Graduate research analyzes optimization of facility design
PhD student Ketan Date’s research is developing new ways for companies to design their facilities and increase efficiency.
Since beginning his graduate research at the Department of ISE in 2013, Date has applied his mathematical background to solving real-world problems.
His research has focused specifically on figuring out how to optimally design layouts of manufacturing facilities. Facilities typically house several machines, but they need to be placed in an efficient manner to optimize production.
Date developed a mathematical model to design efficient layouts that can be used by companies that are moving to new facilities due to increased demand or productivity.
When moving to a new location, companies typically want to redesign the layout of their facilities. So Date considered the way that people work with machines, and the way machines work with other machines, when thinking about layout design.
“If you are taking too much time going from one machine to the other, it costs you something. It’s not a direct cost, but it takes up your time,” he says. “If two machines are communicating more, then they should be placed close to each other.”
Date developed a mathematical model that considers these constraints and comes up with the most efficient design for the layout of new facilities.
But another problem arises when this model is used in situations that involve a large number of machines.
“You have to have a lot of computing power for solving even a modest-sized problem,” Date says of the model.
So he began to figure out a way to use parallel computing to solve problems faster. In parallel computing, multiple processors divide the work among each other and work simultaneously to solve a big problem faster.
In other research, this method has been carried out by using a computer’s central processing unit, or CPU, which carries out the instructions of a computer program.
Date instead decided to use the graphics processing unit, or GPU, which is typically used in gaming applications, to develop faster solutions.
GPUs have more cores — specific units in the processors that perform computations — than CPUs.
“Basically you can divide your work among a large number of chunks, and a single graphics card is able to solve a problem faster than a single CPU” Date says.
This problem has been studied before, but not in a way that specifically dealt with the finite-size facility placement problem that Date studied.
“That’s why I wanted to explore the area and see if we can develop a general theory for placing machines in facilities, plus we wanted to see if we can be efficient in solving those problems,” he says.
His research is highly applicable to today’s industries because manufacturing companies typically redesign the layout of their facilities every two to three years.
Date says he enjoys studying problems like these because they involve both mathematical theory and applications to real-world problems.
He hopes to continue to work in the overlap between applied work and theoretical mathematics, and his goal is to become a faculty member at a research university.
“Being an academic is like the only dream job that I can have, because I love to teach students,” Date says. “I love to guide them and help them out in their problems, and I enjoy research as well.”