Latin Hypercube Sampling and Pattern Search in Magnetic Field Optimization Problems

Latin hypercube is a sampling technique for searching n dimensional space. Like Monte Carlo methods, it retains random qualities, and yet Latin hypercube is consistently more effective than Monte Carlo. Despite this fact, not a single paper has been published in IEEE Magnetics on its use. Field analysis is a long way from delivering vectorized solutions where a vector of inputs can be processed. Stochastic algorithms are exceptionally inefficient compared to their deterministic counterparts. The best optimization tool would be a deterministic method which quickly and effectively interrogates the search space. Latin hypercube sampling, combined with pattern search solutions, comes close to achieving that objective. This concept and an improved solution for the magnetic TEAM workshop problem 22 using these tools were presented by Kent Davey at the 16th International Conference on the Computation of Electromagnetic Fields (Compumag 2007) in Aachen , Germany , June 24-28, 2007. For further information, please contact Kent Davey .

Representation of a two dimensional function with Latin hypercube and Monte Carlo representations.

Development and Analysis of Trapped Field Magnets in Electromechanical Devices

High temperature superconducting trapped field magnets (TFMs) offer great potential as an alternative to 2nd generation YBCO wire, both in cost and performance. Attention is given to the calculation of current distribution within YBCO disks at partial and full activation and comparing this to experimental values. The best results are obtained by treating the current as a sequence of nested current rings. The fields are computed by integrating the elliptic integrals representing the fields from these rings and using variable metric optimization to choose the ring radii to best match the activation field over the un-activated material. A technique for treating the sub-regions of the TFM as voltage fed coils appears most expeditious for computing forces. These concepts were presented by Kent Davey at the 16th International Conference on the Computation of Electromagnetic Fields (Compumag 2007) in Aachen , Germany , June 24-28, 2007 .For further information, please contact Kent Davey

Current density pattern determined numerically using a 1 T activation field and Jc =400 MA/m2 .

Optimization of Gas Turbine Generator-Sets for Improved Power Density

Many future US Navy ships will employ all-electric propulsion systems instead of mechanical drives. To help optimize performance of these systems, studies are under way at The University of Texas at Austin Center for Electromechanics (UT-CEM) to minimize the size of power generation components. These studies focus on increasing the power density of directly-coupled gas turbines and generators (gen-sets). One approach adopted by UT-CEM uses scaling laws of gas turbines and synchronous electrical generators to examine the possibility of increasing power density by operating at higher shaft speeds. Included is consideration of inlet and exhaust turbine ducts and issues involving power electronics. Study results indicate that if inlet and exhaust duct volumes are neglected, the power density of directly coupled gas turbine-generator sets can be significantly improved by scaling to higher operating speeds. However, advantages of scaling to higher speeds are largely negated when duct volumes typically encountered on modern ships are included. This suggests locating power generation equipment near the ambient terminus of inlet and exhaust ducts, so that duct lengths are minimized and it is possible to fully exploit the power density advantages of scaling to higher shaft speeds.

Gas Turbine scaling results