Modified TEM cell design using mixed simulated annealing-deterministic optimization

A global optimization algorithm for TEM cell design is presented. It merges features from the simulated annealing algorithm and from the Hooke and Jeeves' (1961) pattern search method. The proposed method derives from the former algorithm the ability not to be trapped in local minima, while from the latter it derives a more efficient way to search in the cost function domain. Cost functions take into account field uniformity in large cell regions, and fields are computed by means of refined numerical conformal mapping procedures. Suitable aspect ratios are introduced in the cost function to control the dimensions of the cells.