Workforce Development | Young Innovators
Doctoral students leapfrog their way to optimization, innovation
By Upasana Manimegalai Sridhar and Anand Govindarajan
Upasana, the 2012-13 president of the ISA OSU student section, finds a combination of process control and optimization especially interesting. After earning her Ph.D., she would like to experience working in the process control industry and gain an industrial perspective on the best practices and techniques currently being employed. She eventually wants to make the switch to academe, to implement and improve process control and optimization curriculum in a way that will aid the students to gain more of an industry perspective in the field of controls and optimization. Upasana is aware of the benefits that ISA provides to members in terms of technical information, networking, and leadership opportunities. Having received the benefits herself, she believes it is a bound duty on her part to ensure that budding engineers of the future receive this too.
Anand strongly believes automation holds the key to the world's future and has farther-reaching implications than already envisaged. After earning his Ph.D. at OSU, he is looking at enriching his experience in the process control industry before moving on to an academic position. Through his experience in the industry, he wants to create a synergy between the process control industry and academe and develop engineers of tomorrow to think, learn, and also unlearn when necessary. In the near future, he would like to advise and mentor ISA student sections and enhance students' experience with ISA. This, he hopes, will create a chain reaction by encouraging students in turn to become ISA leaders of tomorrow.
We are doctoral students in chemical engineering, with a focus on advanced process control (APC) and optimization at Oklahoma State University (OSU).
With the guidance of our advisor, Dr. R. Russell Rhinehart, professor of chemical engineering, we are focused on developing a recently discovered Leapfrogging optimization technique (see footnote 1) and demonstrating its applicability in APC.
Leapfrogging (LF) is a multiplayer, direct search, optimization technique. Initially, "players" are placed at random spots in the feasible decision variable space. The approach to reaching the global optimum is for a player with the worst objective function to "leapfrog" over the player with the best objective function to a new position into the reflected hyper-volume. This optimization technique is stopped when there is no statistical improvement relative to the data - when all players converge to a common optimum.
In developing the LF optimization algorithm, we provide fundamental understanding, supporting the heuristically (experience-based) developed technique with mathematical truths and analyzing various aspects of the algorithm.
One of our studies has been focused on improving the initialization of the optimizer. We have found that initializing with a higher number of players and then cutting down to approximately 10 players per decision variable for optimization leads to the optimizer finding the global minimum with a lower computation burden and a higher confidence. Additionally, we are also working on improving the leap-over technique by modifying the leap-over distance to improve the speed of convergence of the algorithm.
Some of the other aspects we are exploring include applicability and efficiency to discontinuous/convex functions, the size and location of the leap-to region, and optimization stopping criteria. Our analysis should lead to improvements in the optimizer efficiency. As a test, we will apply the innovations to nonlinear multivariable model predictive control of pilot-scale heat exchanger and distillation processes.
The desirable properties of an optimization technique are that it reaches the global optimum with less computation burden and is robust and user-friendly. Our hope is that LF will benefit industry by providing faster convergence of APC and plant-wide optimization algorithms with a lower burden on computational resources - saving time, energy, and money.
Both of us have been involved with the ISA OSU student section since its inception in the fall of 2010. We have fulfilled multiple roles and have played a significant part in its success over the last two years. We strongly encourage students to make the best use of an ISA student section if it exists on campus. If not, you must consider starting one.
ABOUT THE AUTHORS
Upasana Manimegalai Sridhar, one of the founders of the ISA OSU Student Section is currently Section president, also volunteers as a secretary and fundraiser in the U.S. for Rural Education Welfare and Development Trust (REWARD). She completed her B.S. in chemical engineering from SSN College of Engineering (Anna University) in 2009, and her M.S. in chemical engineering at OSU in 2010. She is presently a Ph.D. candidate exploring optimization techniques for advanced process control and applications to a pilot-scale heat exchanger.
Anand Govindarajan, one of the founders of the OSU Automation Society, and former Section president, received a B.S. in chemical engineering from SSN College of Engineering (Anna University) in 2008 and his M.S. in chemical engineering at OSU in 2011. He is working toward a Ph.D. on advanced process control and optimization of a pilot-scale distillation unit.
1. Rhinehart, R. R., Su, M., and Manimegalai-Sridhar, U., Leapfrogging and synoptic Leapfrogging: A new optimization approach. Computers & Chemical Engineering, 2012. 40(0): pp. 67-81.