ELECTRICAL AND COMPUTER ENGINEERING

Engineering new efficiency into a tried-and-true system

Annette Muetze

Electrical and Computer Engineering Assistant Professor Annette Muetze (15K JPG)

WHENEVER ELECTRICAL CURRENT is employed to create motion, chances are a traction motor is doing the work. In fact, these workhorse systems for converting electricity to motion or motion to energy are so common, even a small gain in their efficiency can produce dramatic energy savings and environmental benefits. Industry continues to find new uses for traction motors and is demanding better performance from them under a wider range of operating conditions.

With a five-year, $399,000 grant from the National Science Foundation in hand, Assistant Professor Annette Muetze will apply modern optimization and design techniques to coax greater reliability and efficiency from traction motors. The grant is part of Muetze's Faculty Early Career Development Award (CAREER) from NSF. Among the most prestigious given to faculty who are just beginning their academic careers, CAREER awards are granted to creative projects that effectively integrate research and education.

Muetze will apply modern mathematical optimization techniques to finite element analysis-based design of electromechanical energy converters. Her modeling and problem formulation will exploit convex characteristics that play an important role in global optimization. Muetze will use these techniques to create streamlined, reliable and highly efficient designs, bringing forth a significant savings related to sub-optimal energy consumption of traction motors.

In addition, Muetze will integrate these modern design techniques into on-campus and off-campus power-engineering education at both the undergraduate and graduate levels. Her methodology will include the development of web-based learning tools, literature-based learning and small-group-based projects.

Quantum dots assemble to mimic membrane pores in living cells

WHILE STUDYING the pores in biological membranes by labeling them with tiny specks of crystal, a team of engineers witnessed a startling result: The specks assembled to form their own artificial pores. The discovery could lead to an entirely new level of manipulation, imaging and understanding of the inner workings of cells.

The specks, known as inorganic semiconductor nanocrystals or quantum dots, measure in millionths of a millimeter. Professors Dan van der Weide and Robert Blick, along with researchers Sujatha Ramachandran and George Kumar, found that by applying voltage to a solution containing both quantum dots and membranes similar to those of living cells, they could press the dots into the membranes. The dots formed rings, which in turn acted as membrane pores.

Natural channels in cell membranes regulate the flow of charged atoms, or ions, in and out of the cell — a process that controls everything from the twitching of muscle fibers to the firing of neurons. Using the artificial pores, the engineers showed they could enhance the movement of ions across the membrane and control their flow from outside by applying voltage. Because the artificial channels elicit bursts of current in the artificial membranes, the team believes quantum dots could perform similarly in cells such as neurons and muscles. They look forward to understanding how the dots behave in vivo in these excitable cells, and will next investigate the properties that cause the artificial pores to open and close.

New mathematics ask, “Do I know you?”

RECOGNIZING A FAMILIAR FACE — something humans do with ease — is what scientists now hope to accomplish with image processing and mathematics. Like the human brain, the face recognition system being developed by Professors Nigel Boston and Yu Hen Hu, and Computer Sciences Professor Chuck Dyer, focuses on facial features that remain unchanged as the face moves, such as a cheek's curve or a nose's slope. Such systems could one day spot wanted individuals in a host of settings, from terrorists walking through airports to serial cheaters casing casinos.

Based on the classical mathematical theory of invariants, the method calculates the mathematical factors that stay the same when a curve or surface is rotated. The resulting graphical fingerprint or signature — called the invariant — can be used to identify the curve or surface at any orientation.

Past mathematicians have used differential equations to calculate invariants, an approach that is very sensitive to noise. The UW-Madison team has instead invented a technique employing integrals. To test its ability to recognize two-dimensional shapes, the researchers obtained the contours of 100 distinct fish species from a database, generated 20 randomly oriented variations of each contour, and fuzzed the outlines to introduce noise. The integral invariant approach proved able to match the variations to the original contours with 98 percent accuracy, outperforming popular competing techniques. With the mathematics in hand, the team is now devising a system for face recognition in three dimensions.

>> Engineering physics >>