Global optimization of slider air bearing design

Hard disk drives continue to increase in areal data density. Densities as high as 1 Tbit/in2 are now being considered. The extremely high areal density requires air bearing sliders with ultra-low flying height (FH), less than 10nm. At this very low FH, slider air bearing surface (ABS) designs must satisfy very strict performance goals, such as uniform FHs and roll profiles across the disk. By using modern optimization techniques, it is possible to optimize slider ABS designs according to multiple design goals.; This dissertation focuses on the development and application of global optimization techniques to the problem of hard disk drive slider air bearing design. Both the stochastic global optimization techniques (Simulated Annealing algorithm family) and the deterministic global optimization techniques (DIRECT algorithm and its locally biased variations as well as some modified versions) are investigated and applied to the slider air bearing surface (ABS) design optimization problem.; We first give a detailed description of the Simulated Annealing family, including the Standard Simulated Annealing (BA) algorithm, the Fast Simulated Annealing (FA) algorithm and the more powerful Adaptive Simulated Annealing (ASA) algorithm. These Simulated Annealing algorithms are then applied to slider ABS optimization. These three main members of the simulated annealing family are shown to produce similar optimized ABS designs with greatly improved performance, i.e. uniform flying heights around the target flying height, flat rolls and improved stiffness. This illustrates that the simulated annealing algorithm is quite suitable for the optimization of ABS designs. Among them, the ASA was found to be the most efficient and robust scheme due to its fastest cooling schedule and its unique adaptive re-annealing mechanism.; An introduction of the new deterministic DIRECT algorithm is then presented through various numerical experiments and slider ABS optimization case studies. The comparison between ASA and DIRECT shows that DIRECT has a much faster convergence rate than ASA. Thus DIRECT can find the global minimum more quickly. It is shown that the DIRECT algorithm outperforms the ASA algorithm, so it is considered to be more suitable for the slider ABS optimization than ASA. Therefore, this dissertation focuses primarily on the DIRECT algorithm.; To further improve the efficiency of the DIRECT algorithm, we then propose three locally biased variations of the standard DIRECT algorithm. These variations generally have faster convergence rates than the standard DIRECT algorithm, and they may dramatically reduce the time needed to find the global minimum in some situations.; This dissertation also reports on two modifications to the DIRECT algorithm: one to handle tolerance (minimum side lengths) and one to deal with hidden constraints. The results show that defining the manufacturing tolerance and hidden constraints can save calculation time for a fixed number of designs generated, and thus improve the efficiency of the DIRECT algorithm.; To make the slider ABS optimization program more flexible, new geometric constraints are introduced. The slider ABS sensitivity optimization issue is also discussed.; Two new versions of the CML Air Bearing Optimization Program based on the Simulated Annealing algorithm and the DIRECT algorithm have been developed and they have been successfully applied to the ultra-low FH slider design and optimization problem for the Extremely High Density Recordings (EHDR) project of the National Storage Industry Consortium (NSIC).