Published in Interface. Vol 2, No. 2, May 1994

Alabama A&M University

by Scott von Laven

Computationally-oriented research at Alabama A&M University (AAMU) involves all five of AAMU's schools: Agricultural & Environmental Sciences, Arts & Sciences, Business, Education, and Engineering & Technology. The subjects under investigation range from aerodynamics to econometrics. AAMU faculty often collaborate with researchers from other universities, government labs, and industry. Graduate students and undergraduates usually participate. Dr. Jeanette Jones, Vice President for Research & Development, oversees this far-reaching and rapidly-growing research program.

A variety of computational techniques have been developed by AAMU materials scientists to predict the behavior of surface layers. Phenomena such as metallic embrittlement by hydrogen and surface melting have been investigated by Dr. Daryush Ila of the AAMU Physics Department and his associates. Their computations range from rough estimates of energy levels to full-scale molecular dynamics simulations. Most computations involve a large number of atoms and a very large number of interactions and thus require large amounts of supercomputer time. A wide array of smaller scale calculations related to optical materials is carried out at AAMU's Center of Excellence for Nonlinear Optics and Nonlinear Optical Materials. Other optics-related work includes modeling optical phenomena such as phase conjugation and harmonic generation.

Dr. H. J. Caulfield, Eminent Scholar in the Physics Department, has several related programs which use the C90. Most of them involve the very computationally-intensive task of global optimization of functions of multiple variables. Two popular approaches to this problem are both taken from the real world. Genetic algorithms evolve optimum solutions in a more-or-less Darwinian fashion; simulated annealing is based loosely on the annealing of metals. Caulfield has shown that both can be viewed as special cases of his concept of generalized exploratory optimization (GEO). GEO is being applied to several problems including (1) pattern recognition filter design, (2) optimization of category definitions for fuzzy control and fuzzy modeling/identification, and (3) design of wavelets for wavelet transforms.

Other traditional supercom-puter applications, fluid dynamics and structural analysis, are also pursued at AAMU, mainly in support of NASA. In the Civil Engineering Department, Dr. G.S. Liaw, Dr. Z.T. Deng, and Mr. K.L. Guo use conventional fluid codes for supersonic flow problems (Figure 1). The parameters of some of the fluid dynamics problems under investigation by this group lie outside the range of validity of conventional fluid codes. In the rarefied gas regime, for example, molecular interactions can be important. At AAMU, a particle technique called direct simulation Monte Carlo (DSMC) is used extensively for such modeling. Fluid properties can be obtained when needed by averaging over particles. Similarly, structural analysis is conducted outside the normal parameter regimes. Conventional (so-called h-version) finite-element structural-analysis codes deal with complicated topologies and material combinations by refining the computational mesh. Dr. Pabitra Saha and his associates in the Civil Engineering Department have developed their own p-version finite-element code which uses high-order polynomial elements (as high as ninth or tenth order) where high resolution is required. This has proven to be computationally more efficient than mesh refinement.

Computational research is also being conducted in the social sciences. In the School of Agricultural & Environmental Sciences, faculty are performing demographic analyses of rural social and economic trends. In the School of Business, faculty have begun studying how the international linking of economies is manifested in correlations between observable macroeconomic quantities. Statistical factor analysis and regression modeling are among the computational tools that will be employed.

The School of Education contributes to the long-term health of all computational research by introducing educators and future educators to the use of computers and to computational methods.

All AAMU students have ample access to computers. One of the dormitories is serving as a prototype for a combination residential/academic facility. Designated the Living/Learning Center, it will provide all its residents with access to campus mainframes and the Internet. Other networking projects currently under development by the Director of Academic Computing, Dr. John Vickers, and others at AAMU will help in meeting the ever-increasing demand for computational resources.