does that reflect on the machinery, and maybe we have to replace this part in the machine so that it doesn’t fail. That way, the data is already there, the equipment is already there, and we’re not spending more money on equipment or trying to get new data. We’re using that data to do the main prescriptive maintenance part,” Kfoury explains.

The first phase of the project involved faculty, students, and Eaton personnel looking at and planning for two applications: injection molding and PCB manufacturing. Phase Two is scheduled to occur in January of next year and will involve the actual implementation of prescriptive maintenance for injection molding at one of the Eaton manufacturing facilities and validation in a factory setting for PCBs.

Dr. George Pappas, assistant professor in the electrical and computer engineering department, is proud of the work that has been done. “This is a truly collaborative project between multi-discipline departments: electrical and computer engineering, mechanical engineering, mathematics, and computer science.”

He explains that when experienced workers, who are able to rely on their experience and intuition, leave or retire, there is an immediate discontinuity in quality control. In light of this, cognitive analytics and machine learning have become invaluable assets. “The key thing is live prediction. Real-time, live-data prediction that actually makes a big difference. The key is to know in real time when something is not right,” says Dr. Pappas.

Through this partnership, LTU and Eaton are creating a framework that can be used by any industry. The potential benefits are manifest and staggering. On a cost-efficiency basis, this new approach will be profoundly advantageous, reducing the cost of traditional maintenance and circumventing the concomitant loss in productivity.

Dr. Kfoury reflects on the exciting and amazingly beneficial opportunity for students involved in this project, who are able to see fundamental research being put into action almost immediately with experiments in the lab and implementation in a real-life setting.

“It doesn’t get any more ‘theory and practice’ than this.”

This material is based upon work supported by the Engineer Research and Development Center (ERDC) under Contract No. W912HZ22C0022. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the ERDC.

by Joe Bedard